Advanced planning...

��

��

�� !" #$�%&#%'&(#)

�� *�+��,�� -�� .� ��!" #$�%&#%$�/#&

��-� 01��,�� 2�� .� ��!" #$�%&#%&% #(

�� ,3�� '� ��!" #$�%&#%&�/#.

�� !� ��,�� ,�� 4��5�� '� ��!" #$�%&#%�(&#.

*� *� �� !��$� ��!" #$�%&#�%%.#/

�� 2�6�� "�# �� $�� $� ��!" #$�%&#�%. #'

4� 27� �� !�� %&�� $� ��!" #$�%&#�%./#(

3� �� !�� '��&� ��!" #$�%&#�%.$#�

8*9 4��: ; <�� !��3 �� + �� (�� !" #$�%&#��'/#

�� $��)�� *�� !" #$�%&#��&%#$

�� !�� !" #$�%&#�( �#/

8� !=�� ,*� *�>,�� %�� (��!�� !" #$�%&#�� #�

�� !��,?� ��,�� (%%%� ��!" #$�%&#�(&(#)

�� :�� !�� (%%�� !" #$�%&#� ''#.

�� 2�6�� "�# �� (%%(� ��!" #$�%&#�.&&#�

9� 0� �� + ��+�� #�� ,��#�(%%(� ��!" #$�%&#�/�'#%

!� ��66,�� @��A�� ,�� -...(%% � ��!" #$�%&#�/(&#.

�� 3 ��%�� (%% � ��!" #$�%&#%% #

*�� (%% � ��!" #$�%&#%%/�#�

?� �� $�� $��(%%.� ��!" #$�%&#%%�&#&

@� 8��6� �� ,3� �B��,3� *�� A��,�� ?�� %�� "��#��(%%.� ��!" #$�%&#%(%(#'

4� @��7��/�� (%%.� ��!" #$�%&#%(&%#&

�� @� ?�� /�� %��(%%/� ��!" #$�%&#�///#�

�� A�$�� (%%/� ��!" #$�%&#�/$&#%

�� -=�A� �� 6

��

�� + -+��

8�� ' �� ?��

��C��#D� �� 6 �� 6��C

�� 8� �� =� �� !��

��

��" �. �#��.��!"#�% #$�%&#�/�(#' ��C��#D� �� *�� "�7 E� A��!"#� �$&# #$�%&#�/�(#. ��C��#D� �� *�� "�7 E� A

��#�#�� 0�� 66�� +� -� � C �+ �� "�� F (%%/�($�/(

?�� 7� A � ��G�� 6C �� H 7�� 7�� 6� � �+�� H �6��+��C �� + � ��H �6 ��H �� + ��#� ��H ��H � ��H �6 �� +� � � ��C �� 7�CH �� A�� 0�6�� + �� 6�� 6� �� + � 6� �� C�� 6 �� + �� 3� �� 6C �� -�7 �+ ��6�� H ��'/H � �� H �� 6� �� +� �� 7�C� �� + �� C��#D� �� D�� +� 6 �� 3� �� 6C �� -�7�

��C��#D� �� 6� � �+ �6 �� I!��

�6 ��

J ��C��#D� �� *�� (%%/� �� 3� ��C

?�� + �� 6�� H �� H � �� A�H �� 6�� 6CH �� + � �6��+� ��H �� >��6� + �� 6 �� 7� �� +� � + �� +� ��

��+�� F � �� 4 �� H *��

��" ��.� . &&, �/ #/ . ( � % ; � �� #+ �� #�� 6�6�

$��K-=�A�K�� 6-�� L�K$��%��K!� ��A�M��

Foreword

During the last three decades trade, industry and also academia became heavily involved in the development primarily intended towards more effective planning and control of logistical operations in supply chains. Lately, these approaches be-gan to be directed towards fresh food supply chains. Competitive fresh food sup-ply chains require that the aspects costs, quality, food safety and technology be taken into account simultaneously in a multidisciplinary way. In recent years the issue of food safety got large emphasis in government, industry and society (con-sumers). The introduction of the General Food Law by the EU from January 2005 on even put more emphasis on the issue of food safety.

It turns out that Advanced Planning and Scheduling Systems (APS) can play an important and integrative role in supporting decision making activities in fresh food supply chains by considering shelf life as an instrument to generate more added value and food safety. Basically the work of Matthias Lütke Entrup is con-centrated on two research questions:

Which requirements must APS systems cover in order to efficiently and effectively support production planning in fresh food industries? How can shelf life be integrated into production planning? How can production planning contribute to optimizing shelf life output?

In his study the author shows how these questions should be answered ade-quately. His results and conclusions are of paramount importance for integrating the issue of shelf life into production planning.

The study provides a wealth of insights and results which are significant both from a practical as well as from an academic point of view. The research starts with an overview of current APS systems and highlights the need of a new genera-tion of planning software which aims at supporting decision making in supply chain management. Although APS gain increasing acceptance in industry, a num-ber of issues remain, in particular at the detailed planning and scheduling level, which are not satisfactorily covered by the decision models to be found in the standard APS software packages. This is truly the case for the fresh food indus-tries. Undoubtedly, the most important planning issue regarding fresh food lies in the consideration of shelf-life. So far, vendors of APS systems have taken many efforts to consider shelf-life issues in their planning systems, however, without covering all of the characteristics being important in Fresh Food Supply Chains (FFSCs) and fresh food production systems. One of the main contributions of the study by Matthias Lütke Entrup is a comprehensive analysis of the planning re-quirements of fresh food industries on one hand and the decision support offered

VI Foreword

by typical APS systems on the other. Software packages from leading players in the market are assessed looking at the scope of shelf life integration and its capa-bilities to generate plans that optimize shelf life output.

Based on the shortcomings of current APS systems, new quantitative planning models are developed and resolved. These models consider shelf life planning problems in specific fresh food industries (yogurt production, sausage production and poultry processing). The models are based on the general block planning prin-ciple and are adapted to the needs of the specific fresh food planning applications. Considerable care has been taken to obtain compact model formulations which can be solved very efficiently by use of standard optimization software. Numerical experiments demonstrate the applicability of the planning models in realistic in-dustrial settings.

As a result, the author makes clear that suppliers of APS software are currently unable to offer APS systems in which the integration of shelf life into production planning has been dealt with adequately. Specifically, product freshness has been modeled by the author as part of the optimization and not as a constraint within the planning function. This is indeed a new and creative contribution of Matthias Lütke Entrup to solving complex planning problems of considerable practical relevance. The applications (case studies) have been selected carefully by the au-thor in such a way that many other application fields in fresh food industries could benefit from his results.

Prof. Dr. Paul van Beek Prof. Dr. Hans-Otto Günther

Acknowledgement

This research could not have been written without the support of many people. Therefore, I would like to thank a number of them for their support and contribu-tions, knowing that the list is, of course, incomplete.

First of all, I am indebted to my academic advisors Professor Dr. Hans-Otto Günther of the Chair of Production Management at the Technical University of Berlin and Professor Dr. Paul van Beek of the Operational Research and Logistics Group at the Wageningen University (NL). Professor Dr. Hans-Otto Günther woke my interest in the field of Production Management and helped me to trans-form my ideas into a full research project. Similarly, I am thankful to Professor Dr. Paul van Beek for his supervision of the work and his critical comments. Working with both of them was a pleasure, they have always been accessible and created a stimulating research environment. Additionally, I thank Professor Dr. Kasperzak for assuming the chairmanship of the promotion committee.

I would also like to thank the entire team of the Chair of Production Manage-ment consisting of Hanni Just, Dr. Martin Grunow, Matthias Lehmann, Ulf Neu-haus, Martin Schleusener, and Onur Yilmaz for their helpfulness and the fruitful discussions. Their comments proved to be very useful and resulted in several im-provements. In addition, I am grateful to Thorben Seiler and Shuo Zhang for their support regarding the development and implementation of the models.

Furthermore, I thank my employer A.T. Kearney for the possibility to conduct this research and the continual support. In particular, I highly appreciate the con-tributions of Dr. Antje Völker, Jan van der Oord and Ferdinand Salehi as well as of Dr. Peter Pfeiffer and all other colleagues of the Consumer Industries and Retail Practice. Dr. Marianne Denk-Helmold and Judith Siefers deserve a special thanks for carefully reading and correcting the manuscript.

The last words are dedicated to my family. I thank my parents for their encour-agement and their trust in me during all the years. Finally, I thank Kathrin for her backing and her care. She made me realize that there are other things in life than yogurt, sausages and poultry.

May 2005 M. Lütke Entrup

Table of Contents

Foreword............................................................................................................... V

Acknowledgement ............................................................................................. VII

Abbreviations....................................................................................................XIII

1 Introduction........................................................................................................ 1

1.1 Introduction to the Field of Research .......................................................... 11.2 Research Objectives .................................................................................... 21.3 Dissertation Outline..................................................................................... 31.4 Conclusion................................................................................................... 4

2 Advanced Planning and Scheduling Systems................................................... 5

2.1 Evolutionary Path of APS Systems ............................................................. 52.1.1 MRP I and MRP II ............................................................................... 52.1.2 Assessment of the MRP Planning Concepts ........................................ 82.1.3 Emergence of APS Systems................................................................. 9

2.2 Structure of APS Systems.......................................................................... 122.2.1 Overview............................................................................................ 122.2.2 Strategic Network Design .................................................................. 142.2.3 Demand Planning............................................................................... 152.2.4 Supply Network Planning .................................................................. 172.2.5 Production Planning ........................................................................... 182.2.6 Production Scheduling ....................................................................... 192.2.7 Distribution Planning ......................................................................... 202.2.8 Transport Planning ............................................................................. 212.2.9 Available-to-Promise ......................................................................... 21

2.3 APS Systems Market Overview ................................................................ 232.3.1 Available Market Studies ................................................................... 232.3.2 Market Size and Segments ................................................................. 242.3.3 Major Providers.................................................................................. 252.3.4 Expectations for the Future ................................................................ 27

2.4 Implementation of APS Systems ............................................................... 272.4.1 Implementation Process Overview..................................................... 272.4.2 Project Definition ............................................................................... 282.4.3 Vendor Selection................................................................................ 302.4.4 Implementation .................................................................................. 31

X Table of Contents

2.4.5 Implementation Risks ........................................................................ 322.5 Assessment of APS Implementations........................................................ 33

2.5.1 Benefits .............................................................................................. 332.5.2 Development Needs ........................................................................... 34

2.6 Conclusion................................................................................................. 35

3 Fresh Food Industries ...................................................................................... 37

3.1 Introduction ............................................................................................... 373.2 Definition and Segments ........................................................................... 373.3 Characteristics of Fresh Food Supply Chains............................................ 38

3.3.1 Structures of Fresh Food Supply Chains............................................ 383.3.2 Economic Characteristics and Developments .................................... 413.3.3 Technological Characteristics and Developments ............................. 473.3.4 Social/Legal Characteristics and Developments ................................ 503.3.5 Environmental Characteristics and Developments............................. 533.3.6 Summary............................................................................................ 57

3.4 Characteristics of Fresh Food Production Systems ................................... 583.4.1 Overview............................................................................................ 583.4.2 Formulation........................................................................................ 593.4.3 Processing .......................................................................................... 603.4.4 Packaging........................................................................................... 613.4.5 Storage and Delivery.......................................................................... 623.4.6 Summary............................................................................................ 63

3.5 Case Study 1: Yogurt Production .............................................................. 643.5.1 Market Segments and Case Study Overview ..................................... 643.5.2 Raw Milk Collection.......................................................................... 673.5.3 Raw Milk Preparation ........................................................................ 693.5.4 Fermentation ...................................................................................... 703.5.5 Flavoring and Packaging.................................................................... 713.5.6 Storage and Delivery.......................................................................... 72

3.6 Case Study 2: Sausage Production ............................................................ 723.6.1 Market Segments and Case Study Overview ..................................... 723.6.2 Input of Ingredients............................................................................ 753.6.3 Grinding and Mixing.......................................................................... 763.6.4 Chopping and Emulsifying ................................................................ 763.6.5 Stuffing and Tying ............................................................................. 763.6.6 Scalding ............................................................................................. 773.6.7 Maturing and Intermediate Storage.................................................... 783.6.8 Slicing and Packaging........................................................................ 783.6.9 Storage and Delivery.......................................................................... 79

3.7 Case Study 3: Poultry Processing.............................................................. 803.7.1 Market Segments and Case Study Overview ..................................... 803.7.2 Transport of Animals ......................................................................... 823.7.3 Stunning and Bleeding ....................................................................... 833.7.4 Scalding and Eviscerating.................................................................. 843.7.5 Chilling .............................................................................................. 84

Table of Contents XI

3.7.6 Rough Cutting .................................................................................... 853.7.7 Fine Cutting........................................................................................ 863.7.8 Packaging ........................................................................................... 863.7.9 Storage and Delivery.......................................................................... 87

3.8 Conclusion................................................................................................. 87

4 The Fresh Food Industry’s Profile Regarding APS Systems........................ 89

4.1 Methodological Remarks........................................................................... 894.2 General Requirements ............................................................................... 904.3 Requirements for Strategic Network Design ............................................. 934.4 Requirements for Demand Planning.......................................................... 954.5 Requirements for Supply Network Planning ........................................... 1004.6 Requirements for Purchasing & Materials Requirements Planning ........ 1014.7 Requirements for Production Planning and Production Scheduling........ 1034.8 Requirements for Distribution Planning .................................................. 1094.9 Requirements for Transport Planning...................................................... 1114.10 Requirements for Demand Fulfilment and Available-to-Promise ......... 1144.11 Conclusion............................................................................................. 116

5 Shelf Life in Fresh Food Industries .............................................................. 117

5.1 Shelf Life of Food Products..................................................................... 1175.1.1 Definition and Limiting Factors....................................................... 1175.1.2 Determination of Shelf Life ............................................................. 1195.1.3 Technological Shelf Life Extensions ............................................... 120

5.2 Shelf Life Characteristics of Case Study Products .................................. 1215.2.1 Case Study 1: Shelf Life of Yogurt .................................................. 1215.2.2 Case Study 2: Shelf Life of Sausages............................................... 1225.2.3 Case Study 3: Shelf Life of Fresh Poultry........................................ 123

5.3 Shelf Life in Fresh Food Supply Chain Management.............................. 1255.3.1 Literature Review............................................................................. 1255.3.2 Role of Shelf Life in Fresh Food Supply Chains ............................. 127

5.4 Conclusion............................................................................................... 128

6 Shelf Life Integration in APS-Systems ......................................................... 131

6.1 Introduction ............................................................................................. 1316.2 SAP APO................................................................................................. 131

6.2.1 System Overview ............................................................................. 1316.2.2 Shelf Life Integration ....................................................................... 134

6.3 PeopleSoft EnterpriseOne........................................................................ 1376.3.1 System Overview ............................................................................. 1376.3.2 Shelf Life Integration ....................................................................... 139

6.4 CSB-System ............................................................................................ 1406.4.1 System Overview ............................................................................. 1406.4.2 Shelf Life Integration ....................................................................... 143

6.5 Summary and Conclusion........................................................................ 143

XII Table of Contents

7 Shelf Life Integration in Yogurt Production ............................................... 147

7.1 Problem Demarcation and Modeling Approach ...................................... 1477.2 Model Formulations ................................................................................ 152

7.2.1 Model 1: Model with Day Bounds................................................... 1527.2.2 Model 2: Model with Set-up Conservation ...................................... 1597.2.3 Model 3: Position Based Model....................................................... 163

7.3 Computational Results............................................................................. 1717.3.1 Simultaneous Optimization of All Lines.......................................... 1717.3.2 Line Decomposition Approach ........................................................ 1737.3.3 Model Combination and “Pick-the-Best” Approach........................ 174

7.4 Conclusion............................................................................................... 177

8 Shelf Life Integration in Sausage Production.............................................. 179

8.1 Problem Demarcation and Modeling Approach ...................................... 1798.2 Model Formulation.................................................................................. 1838.3 Computational Results............................................................................. 1918.4 Conclusion............................................................................................... 195

9 Shelf Life Integration in Poultry Processing................................................ 197

9.1 Problem Demarcation and Modeling Approach ...................................... 1979.2 Model Formulation.................................................................................. 2009.3 Computational Results............................................................................. 2069.4 Conclusion............................................................................................... 209

10 Conclusions and Recommendations ........................................................... 211

10.1 Summary of Results .............................................................................. 21110.2 Discussion ............................................................................................. 21310.3 Recommendations for Further Research ............................................... 215

References.......................................................................................................... 217

Abbreviations

3PL Third Party Logistics Provider APO Advanced Planner and Optimizer APS Advanced Planning and Scheduling ATP Available-to-Promise BBD Best-Before Date BOM Bill of Materials BSE Bovine Spongiform Encephalopathy CAGR Compound Annual Growth Rate CIP Clean-in-Place CPFR Collaborative Planning, Forecasting and Replenishment CPG Consumer Packaged Goods CRM Customer Relationship Management CTP Capable-to-Promise DC Distribution Center DisP Distribution Planning DP Demand Planning EAN European Article Number ECR Efficient Consumer Response EDI Electronic Data Interchange EDIFACT Electronic Data Interchange for Administration, Commerce and

TransportELSP Economic Lot Scheduling Problem EPC Electronic Product Code ERP Enterprise Resource Planning FFSC Fresh Food Supply Chain GMP Good Manufacturing Practice HACCP Hazard Analysis Critical Control Point IFS International Food Standard ISO International Organization for Standardization IT Information Technology KPI Key Performance Indicator LP Linear Programming MDB Model with Day Bounds MILP Mixed Integer Linear Programming MPS Master Production Schedule MSC Model with Set-up Conservation MTO Make-to-Order

XIV Abbreviations

MTS Make-to-Stock OR Operations Research OV Objective Value P&MRP Purchasing & Materials Requirements Planning PBM Position-Based Model PP Production Planning PP/DS Production Planning / Detailed Scheduling PS Production Scheduling QAS Quality Assurance System RFID Radio Frequency Identification Tag ROI Return on Investment SC Supply Chain SCE Supply Chain Execution SCM Supply Chain Management SCP Supply Chain Planning SKU Stock Keeping Unit SND Supply Network Design SNP Supply Network Planning SRM Supplier Relationship Management TP Transport Planning TP/VS Transport Planning / Vehicle Scheduling VMI Vendor Managed Inventory WWRE World Wide Retail Exchange XML eXtensible Markup Language

1 Introduction

1.1 Introduction to the Field of Research

With an approximate turnover of € 100 bn., the food processing industry is one of the major sectors of the German economy. Ca. 50% of this number is generated by fresh food industries such as the meat, dairy, fish, fruit, vegetables, or bakery in-dustry (Lebensmittel Zeitung 2001). Due to factors such as high variability of raw materials, intermediate and final products, fluctuating prices, or variable process-ing times and yields, production planning in fresh food industries is generally a challenging task.

In this environment, Advanced Planning and Scheduling (APS) systems can constitute significant means of support for the planner. Driven by developments in Supply Chain Management (SCM) and Information Technology (IT), APS sys-tems are a shift of paradigm in production planning since they address material re-strictions and capacity constraints simultaneously and not successively as Enter-prise Resource Planning (ERP) systems implemented today by most companies. Hence, APS systems help to avoid high amounts of work-in-progress, to increase service levels, and to shorten planning times. Moreover, APS systems allow opti-mizing the entire supply network by integrating several production sites, distribu-tion centers, suppliers and customers into one planning model. However, imple-mentation numbers of APS systems in fresh food industries remain rather low, because many important requirements of these industries are not yet sufficiently covered.

One of the most distinctive factors to consider in fresh food production plan-ning is the limited shelf life of the products. Shelf life restrictions directly influ-ence scrap rates, out-of-stock rates in the retail outlets and inventory levels. Fur-thermore, consumers tend to buy the product that has the longest possible shelf life. Being able to offer a longer shelf life than their competitors constitutes a piv-otal competitive advantage for food producers. Hence, the provision of shelf life functions is crucial for APS systems in order to succeed in the fresh food industry. Yet, only a few authors have considered the integration of shelf life into produc-tion planning (see Chapter 5.3.1).

2 1 Introduction

1.2 Research Objectives

Therefore, the research will focus on two main research questions:

Research question 1:

Which requirements must APS systems cover in order to efficiently and effectively support production planning in fresh food industries?

The scientific outcome of the first part of the thesis is a profile of three sample fresh food industries (yogurt, sausages and fresh poultry) with regard to APS sys-tems. These three case study industries cover the most important fresh food seg-ments (dairy, processed and fresh meat). In addition, within each of the case study industries, the product with the most challenging production environment has been chosen. For each of the modules of an APS system, the fresh food specific re-quirements are analyzed and their importance for each of the three sample indus-tries is assessed based on a rating score. This list of requirements constitutes an important support for companies operating in fresh food industries. On the one hand, the functional specifications for the planning systems can be defined more easily and efficiently. In addition, the list can also be used to evaluate the capabili-ties of APS software and to decide which software to implement. From a scientific point of view, the structured approach to developing the list of requirements can be used as a guideline for other industries. In literature, fresh food industries have not been subject to intense research regarding APS systems. Most contributions dealing with APS systems are concerned with the automotive or the semiconduc-tor industry when looking at discrete parts manufacturing (see for example Schmelmer and Seiling 2002; Schneeweiss and Wetterauer 2002; Zeier 2002d) or with the chemical industry when looking at process industries (see for example Hurtmanns and Packowski 1999; Franke 2002; Kallrath 2002; Mekschrat 2002; Richter and Stockrahm 2002). Some research is also related to the food industry in general (e.g. Wagner and Meyr 2002), however no author looks specifically at the requirements of fresh food industries.


How can shelf life be integrated into production planning? How can production planning contribute to optimizing shelf life output?

The outcomes of the second part of the thesis are Mixed Integer Linear Pro-gramming (MILP) models that integrate shelf life into production planning and the solution of those models. The models are built around the case studies from the three sample fresh food industries and will support providers of APS systems to develop tools that integrate shelf life. With respect to literature, only very few au-thors integrated the shelf life of the products into their models. The main contribu-tions are concerned with inventory models for deteriorating items or with adding a shelf life constraint to the Economic Lot Scheduling Problem. However, the major drawbacks of these models are that production aspects such as sequence-dependent set-up times, production on multiple lines or production of multiple

1.3 Dissertation Outline 3

products are often neglected. Furthermore, product freshness is only considered as a constraint and is not part of the optimization. The models developed for the three case study industries address these issues.

1.3 Dissertation Outline

According to the two research questions, the dissertation is divided into two sec-tions. The first section (Chapters 1 to 4, see Fig. 1) aims at answering the first re-search question and concludes with a comprehensive list of requirements. The second section (Chapters 5 to 9) covers the integration of shelf life into production planning.

After having introduced the research subject, the dissertation starts with an overview of the current status of APS systems (Chapter 2). The most important functions of each of the software modules are described, and the level of support for the planner is evaluated. The assessment relies on a literature review of APS systems and of production planning and scheduling, as well as on descriptions of selected APS systems. This analysis provides an understanding of what these APS systems can offer.

Fig. 1.1 Structure of dissertation

4 1 Introduction

Then the characteristics of fresh food industries are examined (Chapter 3). An overview is given on major segments, competition, developments and trends. Lastly, the characteristics of Fresh Food Supply Chains (FFSC) and fresh food production systems are emphasized. The results of this analysis provide the foun-dation for the subsequent development of the requirements of fresh food industries regarding each of the modules of an APS system (Chapter 4). The second section of the thesis begins with an overview of the shelf life characteristics of perishable food products (Chapter 5). The reasons and influencing factors for shelf life limi-tations are examined and options to extend shelf life are evaluated. Furthermore, the consideration of shelf life in the Operations Research (OR) literature is ana-lyzed. Thereafter, the impact of shelf life restrictions on production planning is examined qualitatively. Following this theoretical foundation of shelf life and its implications on production planning, several APS systems are evaluated with re-spect to how they cover shelf life (Chapter 6). For the analysis, three important players in the German SCM software market have been chosen (SAP, PeopleSoft, and CSB). Each software package is assessed based on the scope of shelf life inte-gration and its capabilities to generate plans that optimize shelf life output. Based on the deficits of current APS systems, new models are developed and resolved. The models consider shelf life planning problems in specific fresh food industries (yogurt production, sausage production and poultry processing, Chapters 7 to 9). Special attention is paid to short term planning problems with a planning horizon of one week. In Chapter 10, the major findings are summarized and recommenda-tions for further research are provided.

1.4 Conclusion

To summarize, this research aims at making a twofold contribution to the body of knowledge with respect to APS systems and fresh food industries. First, a profile of three sample fresh food industries with regard to APS systems will be devel-oped. This list will serve fresh food manufacturers to define their own list of re-quirements. In addition, based on this list APS providers can close existing gaps and, hence, broaden the acceptance of APS systems in fresh food industries. Sec-ond, for the most distinguishing characteristic of fresh food industries – the short shelf life of its products – the current support of APS systems will be assessed. Then it will be demonstrated by means of MILP models how shelf life can be in-tegrated into the production planning of three sample industries. These models will allow fresh food producers to optimize product freshness with respect to spe-cific products and customers.

2 Advanced Planning and Scheduling Systems

2.1 Evolutionary Path of APS Systems

2.1.1 MRP I and MRP II

The production planning and scheduling processes that have been implemented by most companies over the last 20 years rely on the Material Requirement Planning (MRP I) and Manufacturing Resources Planning (MRP II) logic (Davies et al. 2002). Fig. 2.1 provides an overview on the emergence of the different applica-tions and the related system architectures over time.

Fig. 2.1 Market penetration of planning systems (based on von Steinaecker and Kühner 2001)

The MRP I concept was developed and refined by J. Orlicky at IBM and the consultant O. Wight in the 1960s and 1970s (Walle 1999). It is a mathematical modeling tool to assist order planners in determining the needs of dependent com-ponents, such as raw materials, parts and sub-assemblies in a manufacturing or warehousing environment.

MRP I is founded on the principle of successive planning and includes four main steps (see for example Tempelmeier 1999b; Walle 1999; Günther and Tem-pelmeier 2000; Steven and Krüger 2002):

6 2 Advanced Planning and Scheduling Systems

Step 1: The entire MRP I process is driven by the end-item schedule of the Master Production Schedule (MPS), which indicates which end items have to be completed within a certain time period (Marbacher 2001), as well as confirmed customer orders. The primary demand of each end-item is determined by additionally considering stock balances of end-items. Step 2: Based on a product structure or Bill of Materials (BOM), the pri-mary demand is translated into gross requirements of the related compo-nents by “exploding” the end items through the BOM. The BOM con-tains the complete product description, including the materials, parts, and components as well as the sequence in which the product is created (Chase et al. 1998). Then the system makes a projection on the stock bal-ances by taking into account the previous stock balance and planned or scheduled receipts and calculates the net requirements for any given part (“netting”). Thereafter, simple lot sizing algorithms are applied (for ex-amples see Buffa and Sarin 1987). The results of this calculation are planned orders including a rough indication of timing. Step 3: Each activity required to produce a part is scheduled to determine the capacity utilization of all necessary resources. If a resource is utilized by over 100% of its capacity, the planner tries either to manually shift non-critical orders or to schedule overtime hours. Step 4: Finally, the planned orders are released to the production depart-ment and assigned to specific resources (e.g. machines or manpower). For each resource, the sequence of the orders is determined, e.g. based on priority rules.

MRP I is regarded as the underlying philosophy for all following production planning and scheduling concepts. Its simple way of data calculation was particu-larly suited for the low performing information systems of the 1970s. By applying MRP I, many companies realized significant benefits, especially in the field of in-ventory reduction in multi-echelon production environments. However, the hierar-chical planning approach often led to infeasible plans, as MRP I assumes infinite capacity (Voß and Woodruff 2000; N.N. 2002a). Feedback loops between input and output do not exist, and the orders for critical parts are often inflated to avoid stock-outs (Thaler 2001). Finally, as an isolated unit, MRP I only applies to a small part of the business function (Walle 1999).

The MRP II concept aims at eliminating the shortcomings of MRP I by inte-grating additional planning modules. It generally includes MRP I as one compo-nent. Therefore, it did not fundamentally change but refine the planning logic (Bartsch and Bickenbach 2002). Fig. 2.2 gives an example of the modules and the planning logic, which are usually incorporated in an MRP II system (Grünauer 2001).

2.1 Evolutionary Path of APS Systems 7

Fig. 2.2 MRP II planning concept (based on Grünauer 2001)

MRP II improved upon MRP I in three distinct ways: First, the forecast of the demand of end-items is now embedded in the general business planning of the company. Secondly, the introduction of feedback loops prevents infeasible plans from being generated by considering capacity constraints. Finally, due to the in-creased computational capacities, more actual information concerning production planning can be managed which results in improved decision making (Kuhn and Hellingrath 2002).

The implementation of MRP II was pushed by a performance improvement of the underlying IT-hardware, leading to an integration of formerly separated infor-mation systems into modular Enterprise Resource Planning systems with one common database (Steven and Krüger 2002). The MRP logic is embedded in all major ERP systems, yet ERP systems go beyond the MRP II logic to manage a company’s entire business and overcome functional boundaries within a company. The software is generally compiled in series of modules, each one covering par-ticular functional elements of the company such as sales, accounting, human re-sources, manufacturing, logistics and many others (Walle 1999). On a worldwide scale, the primary ERP vendors are Baan, Oracle, PeopleSoft, SAP and J. D. Ed-wards (O’Leary 2000). In Germany, SAP is by far the market leader, holding 58,3% of all ERP-related license and maintenance fees in 2000 (Kaftan and Kaf-tan 2002). Today, ERP systems constitute the basic architecture (“backbone”) of all business applications including APS systems. The focus of ERP systems is to support cross-departmental and cross-functional transactions. However, real plan-


ning support is only provided for isolated activities such as algorithms for lot siz-ing (Steven and Krüger 2002).

2.1.2 Assessment of the MRP Planning Concepts

The MRP concepts and their implementations in ERP systems have been subject to intense research throughout the 1980s and 1990s. Although many authors have criticized these concepts, the introduction of MRP I and II has led to significant benefits compared to past concepts. Schonberger and Knod (1994) provide a se-lection of advantages that have been realized:

Improved on-time completions of 95% or higher. A high level of on-time completions has a major effect on customer satisfaction. Cut inventories typically in the range of 20% to 35%. The cut in inven-tory has been realized at the same time as the on-time completions in-creased.Improved direct-labor productivity, ranging from 5%-10% in fabrication and 25% to 40% in assembly as well as overtime cuts between 50% and 90%. This is mainly due to less time being required to halt one job and set-up a shortage-list job. Increased productivity of support staff. As the time for expediting is re-duced and planning procedures are partly computerized, less support re-sources are necessary. Standardization of data. Formerly independent solutions have been re-placed by the MRP II package (e.g. purchasing).

However, despite the progress related to the MRP concepts and its widespread distribution (particularly in larger manufacturing companies), many researchers as well as practitioners report numerous weaknesses of MRP (e.g. Schonberger and Knod 1994; Meyr 1999; Günther and Tempelmeier 2000; Li et al. 2000; Zijm 2000; N.N. 2001; Tempelmeier 2001; Bartsch and Bickenbach 2002; Knolmayer et al. 2002; Lang 2002b):

The major criticism of all authors is that capacity constraints are not con-

sidered (e.g. regarding machines or transportation equipment) in any planning phase and that therefore all processes are planned under the as-sumption of infinite capacity. The main objective of the systems is to create a feasible plan and not to optimize the production with regard to time or cost objectives. The capabilities for a real decision support are limited; MRP remains a transaction-oriented tool. In particular, there is no support for making decisions concerning alternative means, for instance reducing throughput times. In most companies that use MRP, the MPS is planned in weekly time buckets that extend one year into the future and are updated monthly. Hence, toward the end of the month the MPS gets increasingly out-of-


date and the scheduled quantities do not correspond to the sales orders being booked.The scheduling decisions rely on lead-times that have been specified in advance. There is no link to the current situation on the shop floor. For security reasons and due to the generally high risk-aversion of planners, these lead-times are regularly greatly overestimated leading to unneces-sary intermediate stocks. Günther and Tempelmeier (2000) estimate that often the waiting times of an order caused by the organization are as high as 85% of the total lead-time. The size of production orders is determined without considering inter-dependencies. Furthermore, MRP overemphasizes the demand explosion while nearly neglecting the lot sizing part.The focus of the systems is site-centric. Other plants or even other part-ners in the Supply Chain (SC) such as customers, suppliers or shippers cannot be integrated in the planning cycle. Consequently, MRP estab-lishes a local optimum without optimizing the entire chain.The amount of time required to establish a plan or to re-plan is substan-tial (the “MRP-run” is usually executed overnight or over the weekend). For this reason, plans are literally “stiff” as it is not easily possible to change them.

Although the benefits of the systems had been significant in the 1970s and 1980s, many companies were no longer satisfied with the planning results at the beginning of the 1990s. The deficits were immanent to the systems and mainly due to the successiveness of the planning process. Therefore, simply employing modern information technology could not eliminate them (Günter and Tempel-meier (2000). To further improve the performance of production planning, the planning philosophy had to change.

2.1.3 Emergence of APS Systems

The development of the APS systems was motivated by the stated drawbacks of the MRP I and II planning logic. Other drivers were the growing integration of business processes beyond site and corporate boundaries (Bartsch and Bickenbach 2002), improved optimization algorithms, and the significant performance in-crease and innovations in hardware technology in the 1990s (Kodweiss and Nadj-mabadi 2001). For example, an analysis performed by Bixby (2002) revealed that the solving power required to solve production planning problems has increased by six orders of magnitude since 1987, which is related to an increase by three or-ders of magnitude of both the algorithmic and the machine speed.

Advanced Planning and Scheduling systems aim in particular at supporting decision-making in SCM. Some authors use the abbreviation “APS” for “Ad-vanced Planning Systems”; however, in this research APS refers to “Advanced Planning and Scheduling”. These systems are not a substitute for an ERP system,


but can be regarded as a layer on top of an ERP system in order to support the planner in making decisions at all levels (see Fig. 2.3; Davies et al. 2002). The transactions are executed by means of Supply Chain Execution (SCE) systems (e.g. order, inventory, transportation or warehouse management system). The APS system has access to the data of the ERP and SCE systems at any time; it can ma-nipulate the data and write the results of the calculation back into the ERP and SCE systems (Corsten and Gabriel 2002). However, the boundaries between APS, ERP and SCE are fluent. Regarding specific APS systems it becomes difficult to distinguish between APS, ERP and SCE functions.

Fig. 2.3 Relation between SCE, ERP and APS systems (based on Knolmayer 2001b)

Most authors use the terms “APS system” and “Supply Chain Planning (SCP) system” interchangeably and define an SCM system as the sum of an APS (or SCP) and an SCE system. Others discuss SCM systems and APS systems as equivalents, or regard APS systems as a subset of SCP systems (Knolmayer 2001a). Although different software providers have launched APS systems inde-pendently at different points in time, they all have a number of common basic characteristics (see for example Ferrar 2000; Grünauer 2001; N.N. 2001; Hieber 2002; Knolmayer et al. 2002; Meyr et al. 2002b; Werner 2002a):

All APS systems are decision support tools and not transaction systems. They prepare plans and provide the possibility to run what-if analyses of multiple production scenarios, but they do not provide facilities for insti-gating or recording material issues or movements. APS systems can simultaneously compute plans and schedules for multi-ple variables and constraints (e.g. materials, resources, demands, etc.), by permitting them to generate plans that are optimized for multiple and user-defined criteria (cost, time etc.). OR methodologies such as Linear Programming (LP) or MILP as well as heuristics are integrated into the software packages to solve these sophis-ticated planning problems. Powerful standard optimization software is


embedded in the APS systems so that even very complex problems can be resolved. Due to dramatic progress in hardware efficiency, APS systems provide a very high processing speed as they use a dedicated server and in-memory processing. They do not rely on a database to store and locate the data used for the calculations, which avoids repetitive read and write transac-tions to and from the database.

APS systems constitute a significant progress compared to ERP systems; many of the disadvantages of ERP systems have been leveled off (see Table 2.1; based on Ferrar 2000; Benninger and Grandjot 2001; Grünauer 2001; von Steinaecker and Kühner 2001; Davies et al. 2002; Knolmayer et al. 2002; Kuhn and Hellin-grath 2002; Steven and Krüger 2002). The primary differentiating factor between APS and ERP systems is the shift in the planning philosophy that had been un-changed since the 1960s (Tempelmeier 1999a). Constraints and bottlenecks, which have previously been neglected, are now taken into account.

Table 2.1 ERP versus APS systems

Areas Traditional ERP APS Systems

Planningphilosophy

- Planning without considering the limited availability of key re-sources required for executing the plans

- Goal: First-cut requirements es-timate, feasible plans

- Push - Sequential and top-down

- Planning provides feasible and reasonable plans based on the lim-ited availability of key resources

- Goal: Optimal plans

- Pull - Integrated and simultaneous

Business driver Manufacturing Coordination Satisfaction of customer demand

Industry scope Primarily discrete manufacturing All industries including process in-dustries

Major businessareas supported

Transaction: Financials,Controlling, Manufacturing, HR

Planning: Demand, Manufacturing, Logistics, Supply Chain

Information flow Top down Bi-directional

Simulation capabilities

Low High

Ability to optimize cost, price, profit

Not available High

Manufacturing lead-times

Fixed Flexible

Incremental planning

Not available Available

Speed of (re-) planning

Low High

Data storage for calculations

Database Memory-resistant


The objective of production planning shifted from generating feasible plans to plans that are subject to company-specific optimization criteria. Therefore, all planning parameters of a specific planning problem have to be considered simul-taneously. Ideally, production lead-times that are fixed in the MRP logic can be reduced to the extent that an order-based pull production can be implemented. In contrast to the MRP logic APS systems are also suited for process industries (Fleischmann 1998).

Due to complex processing conditions, the efficient reorganization of the sup-ply chain is much more difficult in these industries than in discrete parts manufac-turing (Günther and van Beek 2003). Finally, another important difference is the decision support function, which is linked to the capability to quickly create new plans. While it took a considerable amount of time – sometimes even a runtime of 24 hours (Fritsche 1999) - to establish a plan with an ERP solution, APS systems deliver the results much faster due to the memory-resistant data storage. This is especially true for smaller changes of parameters, since in this case ERP systems had to recalculate the entire plan.

2.2 Structure of APS Systems

2.2.1 Overview

Today, APS systems cover most aspects of supply chain planning: from procure-ment to sales and from strategic to operational decisions. Different modules of APS systems support different tasks in the planning process (see Fig. 2.4).

Fig. 2.4 Software modules of APS systems (based on Meyr et al. 2002b)

2.2 Structure of APS Systems 13

The representation is based on the so-called Supply Chain Planning Matrix. Al-though the names of the modules differ depending on the APS provider, the sup-ported planning activities are generally the same (Meyr et al. 2002b). However, most of the implementations do not cover all of the modules. Usually only those desired by the customer are activated and installed. Although not all APS provid-ers currently offer all modules, the clear trend is to supply an entire package cov-ering all planning tasks. The following modules can be distinguished (at this place, only a short overview on the different modules is provided; for a detailed analysis of each module, it is referred to Chapter 2.2.2 to Chapter 2.2.9:

Strategic Network Design (SND) determines the structure of the SC with a long-term planning horizon up to ten years (Neumann et al. 2002). Within a tactical or mid-term planning horizon, the Supply Network Planning (SNP) module aims at efficiently utilizing the company’s ca-pacities. Therefore, the purchasing, production and distribution functions are planned simultaneously. The MPS is one important result of this mo-dule.The Demand Planning (DP) module incorporates both strategic long-term demand estimation and mid-term sales planning (Meyr et al. 2002b). Most of the functions of ERP systems concerning production planning and purchasing are now incorporated in the Purchasing & Materials Re-

quirements Planning (P&MRP) module. However, as many companies have these functions already available in their legacy ERP system, this module is only seldom provided in APS packages (Meyr et al. 2002b) and therefore not presented in detail in the following paragraphs. Production Planning (PP) and Production Scheduling (PS) aim at deter-mining lot sizes and detailed production schedules. Depending on the production type, PP and PS can be performed on one or two planning levels (Stadler 2002). Distribution Planning (DisP) and Transportation Planning (TP) seek to build the most efficient transportation method (Davies et al. 2002). While the DisP module is a more detailed representation of the Master Planning for the distribution part (Meyr et al. 2002b), TP considers short-term fac-tors such as routing or vehicle availability. Available-to-Promise (ATP) is meant to generate quick and reliable order promises (Kilger and Schneeweiss 2002a).

While a representation of the APS modules along the dimensions business process and planning horizon as in Fig. 2.4 is favored by many authors (see for example Corsten and Gössinger 2001b; Kilger and Müller 2002; Rohde 2002; Neumann et al. 2002), some elements are missing or misleading. As planning ac-tivities for different industries can vary noticeably, industry-specific planning so-lutions should be added, in particular at the mid- and short-term planning levels. Moreover, the aspect of collaboration with suppliers and customers should also be integrated due to its importance in implementing efficient and effective SCM (Meyr et al. 2002b). Finally, Tempelmeier (2001) criticizes that the multi-location


structure of a network is not visible and that DP and ATP have nothing to do with planning, but instead with the generation of input data. Likewise, he emphasizes that the MRP calculations should be shifted from purchasing to production since APS systems do not replace the MRP calculations, but rely on their results.

However, despite the deficits stated above, the representation depicted in Fig. 2.4 is still suited to describe the current status of APS systems, because the struc-ture of the APS systems has developed according to the two dimensions of the SC planning matrix (Tempelmeier 2001). As shown later, the weaknesses of the rep-resentation coincide with the actual weaknesses of APS systems in general. In the following chapters, the planning objectives of each of the modules are analyzed in detail. Major input and output factors are described. Special attention is paid to the solution algorithms that are incorporated into the modules.

2.2.2 Strategic Network Design

The SND module provides support for key strategic decisions concerning the con-figuration of the entire SC (Hartweg and Bruckner 2001). The module is essential for industries with frequently changing SCs and material flows (Hauptmann and Zeier 2001). The planning horizon typically ranges from three to ten years (Goet-schalckx 2002). Primary users of the module are business development depart-ments or consultancies (Fraunhofer Gesellschaft 1999). The issues addressed can be structured around four major pillars (Davies et al. 2002):

Product Strategy: number and main characteristics of products as well as markets to be served; Manufacturing Strategy: number and location of plants, sourcing strat-egy, investment decisions, supplier selection; Logistics Strategy: number, locations and echelons of distribution cen-ters, sourcing strategy, investment decisions; and Investment/Divestment Decisions: in-/outsourcing, acquisitions/mergers, new technology introduction.

SND models must tie together all relevant decision variables and constraints re-lated to countries, periods, products, facilities, transportation channels, product flows and inventories just to mention a few examples. Therefore, they are very large (Geoffrion and Powers 1995; Goetschalckx 2002). To limit the model size, products, suppliers and customers are usually aggregated to zones (Vidal and Goetschalckx 1997). Some APS providers also incorporate model size limits to keep the models solvable (Steven and Krüger 2002). The modeling support pro-vided by the SND module generally comprises numerous functions. Typical fea-tures are multi-echelon and multi-period modeling, finite capacities for sourcing, production, distribution and transportation facilities or piecewise cost functions representing economies of scale (Goetschalckx 2002). In addition, the model should also account for governmental issues such as tariffs, duties and transfer prices (Cohen and Lee 1989). Stochastic features have not yet been integrated into most systems, although future estimates incorporate a high degree of uncertainty.


To cope with this, scenarios can be developed that describe the best, the worst and the most realistic case. Either cost minimization or profit maximization can be chosen as objective function. The models are defined as LP or MILP models and can hence be solved with standard solvers. Nonetheless, in order to achieve rea-sonable solution times, a significant level of technical expertise is required to limit the model size (Goetschalckx et al. 2002). For an example from industry, it is re-ferred to Cohen and Lee (1989).

Due to the necessity to reduce complexity and in order to reach a high level of abstraction, the models can only constitute a decision support tool for the SC de-sign team. The results should be evaluated with a “healthy skepticism” (Goet-schalckx 2002), as the planning results have a high influence on all following planning steps. Although the results of the SND module have the highest impact on the SC, it is interesting to notice that APS providers generate only marginal li-cense fees with this application. Many companies hesitate to implement the mod-ule for two reasons: First, the strategic decisions for which the module is used are company specific to a high degree, so that an individual support cannot be pro-vided by standard applications. Secondly, these decisions are not made very often, and therefore most companies prefer to use simple Excel-based solutions.

2.2.3 Demand Planning

The objectives of the DP module are to forecast and plan future demand (Davies et al. 2002). DP is relatively easy to install thanks to the limited interactions with other modules. In addition, the results of the DP module are required as input fig-ures for the other modules. Therefore it seems reasonable to start an APS imple-mentation with this module. Two levels can be distinguished within DP: The long-term demand is generally forecasted for several years on a product group or prod-uct family level. It serves to support the SNP module. The mid-term demand is elaborated on a Stock Keeping Unit (SKU) level with a planning horizon of months or weeks and can generally be partitioned by customers, regions, seg-ments, or distribution channels (Kuhn and Hellingrath 2002). The mid-term de-mand figures constitute the input data for several other modules such as SNP, DisP or ATP. Finally, the short-term demand is derived from the orders in the ERP system.

Forecasting in APS systems relies on three components (Wagner 2002). First, statistical forecasting methods assist the planner in making estimations derived from historical data. Within statistical forecasting, time series methods assuming demand to follow a certain pattern can be distinguished from causal methods, which focus on the relationships between two series (dependent and independent variable). Examples for time series methods are (Davies et al. 2002; Meyr 2002):

Moving Average: smoothing time series to reduce period-to-period varia-tion;Classic Decomposition: decomposing a time series into trend, cyclical, seasonal and error components;


Exponential Smoothing / Holt-Winters: smoothing time series by assign-ing greater weights to most recent observations and including trend and seasonality through decomposition; Autoregressive Integrated Moving Average (ARIMA)/Box-Jenkins: mod-eling a series using trend, seasonal and smoothing coefficients that are based on moving average, auto regression and differential equations; Croston: forecasts the length of periods and the size of demand for spo-radic demand.

Examples for causal methods are (Davies et al. 2002; Meyr 2002):

Multiple Regression: modeling the relationship between one dependent and many independent variables using the least squares method; Econometric models: estimating the relationships between one or more endogenous and exogenous variables using the least squares method to model mutual causality; Neural Networks: network of elementary nodes that are linked through weighted connections.

A major shortcoming of both types of methods is that only inventory exit vol-umes are registered in ERP systems. Unfulfilled demand results in an inventory exit of zero, although there is a demand for the specific product (Tempelmeier 1999a).

Secondly, judgment factors are incorporated to correct and improve the statisti-cal forecast. Data on promotions and marketing campaigns (own and competitors), customer feedback on special products, or cannibalization with regard to product launches can be integrated into the forecast by judgment factors (Seidl 2000). Changes in pricing policies are another primary driver for demand volatility that can only be captured by judgment factors (Bolton 1998). Davies et al. (2002) name several methods for integrating qualitative factors:

Panel consensus: consensus of experts to yield a better forecast than a single expert’s opinion; Sales force composite: average forecast from independent inputs of sev-eral salespeople; DELPHI: iterative process in which experts respond to questionnaires that are tabulated and modified in reaching conclusions.

Thirdly, the collaboration component assures that input for the demand plan-ning process can be collected from all involved departments such as marketing, sales, procurement, or logistics (Rojek 2000). Collaboration is not only necessary to gather all relevant information, but also to get an organizational agreement for the planning results (Smith et al. 1998). In case of an external collaboration, de-mand forecasts from customers are integrated as well.

In addition to the statistical forecasting functions, features that support the qual-ity control of the forecast, the selection of methods and parameters, and forecast-ing based on product life cycles are integrated. Most common measures to control forecast accuracy are the mean squared error, the mean absolute deviation, the


weighted mean absolute percent error, and the mean absolute percentage error (Wagner 2002; Smith et al. 1998). After having calculated the accuracy of a method, most APS systems can automatically generate a proposition as to which method or which combination of methods with which parameter adjustment achieves the highest forecast accuracy; frequent forecasting methods changes, however, lead to plan nervousness (Corsten and Gössinger 2001b). Forecasts based on life cycles should only be generated if the demand curve of the product is relatively similar to the compared product, which is not often the case in real life (Corsten and Gössinger 2001a). Lastly, most DP modules can support the calcula-tion of single-stage safety stocks.

With respect to the DP module, Cap Gemini Ernst & Young (2002a) expect in-creased capabilities in external collaboration to be the most important develop-ments for the forthcoming years. This trend can be observed for other modules as well but it is particularly important for DP. The collaboration function also allows the customers to enter demand data directly into the system. In addition, the APS providers aim to better integrate the DP module with other APS modules. In many cases, APS systems consist of modules that have been acquired from different APS providers, and are therefore not yet as integrated as they should be.

2.2.4 Supply Network Planning

The SNP module aims at synchronizing the flow of materials along the SC. It sup-ports mid-term decisions concerning efficient utilization of production, transport and supply capacities, seasonal stocks as well as the balancing of supply and de-mand (Rohde and Wagner 2002). SNP can be carried out on an intra- or inter-company level. In the latter case, the “strongest” partner in the SC is in charge of the planning as he has the highest degree of added value, and is often within clos-est vicinity to the final customer (Kuhn and Hellingrath 2002). To integrate all demand peaks, the planning horizon has to cover at least one seasonal cycle (Neu-mann et al. 2002). Frequently, the planning horizon covers 12 months and is di-vided into periods of a week or a month (“time buckets”). Inputs to the SNP mod-ule are the determinations of the SND and DP module and data on capacity, costs and stock levels per plant or Distribution Center (DC). Due to the complexity of the problem, only bottleneck resources can be modeled in detail. Moreover, the BOMs of all products or product groups are required to gain information on input-output coefficients (Rohde and Wagner 2002).

In order to solve the optimization problem, LP or MILP techniques are applied; most vendors use a mix of internally developed and third party solvers like ILOG/Ceplex or DASH/Express (Shepherd and Lapide 1999). The most important decision variables include sourcing, production and transportation quantities as well as inventory levels for every product, period and plant. To increase the solv-ability of the model, most vendors distinguish between hard and soft constraints. While hard constraints have to be fulfilled, the violation of soft constraints is only penalized by the model (Davies et al. 2002). For instance, if the demand cannot be fulfilled completely, penalties are imposed for the lacking volume. Another tech-


nique to reduce model complexity is aggregation. Rohde and Wagner (2002) dis-tinguish the aggregation of time, decision variables and data. When aggregating the time, several smaller periods are consolidated into a larger period. The aggre-gation of decision variables generally refers to the consolidation of production and transportation quantities by aggregating for example products to product families. This is generally more difficult for the supply side than for the delivery to the final customers due to the influence of differential supply costs among suppliers and their greater uniqueness in terms of which supplier provides what (Geoffrion 1977). Finally, aggregating data comprises the grouping of, for instance, produc-tion, transport or inventory capacities, or the grouping of product demand data into product family demand (Rohde and Wagner 2002).

The outputs of the model are aggregated production and distribution plans that are transmitted as requirements for the detailed plans to the PP, PS, DisP, TP, and to the P&MRP module. The main purpose of the SNP module is thus to coordinate the detailed plans, yet the achieved planning results cannot be used without further processing (Steven and Krüger 2002). Therefore, for Rohde and Wagner (2002) the most important planning results are the planned capacity usage and the amount of seasonal stock at the end of each time bucket, both of which cannot be deter-mined in the short-term planning modules as they need a full planning cycle. Only the planning results of the first time bucket are obligatory for the short-term plan-ning modules (frozen horizon), while the requirements for the following periods are determined in later planning runs with actualized data (Corsten and Gössinger 2001a; Thonemann et al. 2003).

2.2.5 Production Planning

Among all modules, the importance of an IT support is considered highest for PP and PS (Nienhaus et al. 2003). In the PP process, the aggregated production plan resulting from the SNP module is subsequently disaggregated to provide an opti-mized production plan for each site of the SC. Hence, the responsibility is gener-ally located at the site level. The planning horizon differs between weeks and months with time buckets of days or weeks. The typical granularity is on machine group level; further detailing is then performed within the PS module (Kortmann and Lessing 2000). The production and distribution quantities of the SNP module for each time period constitute the most important input for PP as it sets the frame within which the decentralized PP decisions can be performed. Other directives usually include the amount of overtime to be used, the availability of upstream items in the SC or the amount of seasonal stocks to be built up (Stadler 2002).

According to Kuhn and Hellingrath (2002), the determination of the production plan is done in two steps. First, exploding the BOM or the recipes disaggregate the production requirements of the SNP. The BOM explosion is also frequently per-formed in the P&MRP module or even in the underlying ERP system. The second step comprises lot sizing and a rough scheduling of the resulting production orders for all parts (on machine group level) with regard to capacity restrictions, shift plans or alternative resources.


The output of the PP module is a site level production plan including the corre-sponding capacity and material requirements that have been leveled according to critical capacity constraints. The further detailing of the plan is then conducted within the following PS module. The PP (and also the PS) capabilities of APS sys-tems vary considerably in sophistication concerning type and level of constraints, strength of the solvers or exception alerting (Davies et al. 2002). For instance, Tempelmeier (1999a) criticizes the lot sizing support as insufficient and Davies et al. (2002) state that the support of lot sizing can differ between a simple unit lot sizing and the calculation of economic lot sizes based upon manufacturing con-straints.

2.2.6 Production Scheduling

The purpose of PS is to schedule in detail the resulting production orders from the PP module. In most cases, the planning responsibility is decentralized at the pro-duction department level to respect local particularities (Kuhn and Hellingrath 2002). The typical planning horizon comprises a couple of hours or days (Kort-mann and Lessing 2000). The PS module considers a variety of constraints or manufacturing rules such as changeover times, routing requirements, resource preferences, or demand priorities (Davies et al. 2002). Stadtler (2002) emphasizes that the objectives of PS are mainly time oriented (e.g. reducing makespan, the sum of or the maximum lateness, or the sum of flow or setup times). Cost oriented objectives (e.g. reducing variable production costs, setup costs or penalty costs) can also be integrated. To solve the planning problem, APS systems use rather simple heuristics (Tempelmeier 1999a). Three types are usually available:

Constraint Programming is a technique to compute feasible (not always optimal) solutions to combinatorial decision problems by solving con-straint satisfaction problems consisting of variables, domains and con-straints. In contrast to LP/MILP techniques, the user can influence the search strategy (Klein 2002b). Genetic Algorithms use procedures (such as selection, mutation and crossover) that are recognized in the evolution of the natural world to find solutions for planning problems (Knolmayer et al. 2002). It finds near-optimal solutions within a reasonable time (Klein 2002a), and has only low hardware performance requirements (Stache 1997). Incremental Planning is used to integrate new orders into a given se-quence. In this case, time gaps are searched for that result in only minor adjustments to the old plan in order to avoid plan nervousness (Stadtler 2002).

The generation of schedules can either be performed on a two level planning hierarchy as described above using the two modules PP and PS, or in a single planning step when PP and PS are integrated. The decomposition decision de-pends on the production type (Stadtler 2002):


In process organization with many different machines of similar func-tions, multi-stage production processes and many lot sizes within the planning interval, a decomposition of the overall decision problem into two planning levels is required to reduce the computational burden. However, for an automated flow line with scarce resources, sequence de-pendent set-up times, and a smaller number of products, a two stage planning hierarchy is not adequate because sequence dependent set-up times cannot be adequately represented by time buckets. Furthermore, the lot sizing and sequencing decision cannot be separated in that case as lot sizing depends on scheduling and vice versa. Therefore, with regard to the smaller number of products, one single planning step is preferable.

2.2.7 Distribution Planning

DisP is part of the mid-term planning with a planning horizon in the range of days to months. The objective is the planning of inventory levels of final products and of the distribution of the final products to the customer, with the objective to op-timize the trade off between inventory holding cost and transport cost. Several de-cisions are to be made within DisP (Fleischmann 2002):

The determination of aggregate transport quantities for every transport link in the SC is the most important activity to be performed within DisP. The frequencies of regular transports set target values for short-term de-cisions on shipment quantities and determine the size of the transport lot. A framework for the selection of distribution paths with regard to limits of order size is set in DisP (e.g. direct delivery if order volume exceeds 30 pallets). Furthermore, on the supply side materials are assigned to supply con-

cepts like direct delivery or Logistics Service Provider (LSP).

The data input for the DisP module is provided from several other modules. The structure of the network including the locations of factories, DCs, and suppli-ers, transport modes and paths, as well as the allocation of suppliers and customers to areas have been determined in SND. SNP delivers the quantities to be shipped and variations in seasonal stocks. Finally, demand forecasts and safety stocks are added from DP (Fleischmann 2002). Generally, DisP overlaps with the SNP mod-ule to a large extent and hence, can only increase the planning performance in the case of a transport network with many far-off lying nodes having an identical range of products and materials (Steven and Krüger 2002).

The results of the distribution planning process are the primary input for the short-term TP module. Hence, a tight integration between both is a must. How-ever, as the TP module has often been acquired by the APS providers and not de-veloped in-house, the integration is frequently incomplete. Consequently, most vendors currently focus on an increased integration of DisP with TP because a consideration of the constraints of the transportation capabilities will result in a more efficient distribution plan (Cap Gemini Ernst & Young 2002a).


2.2.8 Transport Planning

Based on the distribution plan, TP as a short-term planning module seeks to build the most efficient transportation method considering constraints such as costs, routing information, availability and speed of vehicles, loading constraints and mix, and timing (Davies et al. 2002). The planning horizon is the same as for PS, that is, it ranges from hours to days (Kortmann and Lessing 2000). Knolmayer et al. (2002) name three major functions of TP:

The Load Consolidation and Vehicle Scheduling function helps to con-solidate the load to destination locations and aims at achieving high vehi-cle fill rates. Route Determination supports the planner to find the best route through a network with regard to time and cost. Carrier Selection allows one to choose from several carriers and usually includes an Internet-based tendering process.

Most APS systems apply a combination of heuristics and LP/MILP procedures to solve the planning problem. The planner can intervene by means of a user inter-face to integrate specific load optimization strategies (Davies et al. 2002). Several additional features are available within specific APS solutions. Some examples in-clude tracking and tracing functions (Bergmann and Rawlings 1998; Lang 2002a) or even a cargo revenue-maximizing assistant to sell overcapacities (Davies et al. 2002). Nonetheless, in spite of all the offered functions, the use of the TP module is generally only reasonable if the company manages a significant own fleet. The TP module’s offered functions are particularly useful for logistics providers (Nos-bers and Plewnia 2001; Steven and Krüger 2002). Regarding the future develop-ment, it is expected that a wide range of functional enlargements will occur in TP. Examples include industry-specific solutions for LSP, fleet management functions or the integration with on-board computers (Cap Gemini Ernst & Young 2002a).

2.2.9 Available-to-Promise

The major objective of the ATP process is to generate fast and reliable order promises to the customer and to shield production and purchasing against infeasi-bility (Kilger and Schneeweiss 2002a). For Kuhn and Hellingrath (2002), order promising is one of the key tasks in SCM because it links planning tasks that are independent of customer orders and the planning tasks related to customer orders. APS systems are particularly suited to support ATP due to their high processing speed (Werner 2002b). In the traditional approach of order promising, orders were quoted against production lead-time if there is no inventory available. This has led to infeasible quotes as supply or capacity constraints have not been considered (Kilger and Schneeweiss 2002a). According to Fischer (2001), the ATP approach of contemporary APS systems can be structured around four activities:

First, the checking of Product Availability is the core of the ATP module. The basis for the product search constitutes available inventories and the quantities


calculated in the SNP (the SNP quantities available for order promising are also called “Available-to-Promise”). The ordered volume of the product is reserved and the available quantity in the corresponding time bucket of the SNP is reduced accordingly. If the ordered volume is not available, an alternative delivery date can be proposed or pre-defined rules can be applied (Schneider and Grünewald 2001) The search can be extended to alternative products (see Fig. 2.5, Number 1), alternative locations (Number 2), or both (Number 3). The system can even review production commitments and re-plan the production. This procedure is called Ca-pable-to-Promise (CTP, Number 4).

Secondly, the Initial Order Promising function aims at confirming the delivery date and quantity to the customer. A company that is able to consistently make re-liable promises over a long period of time creates an important competitive advan-tage (Kilger and Schneeweiss 2002a).

Thirdly, ATP supports measures and decisions regarding temporary delivery inability. This is especially important when only a part of all customer orders can be satisfied with the available volume. In that case, a variety of shortage allocation rules can be applied. Some examples are (Fischer 2001):

Allocation proportional to the volume of customer orders; Allocation proportional to the customers turnover; Allocation proportional to the customer’s demand forecast or Allocation according to predetermined priority rules.

Fourthly, Due Date Control and Re-Promise are also essential for customer sat-isfaction as re-planning of already confirmed orders can never be totally avoided due to unexpected events. The objective of ATP in that case is to identify potential delivery bottlenecks as early as possible.

Fig. 2.5 ATP rule representation (based on Knolmayer et al. 2002)

2.3 APS Systems Market Overview 23

The granularity of the ATP with respect to the product dimension depends on the SC decoupling point (see Table 2.2). In a Make-to-Stock (MTS) environment as it can be found in most Consumer Packaged Goods (CPG) industries, the ATP is generally on a finished products level due to short customer order lead-times. As a high number of configurations is possible in a Make-to-Order (MTO) envi-ronment (e.g. computer industry), the forecast and the ATP are based on product group and component level. Depending on its configuration, an order can consume multiple resources. Characteristics of a Configure-to-Order environment are long production lead-times and difficult forecasts with BOMs being only partially available. In that case, capacities must also be considered so that ATP resembles the CTP procedure. Regarding the time, the ATP module is usually represented in the same granularity as the supply given by the master plan. Therefore, quoting an order means consuming quantities from a particular time bucket (Kilger and Schneeweiss 2002a).

Table 2.2 ATP granularity (based on Kilger and Schneeweiss 2002a)

With regard to the future development of the ATP module, Cap Gemini Ernst & Young (2002a) foresee a further development of so called “Advanced Order Promising”. A very prominent example is “Profitable-to-Promise” in which the profitability of an order can be determined and taken into account to facilitate the fulfilment decision. Another expected development is the further expansion of ATP/CTP functions in the areas of multi-echelon ATP and true CTP, meaning that in fact all modeled capability constraints are considered. Currently, this is not pos-sible in most cases due to the high complexity of the planning problem.

2.3 APS Systems Market Overview

2.3.1 Available Market Studies

Although the APS software market has been subject to intense market research in the recent years, most companies in CPG industries assess the market for APS sys-tems as not very transparent (Lebensmittel Zeitung and PwC Consulting 2002). One of the first comprehensive market overviews was developed by the Fraun-hofer Gesellschaft (1999), which compares 20 APS systems based on a detailed questionnaire. An overview of the most important market studies is given by Kortmann and Lessing (2000). The market studies covered include for example

Manufacturing

Environment

Order Lead-Time ATP Granularity

Make-to-Stock Transportation time Product groups / Finished Goods

Make-to-Order Production lead-time / Transportation time

Product groups / Components / Intermediate Products

Configure-to-Order Production lead-time / Transportation time

Capacity / Components


the Supply Chain Magic Quadrant and the SCP vendor “footprint” of the Gartner Group, the Manufacturing Systems Software Database of Cahners Business In-formation and the Manufacturing Enterprise Applications Comparison guide. They analyzed the results in detail and mention these major shortcomings:

The information for the market studies are given by the APS providers and has not been subject to intensive checks; Lacking comparability of the systems; Very technology-oriented analysis; No integration of reference implementations in the studies; and Few indications on pricing.

Based on these drawbacks, the authors have also integrated an enquiry of users in addition to the interrogation of software providers and give indications on the pricing model of the providers. More recent market studies from 2002 and 2003 are more focused on specific industries or functions. Trier’s (2002) comparison of 18 APS software packages stresses the importance of implementation issues. Con-sulting companies in the Netherlands have published two overviews on APS sys-tems in process industries. The study prepared by Davies et al. (2002) includes a detailed analysis of ten APS software packages focusing on consumer products and process industries. The analysis of Cap Gemini Ernst & Young Netherlands (2002a) concentrates on the Dutch APS market and investigates specifically on semi-process industry functions. Only very few market studies are not limited to the analysis of single APS systems and its capabilities, but provide data on the to-tal market size and development. The most comprehensive analysis is provided AMR Research (2001 and 2003).

2.3.2 Market Size and Segments

The worldwide market for SCM software has a total annual volume of about $ 5 bn. This number includes APS systems as well as SCE software. After a period of dramatic growth, sales reached a peak in 2001. In 2002, the SCM market has seen a decline for the first time and shrunk by ca. 6%. Reasons for this development are the general economic slow down and the end of the Internet hype. However, de-spite this first decrease in SCM software spending, the market is expected to fur-ther increase but at a slower growth rate than in the late 1990s (AMR Research 2003).

With regard to the types of income of APS providers, software license fees and implementation fees generate the largest part of income for the software providers. Although the SCM business is basically a license business, the implementations often include process redesign and other consulting services, which result in a high portion of implementation fees. However, the implementation fees are typi-cally a one-time income so that the software providers have to sell a significant number of new implementations every year. This is increasingly difficult in a sluggish economy as the IT budgets for new projects are the first to be cut in case of cost reduction programs.

2.3 APS Systems Market Overview 25

APS providers obtain most of their income from larger companies. This is mainly due to the fact that the benefits of implementing such software are much bigger in larger companies, which have to handle a far higher complexity regard-ing sites, departments, or products. These customers can already improve their op-erations significantly on an inter-enterprise basis, whereas smaller customers have to integrate their customers and suppliers right from the beginning to fully benefit from employing the software.

Finally, although the implementation of APS systems generally provides a quick Return on Investment (ROI), the investment represents a high risk for smaller companies if the implementation fails or if not all targeted benefits can be realized.

2.3.3 Major Providers

Despite several takeovers in recent years, the market is still relatively fragmented. The major players in the market are i2, SAP, PeopleSoft/J.D. Edwards and Manu-gistics. Due to the fact that most providers offer both APS and SCE functions, a clear distinction between APS and SCE providers is rather difficult. Therefore, the SCM software market (which includes APS as well as SCE) is first categorized to understand the different players. SCM software providers can be divided into five categories (Hellingrath et al. 2002):

Integrated SCM- and e-business suite providers: Solutions that cover al-most every aspect of SCM (APS and SCE). Generally, these providers start with an APS system and extend the system to create an integrated “SCM-Toolsuite”. Prominent examples are i2, Manugistics, and SAP. Specialized SCM suite providers: Providers in this category are similar to providers in the first category, but are more concentrated on special tasks or industries (e.g. AspenTech for the process industry). Their competitive advantage often consists of sophisticated algorithms to solve specific planning problems. Several providers in this category are currently enlarging their range of products and will soon be part of the first cate-gory.Enlarged ERP systems: Providers that develop their SCM functions as an enlargement and complement to their ERP systems. Examples are Baan, PeopleSoft, Oracle, SCT or J.D. Edwards. The SCM part has frequently been added through a company takeover (e.g. Baan acquired Berclain, CAPS Logistics, CODA and Aurum; PeopleSoft acquired Red Pepper and J.D. Edwards which bought Numetrix before). Pillep and von Wrede (1999) emphasize the tight integration of the SCM and the ERP system of the solutions in this category. Therefore, these providers leverage their ERP customer base to sell the APS system. APS niche players: Providers that develop individual solutions for special tasks in the SC or for target groups (e.g. flexis or ICON). By focusing on special activities, they can frequently offer attractive solutions with re-


gard to price and quality and are especially suited for small/mid-size en-terprises. SCE software providers: Companies focusing on transaction-oriented systems such as Inventory Management or Warehouse Management. Due to the high variety of activities along the SC, this category is very hetero-geneous.

Fig. 2.6 SCP magic quadrant (based on Peterson et al. 2002)

When looking specifically into the APS segment, many authors (e.g. Buscher and Jelken 2000; Kortmann and Lessing 2000; Grünauer 2001) use the representa-tion of the Gartner Group (“SCP Magic Quadrant”; Peterson et al. 2002) to illus-trate the market position of the APS systems (see Fig. 2.6). This instrument aims at supporting companies when selecting an APS system and integrates a variety of factors along the dimensions “Ability-to-Execute” and “Completeness of “Vi-sion”. Only providers with an own multi-module package and a credible vision are accepted. The “Ability-to-Execute” is a score for the sustainability of the company and the product and integrates factors such as breath and depth of the product, the technical expertise, service levels, the stability of the product and of the manage-ment, the financial performance as well as the marketing of the company. Hence, the “Ability-to-Execute” score indicates how a provider corresponds to today’s re-quirements. The “Completeness of Vision” assesses the vision of the solution with regard to cost, functions, technology, service and sustainability. Another important factor for the judgment is the integration of the solution components. Therefore, the second dimension of the matrix shows how a provider will correspond to fu-ture requirements.

2.4 Implementation of APS Systems 27

2.3.4 Expectations for the Future

Despite a decline of approximately 6% in revenue in 2002, the total APS market is expected to grow further; AMR Research foresees a Compound Annual Growth Rate (CAGR) of 5% over the next five years, compared to 10% for SCE systems (AMR Research 2003). Because benefits and payback times become more impor-tant in a sluggish economy, companies are shifting their SCM software spending from the bigger “visionary” projects or from buying entire SCM suits to smaller projects in order to optimize specific SC problems. APS providers that are able to tailor their products to these problems (e.g. tank planning, formulation optimiza-tion, or blending) will be successful in the future. Mid-market APS providers are better positioned for these implementation than high-end providers (Cap Gemini Ernst & Young 2002a). Traditional ERP vendors will continue to be successful because they can benefit from their ERP customer base. However, they will also be restricted to their customer base as it will be difficult to sell these solutions “stand-alone”. For “pure” APS providers, the integration with ERP systems will become increasingly important. Many of the specialized SCM providers and niche players will continue to enlarge their functions either by developing their own so-lution, or by acquiring another provider. However, the latter will take place on a selective basis, the market consolidation of the recent years is likely to slow down.

2.4 Implementation of APS Systems

2.4.1 Implementation Process Overview

The implementation of APS systems is a complex and difficult process that ad-dresses both the planning processes within a company as well as the relations to SC partners (Poluha 2001). Usually, many different people and functions (e.g. production, purchasing, sales etc.) are involved in the implementation process. Moreover, deep knowledge of IT and process redesign is required to successfully implement APS systems. Due to this complexity, APS implementations bear a high risk of failure if not managed correctly. To address this problem, all sources propose a structured implementation approach (e.g. Kortmann and Lessing 2000; Davies et al. 2002; Hellingrath et al. 2002; Kilger 2002a and 2002b; Wetterauer 2002). When summarizing the proposed procedures, the APS implementation process can be divided into three major steps with three sub-steps each (see Fig. 2.7). The objective of the step “Project Definition” is the generation of a business case including the analysis of cost and benefits and the set-up of a “Masterplan” for the implementation which indicates which modules when to implement. The second step aims at identifying the right provider and the last step comprises the implementation itself. In the following paragraphs, each of the steps is presented in detail. Finally, in Chapter 2.4.5 the most important implementation risks are analyzed. It should be noted that the described process in not an “optimal” proc-ess. Rather, the process must be adapted to the specific company situation.


Fig. 2.7 APS implementation process

2.4.2 Project Definition

An important driver for an APS project is the realization that some kind of signifi-cant change is required (Kilger 2002a). This can be seen in a negative way as pain

or perceived positively as a potential. To realize the potential, the current per-formance of the SC has to first be analyzed in order to identify areas of improve-ment (Supply Chain Review). These opportunities should then be assessed, quanti-fied (Assessment of Opportunities) and prioritized. This leads to the project roadmap that includes all sub-projects, their scope and objectives, and their im-plementation order and timelines (Prioritization and Project Masterplan).

Supply chain review

A thorough review of the current SC should always be the first step when imple-menting an APS system. A common understanding within the organization about the structure and the processes of the SC, as well as potential improvement areas, has to be developed. The scope of the review should encompass all hierarchy lev-els of the corporation, and all major functional areas. Special attention should be paid to marketing and sales, product management and engineering, purchasing, manufacturing, distribution and IT. Typical weaknesses are low forecast accuracy, high stock levels (raw materials, work-in-progress, and finished products), low performing order management, insufficient delivery performance, low product quality, or high cost levels with regard to sourcing spend, production or transpor-tation cost (Kilger 2002a). As an APS system can only be as good as the data that is used, the availability and quality of the data should also be assessed in the re-view (Davies et al. 2002). Top-level support is absolutely necessary for an APS implementation. On the one hand, an implementation touches numerous strategic areas of the company (e.g. relationships to suppliers or customers). On the other hand, the role of the board is to drive the change management. Besides the top management involvement, the participation of the planners is a critical success factor because much of the planning knowledge is in the heads of these people.


Furthermore, they will have to apply the new system in the future, so that their buy-in from the very beginning is a prerequisite.

Assessment of opportunities

The main objective of an APS implementation is to realize a financial benefit. Therefore, all improvement opportunities have to be assessed in terms of their fi-nancial impact for the corporation (Richmond et al. 1998), which can be deter-mined by the means of a value creation tree (see Fig. 2.8).

Fig. 2.8 Assessment of SCM opportunities (based on Salehi 2003)

The use of this representation method has many advantages. First, it ensures that all levers have been addressed. Second, for each improvement opportunity a requirement has to be defined that is in most cases directly linked to a module of an APS system (e.g. PS to create feasible plans). Therefore, the levers addressed by the same module can easily be derived. Third, it requires that all possible im-provements of logistical Key Performance Indicators (KPI) – e.g. the number of infeasible plans down by 95% - be translated into financial benefits. Finally, the impact on value creation can be derived directly.

Prioritization and project masterplan

Once the benefits of all improvement opportunities are determined, the different APS modules are prioritized along the dimensions “Expected Benefit” and “Ease of Implementation”. The implementation starts with the modules that show the highest expected benefit, and that are relatively easy to implement in terms of both resistance to change and technical complexity. The prioritized modules lead to the Project Masterplan that includes starting dates, timelines and objectives for all


sub-projects with regard to logistical KPIs and financial benefits. A transparent definition of objectives is necessary in order to avoid the implementation of “nice-to-have” technical features and modules, which are not business driven in the end. The Project Masterplan should clearly define the scope of each sub-project in terms of geography, products, business processes covered and users involved.

2.4.3 Vendor Selection

The vendor selection step starts by defining the software requirements in detail. The selection process itself includes a provider long list and short list. For the final decision, a detailed scoring matrix should be applied.

Definition of requirements

The definition of the requirements is a critical step in the selection process. Jehle (2000) proposes a detailed gap analysis between strategic targets and the actual performance of the business processes that are subject to the APS implementation. Many consultancies have developed their requirements based on comparisons with collected best practices for each business process (Salehi 2003). Gronau (2001) argues that all requirements should be assessed and only the most important ones should be ranked with an “A” (absolute prerequisite) to allow the providers to eas-ier check the criteria. Besides the technical aspects, the list should also include re-quirements concerning the vendor (financial health, reference clients etc.).

Vendor long list and filter

The vendor long list contains providers that could potentially be qualified to de-liver an APS system or module containing the required functions. Market studies, trade fairs, or analyst reports (for instance AMR Research, Gartner Group or Fraunhofer Gesellschaft) can provide support for the definition of the long list. Fandel and François (1999) and Treutlein and Kipp (2002) suggest using a soft-ware vendor database that contains the coverage of requirements of specific sys-tems.

As the requirements for APS systems differ considerably across industries, al-ready at this stage of the selection process it should be considered whether the vendor has experience in the specific industry (Tiemeyer 1999). Following this step, a high-level request-for-information is prepared based on the “must-have” requirements that have been defined before and sent out to the providers (Davies et al. 2002). The answers are evaluated mainly if to whether or not all “must-have” requirements have been covered. Moreover, indications on industry and functional focus, expertise of the provider and pricing are considered to select the short-listed providers.

Vendor short list and decision

For the short-listed providers, a detailed request-for-proposal is prepared which is usually built around the following points (Davies et al. 2002):


Functional specifications, Technical specifications (reliability, adaptability, maintainability, secu-rity),Training & support, Vendor profile (financials, experience, reference clients, R&D), Pricing for licenses, implementation, training etc.

The answers should be evaluated using a scoring matrix. Kilger (2002b) pro-poses the use of a 1-5 scale ranging from “function not available” to “function fully available” to rate the functions offered. The most interesting providers are usually invited for a presentation and product demonstration. Within this live demonstration, the required functions can be evaluated in detail. In some cases, appointments with reference clients are scheduled in addition. For the final selec-tion decision, the most important stakeholders in the corporation should be in-volved in order to generate broad buy-in.

2.4.4 Implementation

Once the APS software has been selected, the implementation can start. First, the details of the proposed solution are defined and a pilot system is developed. After a period of testing and improvement, the system is rolled out within the company.

Detailing and pilot

The detailing phase typically starts with a review of the solution design in order to gain an understanding of how the organization will be affected, and to make sure that the required functions of the software are really available on a detailed level (Wetterauer 2002). Furthermore, the integration of the new system into other, re-lated processes is an important issue because the APS system receives and returns data from and to many other processes (Hellingrath et al. 2001; Kilger and Müller 2002). Decisions have to be made concerning which integration mode and which integration technology should be applied. For the integration mode, the full data upload can be distinguished from the netchange of the data.

With regards to the integration technology, some providers offer their own techniques, for example SAP for the integration of its Advanced Planner and Opti-mizer (APO) and R/3; others rely on middleware-products like COBRA (Kilger 2002b). In the next step, the underlying processes of the corporation and the nec-essary data are modeled in the system in order to customize the system to the spe-cific requirements of the company. Persisting functional gaps should now be re-solved, either by developing additional functions or by adapting corresponding business processes.

Testing and improvement

Once the pilot is developed, it is applied to some sample processes to get a deeper understanding of how the solution works and how it can be improved further. On the other hand, it can also become necessary to further change the business proc-


esses at this stage. Intensive testing is a prerequisite for the new system before “going live” (Grunow et al. 2002). A formal test-plan is especially important with regard to the final approval of the software by the management of the company (Wetterauer 2002). The testing should also include a comparison of the output of the APS system and the plans currently generated by the planners. This allows not only to fine-tune the planning parameters, but also to increase the software accep-tance within the organization. Finally, the end-user trainings and the according training plans are developed and the roll out plan is finalized.

Roll-out

In the rollout stage of the project, the improved IT systems and the new processes are deployed to other departments, sites, SC partners or products (Hellingrath et al. 2002). In the case of multi-site companies, site-specific adaptations are made during this phase. The integration of the software has to be accompanied by inten-sive training of the users, which should address the software as well as the new processes. At the end of the rollout, a team has to be established to manage all is-sues after “going live” (Wetterauer 2002).

2.4.5 Implementation Risks

Although the payback times of APS systems are generally relatively short (6-12 months versus 2-5 years for ERP systems), they vary considerably depending on the case (Kortmann and Lessing 2000). Therefore, managing the following im-plementation risks is crucial in making the implementation successful:

Lacking executive management sponsorship and resistance to change are the most frequently mentioned reasons for implementation failures (e.g. Davies et al. 2002; Kilger 2002a; Mekschrat 2002; N.N. 2002c). Careful and ongoing communication is crucial to ensure adequate executive man-agement support. Involving the future users from the very beginning can avoid resistance to change. Bick (2002) stresses the training aspect not only to educate the users but also to “promote” the system. Besides the involvement of the end users, the support of the APS pro-

vider is an important aspect as well. The availability of data is a key success factor for the implementation (Hellingrath et al. 2001; Schmelmer and Seiling 2002). In numerous pro-jects the collection, correction and completion of the data in the underly-ing ERP systems is a task underestimated at the project start. Another common reason for implementation failures is a weak project

management (Davies et al. 2002; Franke 2002). The project manager has to ensure that people with the necessary skills and authority are available and that milestones and budgets are met. Properly defining and tracking of KPIs is necessary to document the success and to maintain the focus of the project.

2.5 Assessment of APS Implementations 33

Many enterprises try to accelerate the implementation process in an at-tempt to receive a higher and quicker ROI by skipping important steps in the Project Definition and Vendor Selection phases (Davies et al. 2002). This often has severe consequences and puts the entire APS implementa-tion at risk.

2.5 Assessment of APS Implementations

2.5.1 Benefits

Numerous articles and book contributions that describe case studies of APS im-plementations and the resulting benefits are available by now. SAP APO is the most frequently cited software (e.g. Bothe 1999 and 2000; Haug 1999; Hurtmanns and Packowski 1999; Grünewald 2001; Bick 2002; Franke 2002; Mekschrat 2002; Nicolai 2002; Richter and Stockrahm 2002), while other reported implementations concern i2 (e.g. Heidrich 2002; Kilger and Schneeweiss 2002b; Schmelmer and Seiling 2002; Schneeweiss and Wetterauer 2002) or other software packages (e.g. Kansky 1999; Tappe and Mussäus 1999; Kirch and Manger 2001; Wagner and Meyr 2002). In the majority of the described cases, the implementation of an APS system has provided significant benefits for the company. Kortmann and Lessing (2000) give a summary on realized quantified benefits. The stated improvements concern SCM in general and not APS systems specifically. However, considering APS systems as an enabler for SCM, the numbers can be used as estimation for savings resulting from APS implementations. Regarding the total cost base of an SC, an improvement potential of 3 to 25% with an average of ca. 10% is men-tioned. Other improvements concern lead-times, inventories, and productivity. In addition to these quantifiable indicators, Davies et al. (2002) name several other intangible benefits of APS systems:

Alignment of various plans existing within a corporation, Increased availability of critical materials and components, Improved feasibility of plans, Increased sales through higher service levels and responsiveness, and More focused planning, planners can concentrate on exceptions and bot-tlenecks.

Although, from a scientific perspective, the realized benefits are assessed less enthusiastically than in the case studies or in the provider’s sales brochures, most researchers concede that significant progress has been made. Corsten and Gössinger (2001a) see major progress in the fact that the decentralized plans are matched roughly and that several companies have access to integrated data. The higher transparency enables to identify and reduce unnecessary safety stocks and to accelerate administrative processes. For Günther and Tempelmeier (2000), the most important benefit is the fact that several sites and production segments can


be integrated within the SNP module, which constitutes a significant advantage compared to the current level of planning in most companies.

2.5.2 Development Needs

Although important benefits have been realized, all authors agree that APS sys-tems have to be further developed to overcome current weaknesses. Tempelmeier (2001) enumerates several currently unsolved problems. First of all, lot sizing is still supported in an unsatisfactory way – if at all. Furthermore, he considers the calculation of derived demand as an important area of improvement since this is often done in the same way as in the standard MRP systems. Also, node design is still not covered sufficiently by APS systems. Typical problems in this field are factory layout design or assembly line balancing. Finally, the integration of sto-chastic features is neglected completely by most systems.

In addition, APS systems offer only limited support to analyze the robustness of a solution. Goetschalckx (2002) names the design of flexible and robust supply chains one of the major research trends with respect to the SND module. The ro-

bustness of a production control system can be defined as “its ability to absorb disturbances” (Biennier and Favrel 1998). An optimal solution provided by the planning algorithm is called robust if it “remains close to the optimal solution if

input data change” (Leung and Wu 2004). Van Beek (2002) defines a supply chain as robust if its design does not have to be adjusted in case of changing inter-nal or external circumstances. In particular on longer-term planning horizons, the robustness of a solution is critical as the long-term data the models are based on incorporate a high degree of uncertainty and the decisions which are taken on this planning level cannot easily be revised. The support of the design of robust solu-tions by APS systems should include for example stochastic features, a sensitivity analysis for all variables, and guidelines on how to build and evaluate scenarios. Examples on how to obtain robust solutions are provided by Goetschalckx et al. (2002) for the strategic planning level, by Leung and Wu (2004) for the tactical planning level and by Biennier and Favrel (1998) for the operational planning level. Knolmayer (2001a) and Dudek et al. (2002) complain that only few provid-ers give information on the methods and algorithms incorporated into their sys-tems. As the power of the optimization engines in the APS systems cannot be as-sessed, it is rarely an important decision criterion for the selection of an APS system. To use the methods effectively, the user must have in-depth-knowledge concerning both, the specific situation of the company and modeling techniques. Therefore, special trainings for the planners are still necessary (Stadtler 2002).

Although the APS providers accentuate the inter-company planning aspect of APS systems, most companies do not yet use the systems for this purpose (Corsten and Gössinger 2001b). Most case examples refer to an inter-site planning within one company. Busch et al. (2002) and Steven and Krüger (2002) emphasize that the current generation of APS systems is only less suited to support a heterarchic SC because the coordination of different decision makers is seldom supported. Therefore, many providers seek to enlarge their collaboration capabilities Lange-

2.6 Conclusion 35

mann (2002) analyses the integration of collaborative functions into different APS modules. Regarding the IT landscape, main obstacles to an intensified collabora-tion are problems with interfaces and lacking standardization of processes (Baumgarten and Thoms 2002).

2.6 Conclusion

APS systems constitute a significant progress compared to incumbent MRP sys-tems, in particular in the field of decision support (Steven and Krüger 2002). Due to the complexity of the total planning problem, they hierarchically decompose the entire problem and use optimization methods to solve partial planning problems within single modules. The results of a higher planning level are then require-ments for the lower level, and in that way a coordination of the planning steps is achieved (Kampker and Wienecke 2001). However, an optimal solution for the entire problem cannot be obtained. The ERP system remains the backbone of the IT landscape, because a tight integration between the APS and ERP system is re-quired to fully profit from the APS technology (Corsten and Gössinger 2001b).

Regarding the further development of APS systems, the necessity to develop specific decision models for each production segment and each planning step is emphasized in many sources (e.g. Mayr 1996; Tempelmeier 1999a; Meyr et al. 2002b; Günther and van Beek 2003). It is crucial to examine the requirements of specific industries and how the software can respond to these needs. This is par-ticularly the case for the PP and PS modules, which have to be adapted to the spe-cifics of different production segments (Günther and Tempelmeier 2000; Meyr et al. 2002b). MacCarthy and Fernandes (2000) classify different production systems for the selection of an production planning and control system. Dudek et al. (2002) have underlined that generic models should be built into the standard software packages that should then be customized to the individual planning problem and that the solution algorithms must be chosen accordingly. Therefore, the analysis of specific requirements of fresh food industries is the focus of Chapters 3 and 4. Upon completion, industry-specific solutions with special attention paid to the PP and PS modules are developed in Chapters 7 to 9.

3 Fresh Food Industries

3.1 Introduction

The purpose of this chapter is to outline the common characteristics of fresh food industries as a basis for the definition of requirements on APS systems in Chapter 4. At the beginning, a definition of fresh food industries is given and the most im-portant segments are presented. Thereafter, the characteristics and recent devel-opments of FFSC and fresh food production systems (Chapters 3.3 and 3.4) are analyzed, as they must be taken into account when selecting an APS system. Fi-nally, the production systems of the three case studies (yogurt, sausage and fresh poultry) are presented in detail (Chapters 3.5 to 3.7), since these case examples will be used in Chapter 4 to demonstrate how specific industry characteristics can be integrated in order to develop an APS profile. Furthermore, the developed MILP models that integrate shelf life into operational production planning (Chap-ters 7 to 9) rely on these production processes as well.

3.2 Definition and Segments

In contrast to the food industry in general, a clear definition for fresh food indus-tries does not exist. The understanding of this research is that fresh food indus-

tries are industries that primarily produce food products with a short shelf life. A shelf life is considered “short” if it is in a range from several days up to 2-3 months. Therefore, dairy and bakery products, fresh and processed meat, fruits and vegetables as well as fish are considered fresh food industries in this analysis. Although some of the products produced by these industries cannot be regarded as “fresh” (e.g. frozen meat or UHT milk), the major part of the output of these in-dustries is sold as fresh products.

The turnover of fresh food industries achieved ca. € 52 bn. in Germany in 2000 (see Fig. 3.1) and therefore over 50% of the entire value of all food industries (ca. € 98 bn. in 2000) and ca. 5% of the German Gross Domestic Product (Wünsche 2002). The most important fresh food segments are dairy products (€ 15.5 bn. product value in 2000), followed by bakery products (€ 12.0 bn.) and processed and fresh meat (€ 9.8 bn. and € 8.1 bn.). The entire product value of German fresh food industries has remained relatively stable during the past two years (€ 52 bn. in 2000 versus € 49 bn. in 1998).

38 3 Fresh Food Industries

Fig. 3.1 Value of products of German food industry (data from Lebensmittel Zeitung 2001)

The case studies presented in this work cover three out of the six fresh food segments (dairy, processed meat and fresh meat). Nonetheless, the results of the case studies can also be transferred to the segments “Bakery” and Fish”, as these production systems are quite similar. In contrast, the category “Fruits and Vegeta-bles” can only partially be covered by the case studies as it shows relatively spe-cific characteristics with regard to the production system. On the one hand, inten-sive processing of the raw material is not required for these products (unless they are preserved or canned, but in that case they cannot be considered as fresh any more); on the other hand, packaging is less important (frequently no packaging at all or simple packaging materials such as nets etc.). However, when looking at the product values of each of the categories (see Fig. 3.1), the three case studies are representative for ca. 64% of the fresh food segments in 2000.

3.3 Characteristics of Fresh Food Supply Chains

3.3.1 Structures of Fresh Food Supply Chains

Before characterizing FFSCs, it is necessary to define the terms “Supply Chain” and “Supply Chain Management” as there is no uniform understanding of these terms neither within the industry nor within the academic world – despite intense discussion and research (Busch et al. 2002). Many different variations of the same theme are offered when defining these two terms (Ganeshan et al. 1999). Several authors provide overviews of definitions that can be found in literature. Among the authors that analyze SCM research (e.g. Kotzab 2000; Pfohl 2000; van der Vorst 2000; Arnold and Warzog 2001; Grünauer 2001; Göpfert 2002), Ganeshan et al. (1999) are to be mentioned who give a comprehensive taxonomy on SCM research, show the evolution of SCM over time and classify the research accord-

8.17.47.0

9.4 9.3 9.81.4 1.4 1.4

5.1 5.24.8

15.515.215.4

11.0 11.412.0

3.43.33.9

8.9 8.99.0

3.13.03.0

3.9 3.94.12.9

3.23.9

7.1 7.27.0

6.36.46.0

2.21.9

2.1

8.27.77.9

1998 1999 2000

3.3 Characteristics of Fresh Food Supply Chains 39

ing to the solution methodology applied. As this research is primarily concerned with the planning, coordinating and controlling of the SC, the working definitions of this research of a SC and SCM also focus on these aspects.

A Supply Chain is defined as “a connected series of activities, which is con-cerned with planning, coordinating and controlling materials, parts, and finished

goods from supplier to customer. It is concerned with two distinct flows (material

and information) through the organization” (based on Ganeshan 1999). Supply

Chain Management is “the integrated planning, coordination and control of all

material and information flows in the SC to deliver superior consumer value at

less cost to the SC as a whole whilst satisfying requirements of other stakeholders in the SC” (based on van der Vorst 2000). Some authors include also a financial dimension into their SCM-definitions, for some examples it is referred to Grünauer (2001).

The term “Supply Chain” was coined in 1982 by Keith Oliver of Bain Consult-ing (Heckmann et al. 2003) and emerged around 1993 to a broader business com-munity (Zuckerman 2002). Various disciplines have contributed concepts that originated outside the original SCM theory but are used throughout the SCM lit-erature (Ganeshan et al. 1999). Arnold and Warzog (2001) name the field of Pro-curement as central origin of SCM, which developed from the concept of materi-als management over integrated procurement to integrated logistics that finally led to SCM. Marketing contributed aspects such as postponement (Ganeshan et al. 1999; for the analysis of the effects of postponed manufacturing on the supply chain of consumer goods it is referred to van Hoek 1998) or distribution policies and channel management research (Kotzab 2000). Furthermore, a central element from Economics and System Dynamics is Forrester’s (1958) research on the so-called “Bullwhip-effect” (see for example Lee et al. 1997). In OR and Operations

Management, several roots can be found such as multi-echelon inventory models, plant and distribution center location models, order allocation schemes, lean manufacturing, just-in-time supply, Vendor Managed Inventory (VMI), and many more. Integrated logistics and partnership building are elements from the Logisticsworld (Ganeshan et al. 1999). Finally, the evolution in IT can be seen as an en-abler for many of the cited concepts and ideas (Beierlein and Miller 2000).

Fig. 3.2 Examples of fresh food supply chains (based on Meulenberg and Viaene 1998)


Looking specifically at fresh food industries, the SC proceeds generally from the farmer to the consumer and may involve one or several additional actors (van Wezel 2001). Meulenberg and Viaene (1998) give some basic patterns of food SCs (see Fig. 3.2). The direct SC from the farmer to the consumer (1) or via a middleman such as a trader (2) has been the dominant FFSC for centuries. How-ever, today this is the exception rather than the rule in developed countries. In many fresh food industries, a SC of type (5) is predominant. Farmers receive feed or seed from industrial producers and deliver their products (e.g. animals or milk) to the processors. Usually, two types of industrial processors can be distinguished (van Donk 2001). The first processor transforms natural materials into intermedi-ate products. Examples for this type of processors are abattoirs in the meat indus-try. On the one hand, they deliver some of their products (e.g. fresh meat) directly via wholesaler and retailer to the customers. On the other hand, they are a supplier to the processor of consumer products (e.g. sausage or ham producer in the meat industry). Retailers usually organize the delivery to the consumer.

In many fresh food industries, a SC of type (5) is predominant. Farmers receive feed or seed from industrial producers and deliver their products (e.g. animals or milk) to the processors. Usually, two types of industrial processors can be distin-guished (van Donk 2001). The first processor transforms natural materials into in-termediate products. Examples for this type of processors are abattoirs in the meat industry. On the one hand, they deliver some of their products (e.g. fresh meat) di-rectly via wholesaler and retailer to the customers. On the other hand, they are a supplier to the processor of consumer products (e.g. sausage or ham producer in the meat industry). Retailers usually organize the delivery to the consumer.

Two important types of vertical cooperation occur in an FFSC. First, due to in-creasing quality concerns of the consumers, a growing number of agricultural commodities are produced under non-market arrangements such as contracts, joint ventures, strategic alliances, cooperative organizations, or integrated forms of ownership. For example, 75% of the German dairy industry is organized in coop-eratives that are controlled by farmers (Murmann and Wolfskeil 2004). Secondly, the management and control of the entire chain shifts from the industry to the re-tailer. While historically the driver of change was located at the food manufactur-ers (similar to the car assembler in the automotive world), retailers are now the major catalyst for change in the chain which is mainly due to their high level of concentration and their resulting bargaining power (Galizzi and Venturini 1999).

Trienekens and Omta (2002) stress that problems and opportunities in FFSCs must be addressed from a multi-disciplinary and farm-to-table perspective. Effi-cient and effective decision making in FFSCs must be based on a holistic view of the SC. Therefore, the following analysis of the major characteristics and devel-opments in FFSCs (Chapter 3.3.2 to Chapter 3.3.5) does not only cover economic and technological aspects, but additionally addresses social / legal and environ-mental issues (see Fig. 3.3). In this context, van Beek (2002) refers to the “Triple-P”-view on a SC: Profit (economic view), Planet (ecologic view) and People (so-cial view).


Fig. 3.3 Elements of fresh food supply chains (based on Trienekens and Omta 2002)

3.3.2 Economic Characteristics and Developments

There have been several important changes regarding the economic situation of FFSCs during the past years. These developments are portrayed in detail in the following paragraphs. All of them lead to a growing pressure on cost and margins of all participants in the FFSC. Many actors yield margins of only 1-2% of their turnover (Poirier and Reiter 1997). Bourlakis and Weightman (2004) report that today food prices in the Western world are probably the lowest they have ever been compared to the average salary. Therefore, cost reduction has been a pre-dominant issue in FFSCs throughout the recent years.

Consolidation

A clear trend towards consolidation can be observed in all fresh food industries and in all stages of the SC. At farm level, the number of farms is steadily decreas-ing while the average farm size is increasing (Meulenberg and Viaene 1998). For example, the number of dairy farms in the Netherlands decreased from 67,000 in 1980 to 36,000 in 1997 (Frouws and van der Ploeg 2000). The concentration proc-ess is accompanied by a growing specialization of the farms (e.g. milk producer or poultry grower) and an increasing degree of vertical cooperation with wholesalers or manufacturers (Meulenberg and Viaene 1998). At the manufacturer’s level, fresh food industries (in particular meat, bakery and dairy) are still characterized by mid-sized companies (Wünsche 2002). As economies of scale are substantial in most fresh food industries (Connor and Wills 1988), concentration is likely to ac-celerate (KPMG Corporate Finance UK 2000; Bourlakis and Weightman 2004). Major concerns for fresh food processors are the overcapacities in the market. For example, Auer (2001) estimates the overcapacities in the German meat industry to be around 30-40% resulting in a tough competition.

The continued consolidation of retailers is perceived as the biggest threat to the manufacturers (KPMG Corporate Finance UK 2000). In 2000, the five leading re-tailers in Germany realized 63% of total retail turnover, 18%-points more than ten


years ago (Michael et al. 2002). In the UK, the share of the top five retailers is al-ready more than 80% (van Wezel 2001). This retail concentration leads to a sig-nificant shift of power in the FFSC. Retailers will increasingly command the re-tailer manufacturer interface (McLaughlin 2002) or even the whole upstream part of the chain – with severe consequences to the manufacturers. Most important, re-tailers may press manufacturers to lower prices. Their incentive to do so is high: As in retailing the costs of goods purchased amount to over 80% of the total sales volume, a reduction of these costs can significantly raise profitability (Dawson 2004). “Every Day Low Price” strategies are more and more adopted among large retailers (KPMG Corporate Finance UK 2000; Hughes 2004). “Slotting Allow-ances” (US term) or “Listing Charges” (European term) which are fees paid by the supplier to gain access to retail DCs and stores are expected to increase further (McLaughlin 2002). In addition, intense promotional support is required for branded products. Therefore, this increasing power of retailers could jeopardize the innovative power of manufacturing companies.

Internationalization

Internationalization is another development for both producers and retailers. Some fresh food companies such as Nestlé or Danone have always been operating on an international scale. Other companies that have traditionally been oriented to do-mestic markets have become increasingly international in the past years. Examples are MD in Denmark or Campina Melkunie in the Netherlands (Meulenberg and Viaene 1998). The EU enlargement in 2004 pushes internationalization further as trade barriers are falling and wage differences are still significant (Hoffmann 2004a and 2004b). This is particularly true for the meat industry as labor involve-ment in this industry is high. However, also in the dairy industry major foreign di-rect investments in the new EU member states of European manufacturers are ex-pected in the near future in these countries (Murmann 2004b). Imports from the new EU member states will increase pressure on prices even more (N.N. 2004c).

Although most retailers still mainly rely on their region of origin, they also en-force their international efforts (Vandenheede 2002). Prominent examples are the current European expansions of the discount stores Aldi and Lidl. The emergence of global retailers represents a new challenge for fresh food manufacturers. Euro-pean and global sourcing as well as European pricing make it harder for manufac-turers to raise revenue (KPMG Corporate Finance UK 2000). Moreover, an exis-tential threat for smaller producers is that many retailers are consolidating their supplier base. Delivering products on a European scale will be a major supplier se-lection requirement in the future. As today most parts of Europe can be reached within 24 hours by truck, European sourcing is also relevant for fresh products with limited transportation times.

Rise of private labels

Many retailers have introduced their own brands (so-called “private labels”) with remarkable success as alternatives to manufacturer’s A-brands and B-brands (van Wezel 2001; Twardawa 2004). Examples in Germany are “ja!”, “gut und billig”, “tip”, or “A&P”. Private labels are usually lower priced, but carry relatively high


margins for the retailer because these products are bought at significantly lower purchasing costs (Meulenberg and Viaene 1998). These products are often pro-duced in the same facilities as branded products, but spread over multiple manu-facturers to easily exchange the producer (van Wezel 2001). Retailers have very strict requirements concerning product quality and characteristics that must be re-spected thoroughly by the manufacturer if he wants to avoid being de-listed.

As consumer confidence in these products has increased, private labels are cur-rently replacing even premium brands (KPMG Corporate Finance UK 2000). Many industry professionals believe that several retailers will exclusively sell pri-vate labels in the future (Michael et al. 2002). The share of private labels varies clearly between countries and categories. While the UK is probably the most ad-vanced in the evolution of private labels with a market share of over 45% (Henson and Northen 1999; van Ossel 2002), this share is much lower in countries like It-aly with ca. 10% (Meulenberg and Viaene 1998). With regard to specific catego-ries, private labels account for the highest share in fresh products, e.g. 45.5% in 2000 in Germany (Lebensmittel Zeitung 2001). An important reason for this is that fresh products have several properties (such as a diminishing product quality or limited shelf life) that make branding more difficult for the manufacturer than for dry groceries. As today many fresh products are not branded at all and retailers have the highest influence on these properties, it is likely that the share of private labels in fresh foods will increase further (Grievink et al. 2002).

Growing importance of the discount channel

The rise of private labels is closely related to the emergence of a specific retail channel, the discount (Otto 2004). This holds particularly true in Germany where companies such as Aldi, Lidl, Plus, or Netto account for 44.6% of total food retail-ing in 2002 (see Fig. 3.4). Discounters show several typical characteristics (Eggert 1998):

Limited assortments: For example, Aldi has a portfolio of only around 750 SKU, 80 out of which are fresh products like dairy or meat products (Brandes 1999); High volume per article and high inventory turns (8 days on average at Aldi);Concentration on only few suppliers; Price leadership in most categories; High share of private labels; Low cost structure in terms of personnel cost or store design; and Strong focus on product quality.

Discounters impose specific requirements on food manufacturers. Processors must be able to deliver high volumes of products with excellent quality at lowest prices. Contracts with suppliers are usually long-term. The quality of products is ensured by intensive audits carried out by discounters at sites of the manufactures.


Fig. 3.4 Food retailing in Germany (data from Lebensmittel Zeitung 2003)

Outsourcing

Outsourcing can be observed for several functions, but it is particularly important with regard to transportation logistics (see for example Dik et al. 2003). Le-bensmittel Zeitung and PwC Consulting (2002) as well as McKinnon (2004) re-port that most food manufacturers and retailers have already outsourced their transportation logistics. Inventory management or commissioning and packing of promotion-packs are subject to outsourcing as well. Third Party Logistics Provid-ers (3PL) are becoming a dominant player in the SC (Bourlakis and Bourlakis 2004) as consolidation takes place also in this industry. In addition, the degree of IT-outsourcing is expected to grow significantly in the years to come and finally, most food chain executives even expect a higher degree of outsourcing in produc-tion (Grievink 2002). There are two essential aspects that drive outsourcing. First, companies increasingly focus on their core business. By outsourcing specific func-tions to contracting partners that can deliver the outsourced services more effi-ciently, firms aim at achieving a competitive advantage. Secondly, specialized contractors have often an improved cost structure in terms of personnel cost as they can conclude other wages agreements. A close connection between the food processor or the retailer and the contracting partner, especially regarding the IT systems, is prerequisite in order to enable seamless business processes.

Decreasing lead-times

As a consequence of the change in the power balance between retailers and food manufacturers, retailers demand a higher logistical performance of the manufac-turer and restructure their SC accordingly. Retailers increasingly require frequent replenishment at short notice, resulting in shorter lead-times for the manufacturer

27.1 30.6 33.6 35.7 38.844.6

33.532.9 32.1 31.4 30.1

30.0

25.626.9 27.8 29.0 29.8

30.4

23.120.4 18.8 18.0 17.1

15.4

120.4115.8114.1112.3110.8109.3

1992 1994 1996 1998 2000 2002


and lower retailer stocks (van Wezel 2001). As the average retail order size will decrease, food manufacturers must either decrease their average batch size in pro-duction or take over a higher share of the SC inventories. Both leads to higher costs for the manufacturer. In many cases, a proper management of these short de-livery times can only be guarantied using a software support such as route plan-ning or other systems (N.N. 2003a). An example for these decreasing lead-times from the Dutch retailer Albert Heijn is given in Table 3.1. McKinnon and Camp-bell (1998) observe a similar development for frozen food SCs as well as Perosio et al. (2001) for fresh produce. The numbers are confirmed by the German retailer Globus that currently has a lead-time from DC to store of six hours resulting in an inventory decrease of 20-25% (N.N. 2003a). Accordingly, the average delivery frequency increased significantly. McKinnon (1999) shows that the average deliv-ery frequency increased from 2.1 in 1995 to predicted 4.3 deliveries per week in 2001 in the UK.

Table 3.1 Lead-time from supplier to DC and from DC to store (based on van Wezel 2001; taken from Willemse 1996)

Lead-time Supplier / DC DC / Store

Past 120-48 hours 48-36 hoursPresent 48-24 hours 18-12 hoursFuture 12-4 hours 18-4 hours

Efficient Consumer Response (ECR) and Collaborative Planning, Forecasting and

Replenishment (CPFR)

ECR is a strategy that focuses on developing closer relationships between retailers and manufacturers. It aims at developing a responsive, customer-driven approach to maximize consumer value, to minimize SC cost (Hieber 2002), and to increase process efficiency (Rode 2004f). ECR is a specific form of SCM in the consumer goods industry (Hill 2000) and comprises a bundle of concepts and measures (Stieglitz 1998). As shown in Fig. 3.5, the basic strategies of the ECR concept can be divided into a supply side (SCM) and a demand side (Category Management).

Efficient replenishment (also called continuous replenishment) probably consti-tutes the most important base strategy within the supply side part. The objective of Efficient Replenishment is to generate an efficient production and distribution based on the demand at the Point-of-Sale (POS; Seifert 2001) in order to avoid the “bullwhip” effect (Forrester 1958). It is of vital importance that the manufacturer gets access to the original, non-biased POS data (Jensen 1999; Baumgarten and Darkow 2002). One of the pioneers in this respect, the largest German retailer Metro Group, recently started its extranet “Metro Link” allowing the suppliers to see the inventories of their products down to article and outlet level (Rode 2004c). However, so far only ca. 20% of all retailers provide their manufacturers with sales data (N.N. 2003d). The tightest form of cooperation constitutes Vendor Managed Inventory (VMI), in which the responsibility of inventories is transferred to the prior stage in the SC (Baumgarten and Darkow 2002).


Fig. 3.5 Concept and basic strategies of ECR (based on Seifert 2001)

On of the most important targets of Efficient Replenishment is the reduction of Out-of-Stock rates. Stock availability is no longer a question of operational effi-ciency. It becomes a key strategic issue for branding and positioning (Ody 2002), as many consumers stop shopping entirely if the desired product is not available. Several authors give estimations on the size of stock-outs. Hausruckinger and Pe-lousek (2003) estimate the value of out-of-stocks between 7-10% of the revenues, Fairfield (2002) mentions a value of 6,5% in terms of revenue and 8,2% in terms of articles, and Stölzle and Placzek (2004) even name 9-11% in terms of articles. In the case of stock-outs, due to the low degree of product differentiation, the revenue is almost lost for the manufacturer; only 26% of consumers buy the same product in another store (see Fig. 3.6). According to Seifert (2001), Efficient Ad-ministration stands for the cooperation in the field of administrative processes be-tween retailer and manufacturers. Besides the creation of performance-oriented terms and condition systems, it comprises the administration of data and informa-tion. Therefore, it is closely linked to Electronic Data Interchange (EDI) and its re-lated concepts.

Fig. 3.6 Stock-out behavior of consumers (based on Seifert 2001; taken from Roland Ber-ger & Partner 1999)


Finally, Efficient Operating Standards aim at introducing common operating standards such as:

Cross Docking: The supplier commissions the pallets so that they can be delivered directly to the retail outlet. In the retail DC, the pallet can then be handled without being stocked. In the case of transshipment, the commissioning takes place at the retail DC (Baumgarten and Darkow 2002).Roll Cage Sequencing: Special roll containers are commissioned with products in the order as required later in the retail outlet (Seifert 2001). Efficient Unit Loads: Unit loads are designed to improve the efficiency of the process (e.g. better pallet utilization or multi-temperature trucks).

On the demand side, the base strategies are summarized under the term Cate-gory Management, which is a joint process between retailer and manufacturer to manage product categories as strategic business units (Seifert 2002a). The most important building block on the demand side is the Efficient Store Assortment, which aims at defining the optimal product offering within a category in a coop-erative way between retailer and supplier. Two additional areas of cooperation are promotions (Efficient Promotion) and new product development and introduction (Efficient Product Introduction). Category Management has found broad reception in the industry (for a detailed analysis see Seifert 2001).

The CPFR concept that has been pushed by the Voluntary Interindustry Com-merce Standards Association can be regarded as a further stage of ECR. While ECR is primarily retail-driven, CPFR is a mutual approach of retailers and manu-facturers. It has the goal to develop integrated business processes and supporting technologies for collaborative forecasting and planning (Hieber 2002), hence it is a standard methodology for two or more companies to work together (Bhambri N.N.). For the implementation of CPFR, nine planning steps are proposed, ranging from an outline agreement between retailer and manufacturer up to the generation of orders. Electronic marketplaces are a main catalyst for the implementation of CPFR, as - in contrast to EDI - e-marketplaces represent a many-to-many relation-ship with high efficiency advantages (Seifert 2002b). Numerous case studies document the benefits of CPFR (e.g. in the fields of forecast accuracy, inventory reduction, delivery reliability and accuracy, lead-times, transportation and distri-bution costs, or personnel productivity; see for example Bastock et al. 2002; Brenchley 2002; Fraser 2002; Kapell 2003).

3.3.3 Technological Characteristics and Developments

The tremendous progress in technology has been the primary driver and enabler for many other developments in FFSCs such as the realization of ECR or CPFR concepts. The most important developments of IT in FFSCs are presented in this chapter. Developments in the process technology of fresh foods are individually analyzed for each of the case study industries in Chapter 3.5 to Chapter 3.7.


Electronic Data Interchange

EDI is the predominant form of IT to coordinate several companies in a SC in the food industry (Hill and Scudder 2001). In particular, retailers are leaders in im-plementing EDI (Hill 2000). EDI is a computer-to-computer exchange of business documents in a standard format (Handfield and Nichols 1999) and focuses primar-ily on ensuring swift, accurate and less redundant commercial transactions (ARC Advisory Group 2000). It relies on two international standards. Concerning the data content, EDI applies the European Article Number (EAN) that allows to un-ambiguously identifying an article (Centrale für Coorganisation 1999). The EAN code is usually put on the products as a barcode and can be read with a scanner. Over 600,000 companies use the EAN worldwide (Seifert 2001). To label ship-ment units, the so-called EAN 128 barcode is used that can additionally represent data on e.g. shelf life or the batch number. Other types of standardized data in the food industry include the number of the shipment unit and the international loca-tion number (Centrale für Coorganisation 1999). Regarding the data structure, EDI is based on the EANCOM standard (“EAN” plus “COMmunications”), which is a consumer industry specific subset of the EDIFACT-standard (Electronic Data In-terchange for Administration, Commerce and Transport; see for example Schmidt 1997; Fraede 1998). It includes numerous standard types of messages such as or-ders, invoices, or receipts (Seifert 2001). EDI offers the opportunity to process scanner data of the POS and to send it to the manufacturers (Kinsey 2000). It co-operates tightly with Computer Assisted Ordering systems that automatically gen-erate replenishment orders in the retail outlet based on inventories and sales (Baumgarten and Darkow 2002).

A large number of studies show that EDI has become broadly accepted in the food industry. Rode (2003b) reports that 87% of all CPG manufacturers use EDI technology, and a further 8% are currently implementing it. Schraft et al. (2001), Lebensmittel Zeitung and PwC Consulting (2002) as well as Hill and Scudder (2002) show similar figures. Today, many retailers expect fresh food manufactur-ers to support EDI communication. Often it is even a prerequisite for a delivery agreement; hence also many smaller manufacturers use EDI (Felger 2004). None-theless, EDI is not suited to provide adequate support for SC collaboration since the defined transactions are limited, concern primarily commercial aspects, and are not real time (ARC Advisory Group 2000).

E-marketplaces

E-marketplaces are a virtual trading area for business transactions. They support all activities to coordinate the exchange processes and are mainly used as a trans-action platform for electronic trade (Kleineicken 2002; Nicolai 2002). Transora, World Wide Retail Exchange (WWRE) and GlobalNetExchange are the three ma-jor marketplaces that have been established in the CPG industry. Other market-places are operated by retailers (e.g. Wal-Mart Extranet or the ReWe Lieferanten Partnernetz ReLiPa). Industry executives see marketplaces as a new and cheaper way to exchange information (Grievink et al. 2002), as EDI requires compatible computer systems that are expensive to set-up and operate (Kinsey 2000). While


EDI costs about $ 150 per hour, internet-based systems are only about $1 per hour (Hauptmann and Zeier 2001). As E-marketplaces are a relatively new instrument in FFSCs, it is not yet clear if they will become the future industry standard.

Radio Frequency Identification Tag (RFID)

Many industry professionals regard RFID technology as a “killer-application” for SCs in CPG. RFID allow products to be tracked via a wireless chip or tag from the factory to the shelves in the store and even beyond (Davis 2002a). This chip can be seen as a “speaking” bar code. It is very likely that RFID technology will be-come an industry standard in the near future, because the development and appli-cation is pushed by the world’s leading retailers, e.g. Wal-Mart or Metro Group (Rode 2004e) and powerful CPG manufacturers such as Unilever, Nestlé, or Kraft Foods (Rode 2003a). The most important benefit of RFID is the higher transpar-ency throughout the entire chain (Davis 2002b), resulting in lower inventory lev-els, lower warehousing and transportation cost as well as in higher sales due to lower out-of-stock rates. In addition, RFID will help to increase food safety, in particular with regard to batch recalls, as batches with critical characteristics can be identified and recalled more easily.

According to Loderhose (2003), the most simple tag currently costs about € 0.50. However, the price is expected to drop down to € 0.05 per piece. A further price decrease down to € 0.01 per piece can be achieved if the development of printable plastic transponders will be successful (N.N. 2004b). As € 0.50 is still too expensive to attach the chip to every single product (compared for example to the price of a package of yogurt), the implementation will start with pallets and transport packages. Wal-Mart requires RFID-labeled pallets and cartons starting 2005 in the US and 2006 in Europe (Rode 2003c), although there still exist im-pediments such as low recognition rates, lacking standardization or privacy and health problems (Rode 2004b; Tamminen et al. 2004). On the product level, it is likely that RFID will first be applied to products with a higher value, a high prob-ability of theft, counterfeit, smuggling, or with higher risks of recalls or out-of stock (Wolfram and Spalink 2004).

From a technological point of view, the Electronic Product Code (EPC) is a prerequisite for the broad introduction of RFID. The EPC is a 96-bit digital num-ber that will allow identifying each individual product, not only each type of prod-uct. The EPC will be included in the RFID chip; all other information concerning a product will be stored in web-based databases (Rode 2003a). As a consequence of RFID, fresh food manufacturers will have to invest in this technology in order not to be de-listed by retailers in the near future although financial benefits exist mainly for the retailers (N.N. 2004a). A big challenge in this context is the inte-gration of the new technology into the existing IT landscape (Brauckmann-Berger and Rindle 2003), often realized by a specific RFID-middleware (Büker and Ber-ger 2004).


3.3.4 Social/Legal Characteristics and Developments

Several socio-economic developments have led to changing and varying consumer requirements on food products. On a strategic level, manufacturers and retailers respond to theses changing demand patterns by measures such as the extensions of their product portfolio, the development of new distribution channels, or by pro-moting specific categories.

Further differentiation of consumer requirements

Several socio-economic trends will cause substantial changes in fresh food mar-kets:

The share of elderly people in the population is expected to increase in the future (for example the share of people aged 55 or older in the Neth-erlands will rise from 23.5% of the total population in 2002 to 27.6% by 2010). Elderly people are less influenced by marketing and advertising, demand more convenience food and place higher importance on healthy food (Eilander 2002). The average household size is likely to further decrease (e.g. in the EU from 3.2 persons in 1970 to 2.1 persons in 2020). At the same time, women are participating in the work environment to a greater extent. The results of these trends are a rising demand for convenience products (e.g. ready-to-cook or ready-to-eat), longer opening hours of stores, and ad-justed packaging sizes as well as an increase of the Home Meal Re-placement segment and of home deliveries (Eilander 2002). Meulenberg and Viaene (1998) point out that the population in Western Europe becomes more multiracial, above all in the big cities, which stimulates the market for ethic food products. The available amount of income will vary more and more within a coun-try. Giles (1999) gives an example from the UK, which is applicable to the whole of Western Europe as well. While currently around 80% of the population fits into a broadly defined middle-income category, this num-ber is expected to decrease to 40% by 2010. 30% of the population will then be characterized by being “time rich, money poor”, the remaining 30% by being “time poor, money rich”. The fragmentation of the society will have major effects on food production (e.g. further product prolifera-tion or shifts in distribution channels). With regard to changing values and life styles, Meulenberg and Viaene (1998) name the replacement of simple traditional dishes prepared from raw products by industrially produced products, the disappearing sea-sonal cycle in food consumption and a shift towards “exotic” foods as main trends. Furthermore, consumers become segmented and fragmented with regard to their values and tastes. Eilander (2002) and Marshall (2004) define major consumer segments. In this respect, van Wezel (2001) emphasizes that consumers are becoming less predictable in their behavior.


The success of discounters and private labels show that consumers have developed a higher sensibility for pricing. For these so-called “smart-shoppers”, the price is the essential buying criterion (Seifert 2001).

Product Differentiation and Innovation

The main answer of the fresh food industry to the presented socio-economic de-velopments is product differentiation and innovation. New product development is ranked as a Top 3 source of competitive advantage in the food industry and even number one for bigger companies in the study of Gilpin and Traill (1999). Van Trijp and Steenkamp (1998) stress that the most important driving force in new product development is the aspiration for differentiation from the competitors. The number of product innovations introduced for example in German retail has in-creased since 1997 by 11.1% annually (see Fig. 3.7). However, the product flop rate reached almost two thirds in 2000 (Michael et al. 2002). Seifert (2001) argues that most new products contain only limited improvements or are so-called “me-too products” (76.7% of all innovations in 1999). The British retailer Tesco for example distinguishes five food development projects types, of which only one is a real new products launch. All others concern improvements of existing products (reformulation, new packaging size, rebranding, promotional products; Francis 2004).

Fig. 3.7 Product innovations in German retail (based on Michael et al. 2002)

The high flop rate can be explained by the fact that customer requirements are not really satisfied in a better way by these “innovations”. On the other hand, Dekker and Linnemann (1998) attribute the low success rate to the predominately empirical methods that are used in new product development. They point out that out of 140 ideas and concepts, only one will survive for more than five years. Three important factors handicap the introduction of new products (Seifert 2001). First, at the retailer’s side, the processes of listing new products are inefficient due to high listing fees and inadequate administrative processes. Second, old stocks

11,778 10,932 9,97811,534

11,879 14,881

20,214

20,944

32,478

30,192

23,657

25,813

1997 1998 1999 2000


are sold first by retailers before a new product is listed which leads to serious de-lays. Third, the sales forecasts provided by the food manufacturers are seldom ac-curate. As a consequence, production capacities are not sufficient if the product is successful and the product gets out-of-stock.

Promotions

An essential difference between Europe and the US is the frequency of promo-tional activities. The planning of promotions has an outstanding significance in the relationship between food manufacturer and retailer in Europe. Particularly in Germany, significant revenue shares are generated by promotions (Treeck and Seishoff 2002). The average annual number promotions per manufacturer are es-timated to be between 25 and 150 (Seifert 2002c). Most promotions are price dis-counts; other types include loyalty rewards, special events, displays, samples, or contests (Seifert 2001). Several studies reveal that most promotions are neither ef-fective nor successful in terms of value creation for the manufacturer. According to Fairfield (2002), manufacturers estimate that only 35% of all promotions are profitable. Moreover, they even cause severe problems in the SC:

Promotional items show exceptionally high out-of-stock rates, sometimes up to 20% (Chappell et al. 2002), which is very annoying for the con-sumer as he goes into a store or outlet just to buy the promoted item. The high out-of-stock rates are mainly due to the fact that volume plan-

ning for promotions is extremely difficult. Only few historical and com-parable data are generally available. Furthermore, unknown influences such as consumer behavior or competitor actions affect the promotion. Therefore, the planning of promotions should be completely separated from the planning of standard items (Treeck and Seishoff 2002). Promotions cause high variations in demand, which results in over-dimensioned stocks, high product surpluses and waste, additional product set-ups in production, and higher administrative costs (Fairfield 2002). Retailers even use the price discounts for promotions to accumulate high inventories resulting in production peaks (Seifert 2001).

Focus on fresh products

Most food processors and retailers expect that the share of fresh and perishable products in retail will continue to grow (Grievink 2002) from 45% to 60% of all supermarket sales after 2005 (McLaughlin 2002). Krampe (2004) stresses that grocery retailers need more fresh products to distinguish themselves from the dis-count channel. For example, freshness is one of the most important competitive advantages the German retailer “Globus” to attract customers. If the lead-time of a product exceeds a quarter of its total shelf life, the product will be rejected (N.N. 2003a). However, most outlets lack the necessary customer frequency to keep in-ventory turns at a high level. McLaughlin (2002) argues that many fresh food manufacturers will even decentralize their production facilities in order to increase freshness and to minimize transportation cost.


3.3.5 Environmental Characteristics and Developments

Environmental aspects as well as food quality and safety concerns are more and more considered when assessing SCs in food industries. On the one hand, these SCs have a substantial environmental impact at all stages of the SC (e.g. farming, distribution, or consumption). On the other hand, there is a large public interest in issues such as health, safety or protection of animals (van Beek et al. 1998).

Environmental impact

According to van Beek et al. (1998), the protection of the environment has be-come an issue at all levels of society. New legal requirements make manufacturers responsible for their products, even beyond their sale and delivery. Therefore, product management must include the entire chain; Fig. 3.8 shows potential envi-ronmental actions in a SC. Bloemhof-Ruwaard et al. (1995) give numerous exam-ples for environment-friendly management practices. At the raw materials pro-curement stage, recyclable materials and renewable resources can be preferred. In manufacturing, focus can be given to both process design (e.g. reduce waste or minimize pollution) and product design (e.g. design to recyclability).

Fig. 3.8 Green supply chain (based on van Beek et al. 1998)

Severe environmental impacts also arise in distribution. McKinnon (1999) shows that the food and drink sector accounts for almost a quarter of all freight movements in the UK in terms of ton-km. 15% of this number require cooling which causes 40% of total energy consumption for these types of freight. Fur-thermore, between a fifth and a fourth of all vehicles run empty (McKinnon 1999). Optimizing fresh food transports has a big environmental impact as it saves energy and hence on the company’s cost position. Finally, at the use and disposal stage, infrastructural measures are required to collect and sort the waste (Bloemhof-Ruwaard et al. 1995).

Food Quality and Safety

For Jongen (1996), quality is “to meet the expectations of the consumer”. How-ever, when looking at specific product categories, it is noteworthy that the quality perception of the consumer has decreased markedly between 1994 and 2001, in particular with regard to meat products (see Fig. 3.9).


Fig. 3.9 Quality perception of consumers 2001 versus 1994 (based on Spiller et al. 2002)

In contrast, dairy products have been able to increase consumer confidence. The quality of food products comprises numerous attributes (see Fig. 3.10). The con-sumer can only judge some of these aspects. For others such as wholesomeness, he has to rely completely on the retailer and manufacturer as well as on the control exercised by authorities (Erdtsieck 1989). Jongen (1996) and Schiefer (2000) em-phasize that quality management is by definition interdisciplinary and requires an integrated approach; the product liability for the food manufacturer does not end at the factory gate, but rather includes the entire SC down to the consumer.

Within all food quality aspects, food safety is probably the most important one. For example, in Germany alone over 85,000 people have fallen ill with salmonel-lose in 1999 (Spiller et al. 2002). Hoogland et al. (1998) state that an average of 120 cases of food-borne illness per 100,000 population were reported from 11 European countries in 1990. For a detailed simulation of the epidemiological and economic effects of salmonella control in the pork supply chain it is referred to van der Gaag (2004). Although food products have never been as safe as today, the consumer is as critical as never before (Felger 2003). Food safety crises such as the foot and mouth disease or the spread of Bovine Spongiform Encephalopathy (BSE) as well as the discussion about genetically modified foods (Kuznesof and Brennan 2004) reinforced this development. According to Grievink et al. (2002) and Ennen (2003), food safety has become one of the highest priorities in the food industry and retailers are being pushed into a position of “guardians” for food safety. From the retailer’s point of view, for manufacturers food safety is the most important issue today (Grievink et al. 2002).

-29

-28

-25

-18

-14

-8

2

6

11

12

25

26

34

0

-8

Pork

Beef

Chicken

Veal

Fish

Sausage

Eggs

Potatoes

Food in General

Fresh Milk

Buter

Fruits & Vegetables

Cheese

Bread

Yogurt


Fig. 3.10 Quality aspects of food products (based on Erdtsieck 1989)

To cope with the food quality and safety challenges, most actors in FFSCs have implemented Quality Assurance Systems (QAS), the mostly applied are:

Good Manufacturing Practice (GMP) covers fundamental principles, procedures and means to design a suitable environment for the produc-tion of food of acceptable quality; Good Hygienic Practice describes the according hygiene practice. For several production types, specific codes have been developed (Hoogland et al. 1998). Hazard Analysis and Critical Control Point (HACCP) is a system for analyzing production or product handling processes to detect hazards and risks of contamination (Tuttle 2001). After an in-depth analysis of micro-biological, physical or chemical hazards and risks, Critical Control Points are identified at which control can be applied (Hoogland et al. 1998). Both the EU directive 93/43/EEC (de Sitter 1998; Spiller et al. 2002) and the German “Lebensmittelhygieneverordnung” (Siebel 2000) require HACCP principles as a standard. The ISO 9000 family of standards and guidelines of the International Or-ganization for Standardization (ISO) is an industry-spanning norm for the set-up and description of a Quality Management system. Although certi-fications according ISO 9000 have been demanded intensively in the mid-nineties, ISO 9000 has lost much of its significance in the recent years (Spiller et al. 2002). Hoogland (1998) gives an overview of the dif-ferent standards and guidelines.


The International Food Standard (IFS) is a standard enforced by retailers for the quality assurance of their private label products.The EU-directive 178/2002 “General Food Law” (effective since Janu-ary 1st 2005) focuses on risk analysis and traceability of food products.

Although implementing QAS is often voluntary, companies not having adopted a QAS may have problems to stay in business (Hoogland et al. 1998), since the QAS implementation is regularly a prerequisite to deliver to a specific retail chain. In addition, most consumers are extremely aware of food quality and safety con-cepts (Allinson 2004). The implementation of the QAS at food manufactures is audited by public authorities, as well as by the retailers themselves. Due to the high number of different QAS, a manufacturer can be audited up to 50 times a year. In order to reduce this number, the industry aims at harmonizing the differ-ent systems (Jahn et al. 2003). With respect to the implementation numbers of QAS, considerable differences can be observed at different stage in the SC. While at the manufacturer’s stage, most companies have already implemented a QAS; the dissemination of QAS is much less frequent at other stages of the SC. Spiller et al. (2002) give an example for the SC of meat (see Table 3.2). A very successful QAS in this industry is the Integrated Chain Surveillance system in the Nether-lands that covers over 90% of the Dutch pork producers (N.N. 2004d).

Table 3.2 Dissemination of QAS in meat SCs (based on Spiller et al. 2002)

Stage in SC Dissemination of Quality Assurance Systems

Feed Industry Broad dissemination of QAS, especially GMP Farmers Only few QAS implementations Animal Trade/Transport Only few QAS implementations Slaughtering General implementation of HACCP, also ISO 9000 Processing General implementation of HACCP, also ISO 9000 Retail On outlet level only very few implementations of QAS Gastronomy Only in system gastronomy in relevant scale

Traceability

Food safety is strongly related to the traceability of products (Felger 2003). Ac-cording to the EU regulation 178/2002, the traceability of food products has to be assured within all production, processing and distribution stages of the SC starting January 1st, 2005 (Dietz 2003). Other regulations such as the EU regulation on feed and food control or the German “Lebensmittelkennzeichnungsverordnung” include traceability issues as well (CSB-System AG 2004c). On top of this legal requirement, many retailers force their suppliers to implement traceability systems (Rode 2004d), in particular by enforcing the standards of the Global Food Safety Initiative or the standards of the British Retail Consortium (CSB-System AG 2004c). Food traceability is the information required to “describe the production history of a food crop and any subsequent transformation or process the crop

might undergo on its journey from the grower to the consumer’s plate” (Gellynck et al. 2002). Current traceability systems are basically concerned with animal health, disease and food safety; nevertheless, they are more and more expanded


into marketing tools. Economic benefits of such systems include reduced disease levels and compensation payments as well as more efficient allocation of testing resources (Gellynck et al. 2002). Furthermore, product recalls that occur relatively often can be managed more effectively (Dietz 2003).

The Internet plays an important role in modern traceability systems, allowing all SC partners to follow the product on its way through the SC (N.N. 2003b). It is most likely that the EAN 128 barcode will constitute the basis for most traceabil-ity systems. The data that has to be provided by the traceability system includes the type and the volume of the merchandize traded, but also the shelf life of the products (Weber 2004). The biggest challenges with regard to ensuring traceabil-ity arise for manufacturers with tank goods (e.g. dairy), due to the fact that the ori-gin of the product cannot be assigned any more once the product is mixed with other products. A clear assignment of origin in this case requires very large in-vestments even for small processors. (Weber 2003).

3.3.6 Summary

The described characteristics and developments in FFSCs are given in Table 3.3. They must be taken into consideration when selecting an APS system for a fresh food producer and are hence integrated in the profile of Chapter 4.

Table 3.3 Characteristics and developments in fresh food supply chains

Field Major Characteristics and Developments

Consolidation

Internationalization

Rise of private labels

Growing importance of the discount channel

Outsourcing

Decreasing lead-times

1. Economic Developments

ECR and CPFR

EDI

E-marketplaces

2. Technological Developments (IT)

RFID

Further differentiation of consumer requirements

Product differentiation and innovation

Promotions

3. Social / Legal Developments

Focus on fresh

Environmental impact

Food quality

4. Environmental Developments

Traceability


3.4 Characteristics of Fresh Food Production Systems

3.4.1 Overview

As most food industries, fresh food industries belong frequently to the so-called process industries, which are characterized by “repetitive production operations

carrying out specific physical (e.g. blending or milling) or chemical reactions”(Günther and van Beek 2003). Process industries usually show a higher complex-ity than discrete manufacturing, which is caused by factors such as the perishabil-ity of products, the high number of end products, a great variety of possible pro-ductions paths, special storage equipment, co- and by-products, or variable recipes (Crama et al. 2001). The characteristics of specific production systems are often described using the Product-Process-Matrix (see Fig. 3.11; Fransoo and Rutten 1994; Rutten 1995; Olhager and Wikner 2000; Crama et al. 2001). The dimension “Product Structure” varies between customized, tailor-made products that are pro-duced in small quantities and commodities on the other side. The dimension “Process structure” refers to the material flow complexity within the factory. In a job shop environment, the material flow can differ per product.

The flow shops can be distinguished further (Rutten 1995): In the disconnected line flow, material flows in batches through the factory. A connected line flow where discrete products flow through the factory in a continuous line is typical for car manufacturers. Finally, the continuous flow is representative for homogeneous materials like oil. Process industries can be found in the continuous flow segment, but even more in the disconnected line flow segment. This is the also the case for most fresh food industries (shaded area) when large to mid-size batches are proc-essed. According to Blömer and Günther (1997), the batch production type be-comes increasingly important due to customer requirements for a higher product variety leading to lower volumes per product.

Fig. 3.11 Product-process matrix (based on Rutten 1995)

3.4 Characteristics of Fresh Food Production Systems 59

Most production systems in fresh food industries – as well as in the food indus-try in general – contain the four steps depicted in Fig. 3.12. A production system comprising “Processing” and “Packaging” as the two most important steps is also named “Make-and-Pack” production (Méndez and Cerdá 2002). The number of products involved increases with each production step. Out of a limited number of raw materials (e.g. raw milk), a still moderate number of intermediate products are produced within the processing step. High product complexity typically occurs at the packaging level due to different tastes and customer individual packaging for-mats. The specifics of each of the production steps are portrayed in the following paragraphs.

Fig. 3.12 Fresh food production system

3.4.2 Formulation

Three types of materials are required in fresh food production: raw materials, in-gredients and packaging material. Generally, the raw materials and also of some of the ingredients have a relatively short expiry date and a high perishability, and therefore have to be processed before the decay. As a consequence, planning and scheduling of the production has to be tightly integrated with the raw materials supply. In many cases, the supply of raw materials is a push supply with high lead-times. Due to long-term contracts, the processor has the obligation to buy a certain volume at a specific point in time. Additionally, raw materials in fresh food production such as raw milk in dairy, animals in fresh meat or grains in bak-ery have a high variability in several dimensions. First, the available quantities of the raw material can vary significantly over time, which is due to the seasonality or the weather-dependence of farm-based products. Secondly, according to the fluctuating quantities, the price of the raw materials can vary as well. Thirdly, the quality of the delivered products is also subject to high variability (van Wezel 2001). Examples are the fat content of raw milk that depends on the seasonality or the weight and the size of animals, which in turn depend on the feed given by the farmers. In contrast to discrete manufacturing, process industries rely on a recipe, not on a BOM. The variable quality of the raw material often leads to variations in the quantities used to produce a product to keep the quality and the characteristics of the finished product stable. In order to include variations of the raw material in terms of quality and price, the recipes are frequently updated just before produc-tion starts. Likewise, the recipe has to provide some kind of flexibility in the choice of raw materials and the quantities used (Rutten 1995). The traceability of products has consequences on the reception of materials (see Chapter 3.3.5). Traceability of products is especially important if products have to be recalled


from the retail in case of a quality issue. Raw materials are the major source of contamination in food production (Schraft et al. 2001). If a quality problem is due to the raw material, all production batches based on the same raw material batch have to be recalled. However, as high volume bulk products are often stored in si-los or tanks where the raw materials of different suppliers get mixed, the raw ma-terial batch is considerably larger than a batch in production. Hence, a requirement is to separate raw material batches of different suppliers as much as possible.

3.4.3 Processing

Van Boekel (1998) distinguishes four basic processes: stabilization (preservation), transformation, separation, and production of fabricated foods. The types of equipment involved are for instance kettles, ovens, pipelines, mixers, or conveyor belts (van Wezel 2001). Examples for processing in fresh food industries are pas-teurization or fermentation processes in the dairy industry, slaughtering and dress-ing processes in fresh meat or mixing and scalding in sausage production. Many of these processes show common characteristics, but not all of the following charac-teristics can be applied to all fresh food processing processes. A general character-istic of all fresh food industries is the divergent product flow within the factory. Based on a few raw materials, a certain number of intermediate products are pro-duced. The high degree of product complexity is typically introduced at the pack-aging step. Other product flow characteristics are co- and by-products. Co-products are produced automatically and are further processed into other products (e.g. raw milk is processed to skim milk and cream that can further be processed into butter). By-products (e.g. poultry blood) can also not be avoided and are commonly further processed in the feed industry (van Sonsbeek et al. 1997). Al-though these processes show some characteristics of continuous or flow produc-tion as they deal with bulk quantities of homogeneous products, there are several reasons why fresh foods are typically produced in batches: First, the volumes re-quired per recipe are limited so that a continuous production would result in sig-nificant surplus. Secondly, several processes require tanks or kettles. Therefore, the batch size is a multiple of the tank or kettle size. Thirdly, for hygiene purposes, frequent cleaning is required for which the production has to be interrupted. Fi-nally, the limited shelf life of the products prohibits the building up high stock levels (van Wezel 2001).

Furthermore, changeover times in processing are relatively long as intensive cleaning, sterilization, and re-tuning are required. Furthermore, the set-up usually causes wastage due to calibration and cleaning processes. In addition, the set-up times are sequence-dependent, which means that producing recipe 1 before recipe 2 necessities other set-up times than the other way round. In many cases, products can be produced in continuity or in blocks if a certain sequence is respected (e.g. the strong flavor after the weak flavor, product with high microbial spoilage after product with low microbial spoilage, etc.). The equipment lay out is in general ac-cording to the product type. The involvement of labor is relatively low in process-ing; even cleaning and sterilization are performed automatically by so-called


Clean-in-Place (CIP) procedures. However, processing is capital intensive and the production is mostly determined by machine capacity (van Dam et al. 1993; Jakeman 1994; van Donk 2001; van Wezel 2001). The processes show a high de-gree of variability of processing times and yields, even if the process is statisti-cally under control (Rutten 1995). This is mainly caused by the variable qualities of the raw materials, although adapting the recipes can level off some of the fluc-tuations. Furthermore, the micro-processes and chemical reactions are very com-plex and often cannot be expressed mathematically, even when processes are rela-tively simple (Rose 2001; Scheiber 2002). To cope with the variability, it is common to use an average for processing time and yield. If quality defects occur, it is possible that the entire batch cannot be used (Blömer 1999). In some fresh food industries, this batch can be reintegrated in the process and mixed with other batches so that the material flow gets cyclical elements. The level of work-in-progress is generally very low for the following reasons: First, the intermediate products are also subject to perishability, and therefore they must be further proc-essed as quickly as possible. Secondly, in many factories there is only limited warehousing space available for intermediates, as these products often have to be chilled, which causes additional costs. However, some products must stay in the warehouse for a certain time because required product tests (blocking time) have to be performed or because they need time to mature (maturation time).

3.4.4 Packaging

Van Wezel (2001) distinguishes a protective and a cosmetic function of packag-ing. One the one hand, packaging should protect its contents against the “outer world”, and should make the product transportable and easy to use. Packaging gives shapeless products a shape (e.g. yogurt) and delays the decay of the product (e.g. by avoiding microbial spoilage). One the other hand, from a marketing and sales point of view, packaging should allow the product to be easily found on the shelf, provide information about the product (especially to fulfil legal require-ments) and address the right emotions (van Dam 1995). Packaging material is also an important factor with regard to the product cost structure as it may represent up to 20% to 50% of the cost of the food product, depending on the food type and the packaging size (Gould 1997). Within the packaging step, the batches are trans-formed into discrete units. Processing and Packaging are usually decoupled; and a buffer stock of intermediates built up to avoid delays.

At the packaging stage, the product variety can increase dramatically as a proc-essing batch can be packed into many different packaging materials (glass, tin, foil, cardboard, cups etc.) and packaging sizes. The variety increases even further considering the fact that many customers require their own packaging design (e.g. private labels for supermarkets). The equipment lay out in the packaging zone is usually according to packaging material as most packaging units can handle one packaging material in many different packaging sizes. The equipment itself is sub-ject to frequent interruptions and is capital and labor intensive. In particular, all input and output related tasks (e.g. palletizing), as well as removing disruptions,


are performed manually to a high extent. To overcome traditional problems with the disposal and exchange of pallets, pallet pools (e.g. Chep) have been developed which are third party suppliers that maintain and lease the pallets (Bowersox and Closs 1996). According to the split of the processing batches into several products, the average order size in packaging is much smaller than in processing which re-quires frequent changeovers. It is important to notice that the full throughput can only be achieved after the machine is adjusted properly which can take much time and which leads to high changeover times that can also be sequence-dependent. Van Dam (1995) differentiates three changeover types:

Format change: change of size of packaging material; Product change: change of unpacked product, frequently related with cleaning;End product change: change of the packaging material specific to the or-der, neither change of size nor change of unpacked product required.

3.4.5 Storage and Delivery

In general, the stock levels of finished products are low, typically in a range from one to a couple of days (Broekmeulen 1998). This is again due to limited store space within the factory, to the perishability of the products and hence the risk of obsolescence as well as to the necessity to chill the products. The products are then delivered either directly to the retail outlets or – more commonly – via DC of the retail chains. Customer ordering is relatively frequent, but with small quanti-ties. An additional complexity for fresh food producers emerging in the recent years is that the distribution function from the manufacturer to the retail DC is in-creasingly performed by the retailer itself or by LSPs under the control of the re-tailer (Loderhose and Kapell 2003; Biehl 2004). With respect to the decoupling point, a combination of MTO and MTS can be observed in fresh food industries (Soman et al. 2002). The decoupling point is the point in the logistic chain at which a product is assigned to the specific, known customer order (Lehtonen 1999). At this point, the forecast-driven activities are separated from order-driven activities. It constitutes the main stock point from which deliveries to customers are made (van Donk 2001). MTS in fresh food industries means that the customer order is served by products from the stock of final products. In contrast, in an MTO environment, the products are packed and sometimes even produced based on a customer order. An important measure in an MTO environment is the com-pliance to the schedule whereas for MTS production, the accuracy of the forecast is most critical (Friend 2002). Although the larger production volume is produced on an MTS basis, there seems to be a clear tendency toward MTO production (ten Kate 1995; van der Vorst 2000; Soman et al. 2002). First of all, fresh food produc-ers are forced to increase their number of SKU’s to cover all client specific re-quirements. Furthermore, retailers and wholesalers order frequently, but irregu-larly. Substantial variations in demand from one week to another can be observed often (van Dam 1995; van Donk and van Dam 1996) bearing a high risk of stock


obsolescence or stock outs in case of an MTS production. Finally, although typical lead-time requirements from processor to DC shortened and are expected to de-crease further (van Wezel 2001) supporting an MTS production, retailers do not accept two subsequent deliveries with the same expiry or Best-Before Date (BBD; Soman et al. 2002).

3.4.6 Summary

The characteristics of all production steps are summarized in Table 3.4. As not every characteristic is present in every industry, the production processes of the case studies are described in the following paragraphs (Chapters 3.5 – 3.7).

Table 3.4 Characteristics of fresh food production systems

Production Step Characteristics

High perishability of raw materials Variable available quantities of raw materials Variable price and quality of raw materials Flexible recipes Mixed raw materials

1. Formulation

Push supply with high lead-times Divergent product flow Co-products and By-products Batch production High changeover times and costs for set-up and cleaning Calibration of equipment after set-up Sequence-dependent set-up and cleaning times and costs Lay out according to product type Low labor involvement High capital intensity Variable processing times and yields Low level of work-in-progress

2. Processing

Blocking and maturation times Decoupled from processing High product variety Lay out according to packaging material Frequent interruptions High capital and labor involvement Pallet pools Small order size and frequent changeovers High changeover times

3. Packaging

Sequence-dependent set-up and cleaning times and costs Perishability of final products Low stock level and frequent ordering with small quantities Distribution to DC organized by retailer MTO and MTS in parallel High variations in demand

4. Storage and Delivery

Short lead-time requirements


3.5 Case Study 1: Yogurt Production

3.5.1 Market Segments and Case Study Overview

For young mammals, including human infants, milk is the first food ingested and continues to be the sole constituent of the diet for a considerable period of time. Today, milk is probably the most important single agricultural commodity. Many animals such as cows, goats, sheep, and camels are exploited to produce milk for human consumption. However, in many parts of the world the cow is of over-whelming importance to milk production (Varnam and Sutherland 1994).

Among all dairy products, dairy fresh products represent the second largest share in terms of weight (see Fig. 3.13). Within the dairy fresh products, most categories belong to fermented products. The fermentation of milk is a fairly sim-ple, cheap, and safe way to preserve milk (Walstra et al. 1999). Fermented dairy products are milks adjusted for fat and non-fat solids and modified by incubation with dairy cultures. While originally yogurt products were only popular in the Middle East, Balkans and Caucasus regions, nowadays a wide rage of such prod-ucts is produced worldwide. They differ in aroma, flavor, and fat content as well as presence and variety of fruits (Rosenthal 1991). Besides yogurt, fermented products also include buttermilk, kefir, or sour milk (Klostermeyer 1996).

Yogurt is arguably one of the most popular dairy products (see Fig. 3.14; the classification given by Lebensmittel Zeitung differs from the classical segmenta-tion that relies on a technological perspective, it rather has a retail category per-spective as, for example, fruit yogurt can be either made of set or of stirred yo-gurt.).

Fig. 3.13 Production of dairy products in Germany 1999 in Kt (data from Lebensmittel Zei-tung 2001)

564

548 427

1,950

1,754

678 3,431

2,893

3.5 Case Study 1: Yogurt Production 65

Fig. 3.14 Production of selected dairy fresh products in Germany in Kt (data from Le-bensmittel Zeitung 2001)

Yogurt is a “semisolid fermented product made from standardized milk mixed from a symbiotic blend of yogurt culture organisms” (Chandan and Shahani 1993). Several types of yogurt exist; the two main types are set and stirred yogurt (Varnam and Sutherland 1994). While set yogurt is incubated and fermented after packaging in the retail cups, stirred yogurt is fermented before packaging. Besides these two main types, several special types of yogurt occur such as drinking yo-gurt, concentrated (strained) yogurt, or frozen yogurt. The success of yogurt can be attributed to the following factors (Rosenthal 1991; Chandan and Shahani 1993; Tamime and Robinson 1999):

Health-related glamour of fermented milks and increase of low fat prod-ucts,Achievement of a desirable taste by using special sweeteners, High versatility of taste, color, and texture Intense marketing and merchandizing activities, Relatively low costs of the product, and Longer shelf life than fresh milk.

Yogurt is one of the few dairy products with a relatively high and stable growth rate in terms of volume produced (e.g. 5.5% CAGR for fruit yogurt between 1996 and 1999, see Fig. 3.14). Although the yogurt segment for specialties and branded products still shows higher margins than other dairy products, the dairy industry in total faces strong competition (Murmann 2004a), resulting in relatively low mar-gins which are expected to decrease even further due to several factors:

257338

283 279152153 158 161

535531

578 625192

192206

229

118

148153

183

193

218209

221

1,1951,1841,0811,019

2,8932,771

2,661

2,466

1996 1997 1998 1999


Overproduction and a stagnating demand lead to a consistently high milk surplus in almost all developed countries. The enlargement of the European Union by ten Middle and East Euro-pean countries with a strong agricultural sector in 2004 will further in-crease the surplus. Discount channels and low margin private labels are getting more impor-tant every year, in some dairy categories the share of private labels has surpassed 50% of the total sales (Hemmelmann 2003). The cut in EU subsidies will lead to a further concentration of the dairy industry in Europe and to an intensified competition (Murmann 2004a). For instance, Frouws and van der Ploeg (2000) report that the number of dairy companies in the Netherlands decreased from 43 to 12 in the recent years, with two companies dominating the market.

Many dairy manufacturers respond to these challenges by focusing on the higher value added yogurt business. Although the advertising intensity and the flop rate of new products are already very high (Heymann 2002), product differen-tiation has been identified as the main success driver in this category (Heimig 2003; Kirschmeier 2003). Only by delivering a constantly high number of innova-tions, manufacturers can secure or even increase their market share. Therefore, many manufacturers are confronted with a high and still increasing product com-plexity, which makes production planning a complex task.

The case study is based on the production of stirred yogurt (see Fig. 3.15). On the one hand, stirred yogurt comprises most of the fruit yogurts and hence covers the biggest yogurt sub-segment to a large extent. On the other hand, the produc-tion of stirred yogurt has to cope with a very high product variety, making produc-tion planning for stirred yogurt more challenging than for other yogurt types.

Fig. 3.15 Scope of yogurt case study


An overview of the stirred yogurt production process is provided in Fig. 3.16. Walstra et al. (1999) name four objectives of the processing process in dairy pro-duction. The most important objective is to ensure the product safety for the end consumer. Other objectives include the quality of the process and of the product (e.g. in terms of eating quality, nutritional value, appearance, or emotional values) as well as a strong focus on maintaining processing cost at a low level. In the fol-lowing paragraphs, each of the production steps is explained in detail. In addition, the main characteristics of the raw material and the most important ingredients are described.

Fig. 3.16 Processing steps in stirred yogurt production

3.5.2 Raw Milk Collection

Raw milk has several unique characteristics that make the dairy SC and produc-tion system different from other fresh food production systems (Rosenthal 1991; Tamime and Robinson 1999; Walstra et al. 1999):

Once separated from the cow, raw milk has no protection from outside contamination. Raw milk is highly perishable and the same holds true for most intermediate and final products. Raw milk is an excellent culture medium, which provides at its natural temperature of 37°C the optimal conditions for boosting populations of microorganisms. Therefore, milk must be produced under conditions that prevent contamination; a clean environment is of prime importance. Furthermore, milk should be kept at a temperature of 4°C along the entire transportation chain from the farm to the dairy plant. The composition of the raw milk varies (e.g. with regard to fat and pro-tein content) from day to day, even within a particular breed, depending on such factors as the breeding policy, the age of the animal, the health of the udder, the feeding management, climatic conditions and seasons of the year, and also on the intervals between milking. Raw Milk is delivered to the dairy throughout the year, but in varying

quantities. As the raw milk must be processed within a very short time, processing capacity of a dairy can never be fully used during most of the year.Raw Milk contains several components that can be separated in various ways (e.g. cream and skim milk, powder and water etc.) so that a wide variety of products can be made.


Milk is normally collected twice a day from the cow and cold-stored at the farm in a milk tank (Rosenthal 1991). It is then supplied to the dairy usually by a tanker. Tank milk can be kept at low temperatures for a longer time and has lower transportation cost (Walstra et al. 1999), and is therefore the predominant method of milk collection today. Raw milk accounts for 50-90% of all costs of a dairy, depending on the focus of production (Murmann 2004b). However, this figure is rather at the lower end in yogurt production. Fig. 3.17 depicts a sample dairy SC. In this example, 400 farmers with 20,000 cows on an area of 200 km2 are covered by a typical dairy production. Nonetheless, a clear trend towards bigger units of both farmers and dairy manufacturers can be observed. In this context, Wouda et al. (2002) elaborate on the rationalization of a dairy network using an MILP model based on an example from industry.

On reception of the raw milk at the dairy factory, the intake is either metered or weighed. After taking a sample for the chemical and microbiological analysis, the milk is cooled to below 5°C using plate coolers and stored in a silo (Tamime and Robinson 1999). The silo usually consists of an inner tank of stainless steel, which is covered with isolation material. An agitation system is also part of most silos (Chandan and Shahani 1993). In some cases, the raw milk is clarified prior to stor-ing, meaning that solid impurities are removed from the milk by filtration or cen-trifugal separation (Rosenthal 1991). As the silos contain a volume of for example 300,000 kg of raw milk, variations in composition, quality, and properties of the delivered batches can be leveled off at this stage to a certain extent. Generally, the raw milk should not remain in the raw milk silo longer than one or two days (Wal-stra et al. 1999).

Fig. 3.17 Structure of dairy supply chain (based on Walstra et al. 1999)


Traditionally, the payment system for the raw milk has solely been based on the value of fat and the amount of milk. However, in the recent years more and more attention has been given in assigning specific values to proteins and nonfat solids in milk, which led to the generation of multi-component pricing schemes. These pricing systems rely on the testing results at the raw milk reception and integrate hygienic and quality issues (Rosenthal 1991). Therefore, an adequate planning support for a dairy production must offer a variable pricing component for raw materials.

3.5.3 Raw Milk Preparation

Raw milk is subject to a number of preparatory treatments before it is fermented into yogurt. The entire preparation process (see Fig. 3.18) is generally highly automated. Teixeira and Shoemaker (1989) describe a typical process control set-up. As the raw milk components vary significantly, it is first necessary to stan-dardize the raw milk in order to meet the compositional standards for yogurt. For this reason, the raw milk (1) is pumped from the storage silos to “balance tanks” (2). Milk powder (3) is widely used in the industry to fortify liquid milk for the manufacture of a thick smooth yogurt. Although different milk powders can be used, skim milk powder is the most widely applied (Tamime and Robinson 1999). Other ingredients include condensed skim milk or cream. The standardization process lasts between one and two hours (Nakhla 1995).

Fig. 3.18 Process line for raw milk preparation (based on Spreer 1995)


Then, the raw milk is pre-heated in a plate heat exchanger (4) to temperatures of 85°C for 30 min or 95°C for 10 min. The heating kills contaminating and com-petitive microorganisms, produces growth factors by breakdown of milk proteins and creates a desirable body and texture of the product (Chandan and Shahani 1993). The subsequent concentration takes place under vacuum in an evaporator (5). The concentration of the milk is much more important for set yogurt than for stirred yogurt (Walstra 1999). The condensate from the evaporator is usually used to pre-heat the incoming milk. Then, the milk is pumped to the homogenizer (6), which is a high-pressure pump forcing the mix through extremely small orifices, thereby causing extensive physiochemical changes in the colloidal characteristics of the milk (Chandan and Shahani 1993). Finally, the homogenized milk is fed back to the plate heat exchanger where it is again heated to 85-90°C in a holding tube (7), cooled and transferred (8) to the fermentation tanks (Tamime and Robin-son 1999).

3.5.4 Fermentation

In contrast to set yogurt, stirred yogurt is fermented in tanks before filling and packaging (see Fig. 3.19). Starter cultures are added in order to inoculate and in-cubate the mix. The starter culture is a critical ingredient in yogurt manufacturing. It influences not only the quality of the product, but also the fermentation time. Depending on the type of the starter concentrate (bulk or frozen), the kind of product, and the fermentation temperature, the fermentation time varies between 2.5-3.0 hours at 40-45°C and 16 hours at ca. 30°C (Tamime and Robinson 1999). The starter culture is usually metered directly into the mix (2), while the cooled mix (1) is pumped to the fermentation tanks (3).

Fig. 3.19 Process line for stirred yogurt production (based on Spreer 1995)


The tanks are generally designed with a cone bottom to facilitate draining of relatively viscous fluids after incubation (Chandan and Shahani 1993). Besides pure fermentation tanks, multipurpose tanks exist as well that can be used for milk processing (e.g. heating), fermentation and cooling. After incubation, the coag-ulum is cooled quickly (within 60 to 90 min at maximum) in a plate cooler (4) to 5-22°C depending on the type of product and kept in intermediate tanks (5). The yogurt is retained in this tank only for a short period of time, maximum overnight (Tamime and Robinson 1999).

3.5.5 Flavoring and Packaging

The flavoring and packaging step has to cope with a high number of product vari-ants, which is caused by the variety of tastes and packaging materials. Before the yogurt is filled into the retail containers (see (7) in Fig. 3.19), it is mixed with fruits and other ingredients (6). The fruit preparations for yogurt are specifically designed to meet the requirements for different product types and constitute gen-erally between 10% and 20% of the final product’s weight (Chandan and Shahani 1993). Several types of fruit preparations are available such as canned fruits, fro-zen fruits, fruit puree, or syrup. Other ingredients added at this stage are flavoring agents or coloring matters (Tamime and Robinson 1999). With regard to yogurt packaging materials, two types can be distinguished: the unit container (the vessel that comes into contact with the yogurt) and the outer shipping container. The most important unit containers are glass bottles, semi-rigid unit containers usually made from plastics, and flexible containers such as paper cartons (“Tetra Pak”) or plastic sachets. The lids, in particular used for the plastic cups, are widely made of aluminium foil (Tamime and Robinson 1999).

Fig. 3.20 Packaging line for yogurt (based on Spreer 1995)


For filling and packaging, a variety of different machines are available which are generally very capital-intensive and which achieve a performance of up to 100,000 retail containers per hour. An example for a packaging line is given in Fig. 3.20. The retail containers are fed to the line on a conveyor belt (1) and steril-ized (2). A rinsing unit (3) turns the containers so that they can be cleaned with water steam and sterile water. The containers are filled in the filling compartment (4) and closed with sterilized a lid (Spreer 1995).

The unit compartments that get in direct contact with the final product must be cleaned regularly and carefully. A cleaning sequence, including the removal of product remains and the subsequent sterilization, can last up to several hours. In addition to the decrease of the available filling capacity, cleaning also causes a material loss (yogurt volumes remaining in the tubes or calibration losses when setting up the unit for the new product). However, cleaning and sterilization can be avoided if the preceding product has no or only a minor negative effect in terms of color, taste, or microbiology on the following product or product variant. These products can then be produced in continuity without intermediate cleaning.


After packaging, the final cooling of the product takes place in the retail container to a temperature below 10°C since the growth of yogurt organisms (e.g. Strepto-coccus thermophilus or Lactobacillus delbrueckii) is reduced below this tempera-ture. The maximum effect on yogurt quality is achieved if the cooling is carried out as slowly as possible. A retaining period of 48 hours before the dispatch is ad-vantageous to achieve the final stability of the coagulum (Tamime and Robinson 1999). Therefore, many companies have introduced blocking or maturation times for the final product.

Refrigeration is also required during transport and in the retail outlet as well. The shelf life of yogurt produced under regular conditions is about 8-10 days at temperatures below 10°C (see Chapter 5.2.1). However, following the trend to concentrate production capacities, to extend the delivered markets and to increase the product portfolio, many manufacturers have increased the shelf life of their products to up to four weeks (Spreer 1995), especially using aseptic packaging technology.

3.6 Case Study 2: Sausage Production


Sausages are products in which “fresh comminuted meats are modified by various

processing methods to yield desired organoleptic and keeping properties” (Savic 1985). The production process dates back thousands of years to ancient Greeks

3.6 Case Study 2: Sausage Production 73

and Romans, and even earlier. The term “sausage” is derived from the Latin word “salsus” meaning salt or salted (USDA et al. 1999).

According to Pearson and Gillett (1996), consumers eat sausages for four rea-sons. The development of sausages was initially driven by economic factors such as utilizing lower-quality meats from cheaper cuts or edible by-products (Xiong and Mikel 2001). Savic (1985) stresses that within sausage production, the raw material of slaughtered meats are converted into products of higher value. Con-venience is the second reason for the success of sausages. Sausages take only little time in preparation (either only warming or even ready to serve). Thirdly, the great and still increasing variety of sausage products in terms of flavors, textures and shapes makes it possible to serve many different variants. Finally, sausages are also of good nutritional value since most of them are excellent sources for high quality protein, essential minerals, and all B vitamins (Kim 2001). Sausages can be categorized into three major segments (see for example Savic 1985; Neuhäuser 1996; Pearson and Gillett 1996):

Scalded Sausages comprise ready-to-eat products made from commin-uted and well-homogenized cured meats, fatty tissue, water and season-ings usually smoked and scalded. Examples are Frankfurters, Bologna (Mortadella), Knockwurst, or Wieners. Fermented Sausages are not heat processed and are made from cured or uncured, fermented and often smoked meats. Most of these sausages are microbiologically stable at ambient temperatures (Varnam and Suther-land 1995). This category can be divided into dry and semi-dry sausages and includes for instance Salami, Pepperoni, Summer Sausages (Cervelat Sausage), Meat Sticks, or Braunschweiger (Sielaff 1996). Cooked Sausages are ready-to-serve products usually made from previ-ously cooked fresh or exceptionally cured raw materials, subjected to fi-nal cooking after stuffing. Oelker (1996b) names the three sub segments Liver Sausages, Blood Sausages and Brawn Sausages.

Some authors, in particular from North America, name fresh sausages as a separate category (see for example Savic 1985; Xiong and Mikel 2001). Fresh Sausages (e.g. fresh “Bratwurst”) are neither cured, smoked, fermented nor cooked and must be kept under refrigeration prior to eating (Savic 1985). How-ever, in German and many European classifications (e.g. Neuhäuser 1996; Le-bensmittel Zeitung 2001) fresh sausages are assigned to fresh meat as they show similar characteristics.

In 1999, over 56% of all processed meat in Germany were sausages (€ 5.2 bn. out of € 9.8 bn. processed meat, see Fig. 3.21 and Fig. 3.1). Other processed meats include for instance cured products such as ham or bacon. During the last years, the total sausage production in Germany remained relatively stable at around € 5.2 bn. Among the sausages, scalded sausages constitute the most important segment (over 50% in 1999) and the only segment with a positive growth rate (2.6% CAGR between 1996 and 1999).


Fig. 3.21 Sausage production in Germany in € mil. (data from Lebensmittel Zeitung 2001)

This research is particularly concerned with the segment of scalded sausages and specifically with the production of larger diameter, sliced sausages (see Fig. 3.22). In contrast to small diameter scalded sausages that are usually eaten after immersion in hot water (e.g. Frankfurter or Wiener), larger diameter sausages are typically sold in slices and have a shorter shelf life. The scalded segment and the larger diameter product groups were chosen for two reasons. On the one hand, scalded sausages constitute the most important segment in terms of production value. On the other hand, production planning is particularly important for scalded and sliced sausages as they show the highest criticality with regard to shelf life due to their larger surface (see Chapter 5.2.2).

Fig. 3.22 Scope of sausage case study

780

814772 755

2,6822,6472,6832,482

1,7821,865 1,808 1,766

5,2035,2275,361

5,044

1996 1997 1998 1999


Therefore, the production of scalded and sliced sausages represents the most challenging production environment. The results of the analysis can then easily be transferred to scalded sausages sold unsliced, to small diameter scalded sausages and finally to other sausages segments. Fig. 3.23 provides an overview on the most important production steps in scalded and sliced sausage production. Each of the steps is explained in the following paragraphs.

Fig. 3.23 Processing steps in scalded and sliced sausage production

3.6.2 Input of Ingredients

Xiong and Mikel (2001) classify the sausage ingredients into four categories: raw meat, salt, water/extenders and seasonings. The most critical component is raw meat; a proper selection of the meat ingredients is precondition for the production of sausages of uniform quality standards (Savic 1985). Although raw meats used for sausages are generally lower-valued materials, they must be fresh with very low microbial counts. Primary sausage materials are beef, veal and pork; however, the use of poultry in sausage production is rapidly increasing, either as pure poul-try sausage or blended into other sausages.

Three types of basic raw meats can be distinguished: Skeletal muscle meats are the principal raw materials as they have high protein content and are excellent meat binders. Fatty tissues are incorporated in order to support specific taste char-acteristics, as well as for economic reasons. Finally, so-called variety meats or “fillers” can be integrated (e.g. skin or pig backfat). Although they have only infe-rior binding qualities, they are also palatable and nutritious if processed properly (Savic 1985; USDA et al. 1999).

With regard to production planning it is important to notice that a considerable amount of rework is integrated into the production cycle. This is especially true in sliced sausages production as slicing causes a lot of recyclable quantities. Raw meat trimmings vary significantly in kind, size and quality. In order to achieve a consistent final product, they are generally classified according to animal species, fat, moisture and protein content. The German classification system is included in the “Leitsätze für Fleisch und Fleischerzeugnisse” (Tändler 1984b).

For a long time, sausages have been made from meat removed from the car-casses immediately after slaughter (warm or pre-rigor meat). However, driven by technological developments (e.g. refrigeration) the slaughterer and the sausage manufacturer are generally separated today and the meat is transported in a chilled or frozen state from the slaughterer to the sausage manufacturer. All raw meat ma-terials must be stored at a temperature between 0°C and 3°C if unfrozen or at –8°C to –20°C if frozen. Unpacked fresh meat can be stored up to a week if stored around 3°C and hygiene conditions are perfectly met (Troeger 1984).


Furthermore, salt and nitrite are the most critical non-meat ingredients. Salt is used to preserve the product, to enhance the flavor and to improve the binding properties. Nitrite as a curing agent helps to limit both bacteria growth and the oxidation of fats in meats. Water is added to improve the consistency of the mix-ture and to substitute fats whereas extenders such as whey protein concentrate or caseinates are applied to increase the overall yield and to improve binding quali-ties and slicing characteristics. Finally, seasonings add a specific flavor to the sau-sage (USDA et al. 1999).

3.6.3 Grinding and Mixing

As the raw meat can vary notably in terms of the size of the trimmings delivered, the meat chunks are first ground to form uniform cylinders of fat and lean materi-als. The meat is compressed by a worm feed, cut by the grinder knife and pressed through the holes of the grinder plate. The obtained meat cylinders are then tum-bled in a mixer to obtain a uniform distribution of fat and lean particles (Pearson and Gillett 1996). The mixer is also used to preblend the different raw meats batches. After a first tumbling, a chemical analysis is performed. According to the results of the analysis, different batches of muscle meats, fatty tissues and variety meats are put together with water and curing salt to achieve the properties required in the recipe (Neuhäuser 1996). LP models are often applied to perform the calcu-lation and to optimize the yield. An example of an industry application is given by Jank and Wäscher (1999).

3.6.4 Chopping and Emulsifying

In order to further reduce the size of the meat particles, the preblended batch is transferred to the chopper (also named cutter), of which two types can be distin-guished. Smaller batch sizes are processed in simple choppers composed of a re-volving metal bowl that contains the meat while knife blades rotating on an axle cut through the revolving meat mass (Klettner 1984; Pearson and Gillett 1996). While this type of chopper is operated in batch mode, a Continuous Cutter Auto-mat can be operated continuously and achieves high processing volumes (Neu-häuser 1996). At the chopping stage, seasonings and ice are also added to the sau-sage dough. Ice is particularly important as chopping has to be performed at very low temperatures; but, the rotation of knifes bears the danger of heating up the mixture (Hammer 1984). The chopper renders a fine meat-fat mixture, usually called “emulsion” (Savic 1985).

3.6.5 Stuffing and Tying

At the stuffing and tying step, the sausage dough is encased either in artificial (cel-lulose, polymer or collagen), or natural (derived from slaughtered animals) casings


(see (1) in Fig. 3.24) and tied with thread or fastened with metal clips (see (2) in Fig. 3.24; Müller 1984; Pearson and Gillett 1996; Jakob 2002).

At this stage, the size and the shape of the future product are determined. With respect to the production of sliced sausages, the most important determination at this point is the diameter of the sausage, which can vary between ca. 5 cm and ca. 12 cm. The length of this intermediate sausage is determined by production re-quirements of the packaging machines. The tied sausages are suspended from a smoke sticks so that the entire sausage is not in contact with the rod or with an-other sausage (Savic 1985). 12 to 18 smoke sticks are placed on a trolley that is then transferred to the smokehouse.

Fig. 3.24 Stuffing and tying machine (based on Zweig and Sielaff 1996)

3.6.6 Scalding

The scalding of the intermediate sausages is performed in special scalding cham-bers (see Fig. 3.25). The entire process consists of four steps, which are all fully automated and carried out within the chamber (Savic 1985; Oelker 1996a). As moist products absorb the smoke only unsteadily and to a minor extent, the sau-sages are first dried. Afterwards, hot smoke is introduced to the chamber (1), lead-ing to the start of the coagulation and of the binding of the proteins. Scalding fol-lows immediately after smoking, mostly by hot air that streams in turbulences through the chamber (Zweig and Sielaff 1996). The scalding temperature of about 72°C has a major impact on the building of the typical structure of scalded sau-sages, on the creation of the specific aroma and color, as well as on the inactiva-tion of microorganisms (Brauer 1999).

In the end, the sausages are chilled by cold-water spray. Depending on the di-ameter of the products, the entire process can take up to eight hours (Neuhäuser 1996). Finally, the chilled products are transferred to the warehouse of intermedi-ate products.


Fig. 3.25 Scalding unit (based on Zweig and Sielaff 1996)

3.6.7 Maturing and Intermediate Storage

After scalding, the intermediate sausages stay a certain period of time in the in-termediate products warehouse. The entire time spent in the warehouse can be separated into a minimum or maturation time and an additional time caused by production planning. The maturation time, which varies between different prod-ucts depending on kind and diameter, is required to avoid a later sticking of the slices at the packaging stage. The time caused by production planning is due to the fact that only economic batch sizes are processed in the prior production steps. Therefore, the intermediate storage serves as a decoupling point between produc-tion and packaging. Scalded sausages are relatively sensitive to storage conditions. On the one hand, the storage temperature must be between –1° to 2°C to avoid microbial spoilage. A decrease below –1°C could freeze the product surface and impair its appearance (Cano-Munoz 1991). On the other hand, too much light has a negative impact on color and taste of the sausage (Wirth et al. 1990). A special characteristic of the intermediate storage is that most of the products loose mois-ture depending on the permeability of the casing. As most final products are sold on a per kg-basis, the optimization of the intermediate inventories can have a sub-stantial impact on the company’s financial performance.

3.6.8 Slicing and Packaging

Slicing and packaging is generally one single operation; the intermediate sausage is cut into slices, which are then directly transferred to the packaging machine by means of a short conveyor belt. Before slicing, non-edible casings are removed from the intermediate sausages. Typical for sliced sausages is the use of flexible


packaging material such as foils under a vacuum environment. As the vacuum is drawn, the flexible packaging collapses around the product, squeezing out most of the air from the pack (Bell 2001). The sliced sausage packages are then put into simple transportation or display cartons. The entire slicing and packaging process involves significant manpower input, in particular with regard to the final packag-ing into cartons. Slicing causes a loss of several percentage points, mainly due to the conic ends of the sausage, which cannot be processed. Furthermore, the set-up of a new product is usually associated with some ramp up losses.

Additionally, in the regular process there are always slices that do not fully cor-respond to the final product requirements and are therefore sorted out. However, as mentioned above, a high percentage of the losses can be re-integrated into the production process. This recycling quota can be further increased by investing in new technologies, however a tradeoff has to be made between the investment and the volume and value of the meat recycled in addition. With respect to packaging, the compliance to hygiene regulations is of particular importance as the degree of microbial spoilage at slicing and packaging is decisive for the shelf live of pre-packed scalded sausages (Wirth et al. 1990). Tändler (1984a) shows how the shelf life of a product depends in particular on the initial spoilage. The shelf life of pro-fessionally packed products is between one and four weeks, when chilled up to 10°C (Andrae 1996; see also Chapter 5.2.2).


The finished products are usually packed on euro pallets and wrapped up in shrink foil (Andrae 1996). The stock level of finished products is generally much lower than the stock level in the intermediate storage because at finished product level, no maturation time is required. Furthermore, many products are packed to order, which means they are packed only in case of a specific customer order and deliv-ered directly.

A delivery performed by the producer is relatively seldom in today’s business environment. It is more common that an LSP is in charge of the transport. How-ever, an increasing portion of the retail customers is picking up the goods with its own fleet within their supply logistics. The sliced sausages are usually delivered to a retail DC or directly to the retail outlets. The storage and the delivery trucks must be refrigerated. A major concern in the industry is the interruption of the re-frigeration chain in retail outlet, between unloading of the truck and packing the products into the shelf. If this period becomes too long, a negative impact on shelf life cannot be excluded.


3.7 Case Study 3: Poultry Processing


Although poultry represented in 1999 only about 17% of the entire meat produc-tion value in Germany (see Fig. 3.26), it is the only sub-segment that has contin-ued to increase over the last years (see Fig. 3.28; see for example Parry 1989; Ris-tic 1993; Lebensmittel Zeitung 2001; Bundesverband der Deutschen Fleisch-warenindustrie e.V. 2003). Chicken and turkey make up over 95% of the total poultry production in 2000 (see Fig. 3.27). The growth rates of both categories be-tween 1999 and 2000 were over 10%. Surveys of consumer attitudes suggest that there are three main reasons for this increasing popularity. Poultry meat is rela-tively cheap, has a healthy image (low fat), and is available in many varied forms when processed further (Richardson 1989; Kim 2001). However, the number of innovative products in the meat industry is rather limited compared to the dairy industry (Hitchens et al. 1998).

Fig. 3.26 Meat production in Germany 1999 in € mil. (data from Lebensmittel Zeitung 2001)

Fig. 3.27 Poultry production in Germany 1999 in Kt (data from Lebensmittel Zeitung 2001)

1,264

465

3,580

2,067

518576

266

294

42

44826

914

1999 2000

3.7 Case Study 3: Poultry Processing 81

Nossent et al. (1995) name several macro-trends that affect the meat processing industry and that all lead to increasing cost pressure:

Scaling-up and concentration of companies, Increasing international competition, Growing awareness of the “human factor”, especially regarding quality issues,Ongoing automation and mechanization of the production process.

With regard to fresh poultry production and distribution, several trends can be observed:

First, compared to other food and fresh food products, the discount chan-

nel in Germany started relatively late to sell fresh poultry since it has a very short shelf life and requires specific chest freezers. However, in the attempt to complete their assortment and following their strategy to cover all basic consumer needs (Hoffmann 2003a), the discounters introduced fresh poultry with significant success in 2002 and 2003 (see Fig. 3.28). The increase of the low price discount sales (30.2% of all poultry sales in the first six months in 2003) also puts the margins of the other channels under pressure (Holler 2003; Hoffmann 2004c). Second, a raise of self-service packages and a considerable decrease of poultry packaged at the counter are noticed in the industry (see for exam-ple Hoffmann 2003b; Wessel 2004a). Traditional retailers respond to these challenges mainly by stressing the freshness of their products and a broader assortment (N.N. 2004e). Finally, almost all European poultry markets face fateful overproduction;additionally, many overseas producers (e.g. from Brazil) have increased their exports to Europe (Wessel 2004b).

Fig. 3.28 Fresh poultry in % per outlet type (data from N.N. 2003c)

37.3%37.9%39.7%42.3%43.3%

7.5% 9.6%17.9%

30.2%

6.1%

15.7%

15.8%

15.6%

14.7%

17.0%

17.4%14.0%

16.3%

8.0%

15.0%

7.9%10.0%9.5%

5.3%

10.3%

4.5%6.7%7.6%8.8%8.3%

1999 2000 2001 2002 1. HY 03


Although some poultry is still killed and dressed on an individual basis, today the majority is processed in large-scale operations which are highly automated and which require a high throughput to justify the capital expenditure (Varnam and Sutherland 1995). A large-scale operation is also prerequisite to deliver to volume-oriented discounters. An overview of such a process is depicted in Fig. 3.29. As for the two previous case studies, each step is presented in detail in the following paragraphs.

Fig. 3.29 Processing steps in poultry processing

3.7.2 Transport of Animals

The contracts between the poultry processors and the growers of the animals are usually long-term. That means that the processor has the obligation to take the animals at a specific point in time, which is highly predictable since the time re-quired to grow a bird has only minor variations. Therefore, the animal supply must be considered as fixed in the short-term, similar to the milk supply in the yogurt case. Furthermore, meat processors exercise considerable influence on growers to ensure high quality standards throughout the entire meat chain (Thiemig 1996b). The grower is typically guaranteed a fixed payment per kg of living bird or per bird (McCorkle 1988). The performance of the growers has increased significantly in the last decades. According to Ristic (1993), between 1973 and 1993, the aver-age age of a broiler decreased from 50 to 35 days, the final weight increased by 0.47 kg and the feed utilization (kg feed required to produce one kg living animal) improved from 1.880 to 1.764. The birds differ significantly in terms of weight and their meat-fat-bones composition, which is influenced by a couple of factors such as the gender of the bird, the seasonality, the feeding, or the age.

To catch the birds, several systems have been developed. Loose crates are the earliest system. They are taken from the truck into the shed and manually filled with birds. Fixed crates are constructed as fixtures on the truck and are also filled manually. Modular systems consist of a metal frame with several compartments. However, due to its weight, a forklift is required for taking the frame into the shed. Finally, mechanical methods exist to harvest the birds that resemble a combine harvester and have a very high throughput (Parry 1989). For the transport, usually open-sided vehicles are used to provide beneficial cooling and chilling to the birds. Long transportation times should be avoided not only for economic reasons, but also to minimize damages during transportation, which can concern up to 8-25% of the animals. Stress factors for the birds that influence the loss rate include handling, confinement, vibration, noise, or air movements (Parry 1989; Turkki 1994). The delivery at the slaughterhouse must be coordinated with the slaughter-ing interval; specific sobering times have to be respected because they are impor-


tant for the evisceration (Thiemig 1996a). Feed withdrawal 8-10 hours before transportation is optimal in order to minimize yield loss (Jones and Grey 1989).

The birds are then brought to the reception area of the slaughterhouse. Good ventilation and even sprinkling equipment is necessary to prevent overheating of the birds in summer. Following a certain time for the birds to calm down after the transport, the living birds are manually hung on the overhead conveyor of the kill-ing line. Shackles are suspended from the conveyor to which the birds are at-tached, head downwards. The conveyor travels at a prescribed speed, which de-pends on the capacity of the slaughtering machinery. According to legal requirements, the poultry must not remain suspended for more than three minutes before stunning (Parry 1989). A scale built into the conveyor weighs the birds; the growers are paid based on this weight.

3.7.3 Stunning and Bleeding

In most countries, domestic poultry must be stunned before being slaughtered. Ex-emptions are only made for religious reasons, in particular with respect to Moslem and Jewish slaughter. Stunning is prescribed for four reasons (Gregory 1989):

To minimize the chance of the birds feeling pain - stunning must make the animal completely insensible, similar to a state of surgical anesthesia (Grandin 2001), To minimize distress, To immobilize the bird to allow an easy and accurate neck cutting, and To prevent convulsion during bleeding out.

Most common stunning method is electrical stunning, although chemical and mechanical stunning methods also exist (Thiemig 1996a). Electrical stunning is usually carried out by dragging the heads of the birds that are hung on the shackle-line through water in which an electrode is submerged. The shackles of the line simultaneously touch the second electrode, causing an electric current (Parry 1989).

The next step, bleeding, is either initiated manually 5 to 10 seconds after stun-ning by passing a knife across the side of the neck which is typical for turkeys, or mechanically which is the more common method for chicken. In that case, the head is guided across a single, revolving, circular blade (Parry 1989). Bleeding re-duces the microbial contamination to a large extent (de Lourdes Pérez-Chabela and Guerrero Legarreta 2001). In contrast to beef and pork, poultry blood is not an edible by-product as it cannot be gained in a hygienic way (Freudenreich und Bach 1993). However, it has a significant value as fertilizer or feed component (Nielsen 1989).


3.7.4 Scalding and Eviscerating

Scalding is an absolute prerequisite to permit the complete removal of feathers (Varnam and Sutherland 1995). Two different systems are applied: while immer-sion in hot water is most common, spray scalding has the disadvantage of a high water usage and causes quality defects (Parry 1989). The scalding temperature is between 52°C for turkeys and 60°C for chickens. Higher temperatures have a positive effect on preventing microbiological spoilage, but a negative effect on consistency and color of the meat (Thiemig 1996a).

Defeathering involves a number of rubber flails or “rubber fingers” on a series of plucking machines. These machines consist of banks of counter-rotating domes or discs with the fingers mounted on them. The number and length of the ma-chines depends on the line speed. The feathers are extracted by the cohesion be-tween the rubber and the feather; passing the birds through an arc flame singes the remaining feathers. A spray wash is applied as defeathering significantly increases the level of contamination on the skin of the bird and bears the danger of cross-contamination (Mead 1989).

Finally, feet and head are removed and the carcasses are brought to the evis-ceration area. For poultry, the volume of all slaughter by-products (feathers, heads, feet, blood etc.) is around 40% of the living weight (van Sonsbeek et al. 1997). The evisceration area must be physically separated from the defeathering area. Evisceration is also performed mechanically and fully automated (up to 6.000 broilers per hour) with the birds suspended from the shackles of the con-veyor (Gill 2000). The edible viscera (heart, gizzard and liver) are removed from the remaining viscera on the slaughterhouse floor (Ockermann and Hansen 1988). They are transported to a central processing area where they are sorted, chilled and packed. A neck cracker separates the neck from the spinal column by pressure (Parry 1989). As incomplete evisceration can lead to rapid spoilage (in particular Salmonella at this stage), the birds receive another spray washing from interior and exterior directly after evisceration (Varnam and Sutherland 1995). For the fur-ther processing of slaughter by-products, it is referred to Ockermann and Hansen (1988) or Freudenreich and Bach (1993). Van Sonsbeek et al. (1997) propose a mixed-integer approach for strategic decisions in the slaughter by-product chain in order to find the optimal processing structure.

3.7.5 Chilling

After evisceration, the birds must be chilled from over 30°C below the legal maximum of 7°C, in most cases down to 2-4°C at the warmest point of the car-cass, for two reasons. On the one hand, rapid cooling is required to prevent prema-ture spoilage (Radespiel 1996; Varnam and Sutherland 1995). As the growth of microorganisms is a temperature-dependent process, it is absolutely necessary to reduce the temperature of the meat immediately after dressing (Cano-Munoz 1991). On the other hand, cutting can be performed much more easily and accu-rately if the carcass is cooled. Two major types of poultry chilling systems can be


distinguished: immersion chillers and air chillers. The immersion chilling systems consists of a series of large tanks in which the carcasses are cooled. The water moves in the opposite direction as the carcasses so that the carcasses leaving the tanks meet the cleanest water (Varnam and Sutherland 1995). However, in spite of the fact that the per-unit cost of immersion chillers are significantly lower than that of air chillers, they are increasingly replaced by air chillers as immersion chillers can constitute an important source of microbiological contamination. To-day, air chilling is used extensively in particular in Europe. The carcasses are transferred automatically by re-hanging equipment from the evisceration area to the chiller (Veerkamp 1989). After a preliminary period in an air blast tunnel, the cooling takes place in a chill room. The risk of cross-contamination is much lower (Varnam and Sutherland 1995); nevertheless, the meat loses weight through sur-face evaporation, which depends on the difference in temperature and the relative humidity between the meat and the environment (Cano-Munoz 1991), but also on the time spent in the chill room and the air circulation. The weight loss is usually between 1.5% and 2.5% of the eviscerated carcass, depending on the size and the type of the carcass (Wirth et al. 1990). Spraying the poultry with water by using a shower at the beginning of the chill blaster, or by using spray nozzles at the en-trance of the evaporative air chillers can help to limit the weight losses (Veerkamp 1989).

3.7.6 Rough Cutting

After chilling and grading, the bird can be sold as a “ready-to-cook” poultry (meaning as a whole carcass) or cut-up and further processed. The portion of fur-ther processed poultry has grown significantly during the last decades for several reasons (Baker and Bruce 1989):

Selling ready-to-cook carcasses is a highly competitive business due to the commodity character of the products. By further processing, unpredictable and unsaleable overproduction can be stored as products with a longer shelf life. Further processed products provide an excellent market for underutilized parts of the carcass such as necks and backs. The consumers have become much more convenient, thus the out-of-home and ready-to-eat market increased considerably.

The cutting and further processing process can be divided into a rough cutting as well as a fine cutting / processing part. In rough cutting, the entire carcass is cut-up into halves, quarters, or into its major components. At the end of the rough-cutting line, the rough-cuts are stored in bulk containers and put in an intermediate storage or directly brought to the fine cutting department. Several automatic cut-ting machines have been developed, however, at the beginning the cuts were im-precise and led to yield losses and quality problems. Today, in particular in the chicken segment, several processors operate fully automated cutting lines (Parry 1989).


3.7.7 Fine Cutting

The fine cutting and further processing of the meat chunks comprises a variety of operations and machines. These types of processes include for example portion-ing, deboning, tumbling, massaging, battering, or marinading. A high number of different final products are made from poultry meat such as cut-portions, rolls, es-calope, grill steaks, or nuggets. To produce a product, the rough-cut pieces are fur-ther dressed according to specific cutting-patterns. For example, the breast can be cut into escalope pieces of a specific weight. The involvement of machines varies considerable between cutting-patterns, from a simple putting together of some ta-bles up to a automated cutlet-cutter line. Nonetheless, each cutting-pattern requires a more or less important set-up and cleaning effort. The fine-cut products are again stored in bulk containers. As the surface of the fine-cut products is much bigger than that of the rough-cut parts and therefore the danger of microbial spoil-age increases, the final products should be packed as quickly as possible. A pre-requisite is the storage at around 0°C, a deviation of only 2°C can already have severe consequences in terms of microbial spoilage (Wirth et al. 1990). If a rough-cut or fine cut product cannot be sold as a fresh product, it is usually frozen. How-ever, fresh products yield a significantly higher revenue, therefore freezing is al-ways the second-best option. Several types of freezing equipment exist in the in-dustry, the most common being air blast tunnel freezers, liquid immersion freezing, and plate freezers (Veerkamp 1989).

3.7.8 Packaging

According to Bell (2001), both chilled and frozen meats are usually protected by flexible plastic packaging, which can take a variety of forms. Non-preservative packaging can be distinguished from preservative packaging. Non-preservative packaging such as wrappers, over-wrapped trays and loose-fitting plastic bags and pouches protect the product from water loss and contamination, but do not create in-pack conditions very different from the ambient conditions. Therefore, products in these packages are highly perishable and have a very short shelf life. More common in industrial practice is preservative packaging, which is characterized by the ability to extend shelf life through the modification or restriction of microbial growth as the in-pack conditions differ considerably of those of the ambient envi-ronment. The most important preservative packaging methods are vacuum packag-ing and modified / controlled atmosphere packaging. In these types of packaging, the atmosphere around the product is either withdrawn completely, or replaced by a gaseous environment. For instance, a high oxygen modified atmosphere packag-ing can extend the shelf life of fresh chicken meat from 2-4 days if only wrapped to 4-8 days (Bell 2001). The packaging process is generally relatively labor-intensive. Changeover times can be important due to the necessity of adjusting new packaging formats and labels. Furthermore, cleaning and sterilization are im-portant tasks as well, as microbiological spoilage at this stage can be extremely damaging to the customer’s health. A special characteristic of most meat packag-

3.8 Conclusion 87

ing lines is that scales are built in that weigh each product and stamp the weight on the packaging since most meat products are sold according to their real weight, not according to a fixed weight (e.g. 200g).


After packaging, the products are stored in a chilled warehouse. Since many retail customers demand individual packaging, the number of products increases mark-edly in the final products storage. The delivery to the retail customers is performed in refrigerated transport facilities that are designed to circulate cold air around the inner walls of the unit (Gill 2000). Due to the high perishability of the products, the delivery interval is usually daily. For fresh meat products, the portion of direct deliveries from the processor to the retail outlet is higher than in most other cate-gories; an additional handling step in a retail distribution center would require too much time. Although most retail outlets that sell fresh meat operate a facility on the premises to produce primal cuts, the centrally cut and packaged meat volume is continuously growing, especially in the discount channel (Gill 2000).

3.8 Conclusion

After a short introduction of the major fresh food segments, this chapter has dis-cussed structures, characteristics and developments of FFSCs and has looked at the specifics of fresh food production systems. The production systems of the three case studies of this research (yogurt, sausages and fresh poultry meat) have been analyzed in detail. FFSCs are subject to significant changes. FFSC partici-pants consolidate on all stages in the SC. Intensive pressure on costs and margins is exercised by the rise of private labels and discounters. Product proliferation and promotions steadily increase the complexity in FFSCs. All actors must invest sig-nificantly into new technologies and implement extensive food quality and safety systems. Fresh food production systems must in particular cope with perishability and variability at all production steps. Other major drivers of complexity include extensive and sequence-dependent changeover-times and costs as well as short lead-time requirements.

The following profile development of fresh food industries with regard to APS systems (see Chapter 4) is based on both the generic description of FFSCs (see Chapter 3.3) and of the production system (see Chapter 3.4), as well as on the de-tailed analysis of the three case study production systems (see Chapters 3.5 to 3.7). Furthermore, the MILP-models developed in Chapters 7-9 aim at integrating many of the described characteristics in order to provide an effective planning support.

4 The Fresh Food Industry’s Profile Regarding

APS Systems

4.1 Methodological Remarks

APS providers usually focus on one or two specific industries – either because the first successful implementations occurred in these industries or because the soft-ware engineers already had experience in one of these industries. No provider can currently cover all industries (Hauptmann and Zeier 2001). Hence, the imple-mented manufacturing processes, the terminology used, the planning processes and the reporting capabilities are focused on these industries (Kilger 2002b). Sev-eral authors (for example Fraunhofer Gesellschaft 1999; Cap Gemini Ernst & Young 2002a; Kilger 2002b; Zeier 2002d) provide surveys on the industry focus of APS systems. Although many providers claim experience in the food industry, it has to be analyzed in depth in which food industry the experience has been gained since fundamental differences exist between different food industries. However, no leading APS system originates from a fresh food industry. Hence, for the selection of an APS system, the specifics of the considered industry must be transformed into requirements. Then, it can be deciphered as to which system best fits to the requirements. Moreover, as shown in Chapter 2.6, it is crucial to exam-ine the requirements of individual industries as a basis to develop specific decision models for each production segment and planning step. Generic solution proce-dures will not be able to cover all aspects of specific decision problems. Hence, the models developed in Chapters 7 to 9 are based on the following requirements. To help identify and categorize software requirements of industries, some authors (for example Braun 1999; Zeier 2002a) propose the so-called core-shell-model. This model classifies the software functions into four types in order to distinguish industry-spanning and industry-specific requirements:

General or module-independent requirements are relevant for all indus-tries and have been implemented by almost all APS providers; Module-specific requirements are relevant for a specific APS module, but industry-spanning as well; Production process-specific requirements vary by operation type (Schaub and Zeier 2000; Hauptmann and Zeier 2001); Industry-specific requirements are only applicable for a single industry.

In the following analysis of the three case study industries (yogurt, sausage and poultry), emphasis is given to production-process and industry-specific require-

90 4 The Fresh Food Industry’s Profile Regarding APS Systems

ments. General and module-specific requirements that are valid for almost all in-dustries are only considered in this research if they are particularly important to one of the case study industries. Descriptions of these basic requirements (e.g. in-terfaces to MS-Office, user-friendliness, etc.) can be found for example in Zeier (2002a). As most companies do not implement all modules of an APS system at the same time, the requirements are structured around the typical modules of APS systems. All requirements are scored with regard to their importance for each case study industry. The scoring method is derived from the five-level scoring pro-posed by Jimenez et al. (1998) or Kilger (2002b), which is usually used to assess the degree of requirement’s coverage of specific APS systems (see Table 4.1). The score ranges from indispensable functions, which basically lead to the exclusion of the provider from the bidding process in case of non-fulfilment (Gronau 2001), to functions with no value-added to the considered industry. As companies within the same industry and even the different sites of the same company can vary consid-erably, the scoring should be understood as a rough indication of the require-ment’s importance and does not replace a company -and site-specific evaluation of requirements.

Table 4.1 Explanation of scores

Score Explanation

5Indispensable function for the considered industry (“must-have”); non-fulfilment will exclude provider from bidding process

4 Important, but not indispensable function (“should-have”)

3 Desirable, but dispensable function for the considered industry (“nice-to-have”)

2Function contains some value for the considered industry, however only to a limited extent

1 Function not important at all for specific industry

4.2 General Requirements

First of all, an APS system must fulfil some basic requirements in order to be suited for fresh food industries (e.g. food industry specific data structures or mod-ule-spanning reporting capabilities). Most of these requirements can be considered as indispensable; a non-fulfilment of one of these requirements makes the APS systems unsuitable for the case study industries.

GEN1 - Data exchange with fresh food ERP systems: The basic data for APS systems is provided by the company’s ERP systems. Although the SAP R/3 sys-tem has a widespread distribution in all industries, fresh food companies often use

4.2 General Requirements 91

other industry-specific ERP systems instead or in addition. A prominent example is the CSB system in the meat industry. The APS system must be able to retrieve real time data from these systems, to manipulate it and to write the results back to the ERP systems.

GEN2 - Support of EDI data: Although the EDI data is first transferred into the ERP system of the producer, the APS system must be able to work with these types of data. Examples of these data types are the EAN, the EAN 128, the inter-national location number or the number of the shipment unit and EANCOM mes-sages.

GEN3 - Support of the EPC and RFID: Even though most retailers do not yet require RFID, the APS systems should be prepared to handle RFID data such as the EPC, as it is most likely to become the future standard (Cap Gemini Ernst & Young 2002d). At the very least, the provider should aim at developing these functions soon.

GEN4 - Support of Internet-interfaces / XML: The Internet is increasingly used for inter-company communication. The support of XML-messages (eXtensible Markup Language) is a prerequisite for APS systems to allow web-enabled com-munication. Friedrich (2002) emphasizes that there is no common XML standard yet. Instead, ca. 100 XML-“dialects” are currently used which are generally not compatible with one another. The APS system must be able to manage the most important food industry dialects. As EANCOM and XML-messages will exist both in the mid-term, the APS system must support both types (Zeier 2002b).

GEN5 - Support of Internet-based marketplaces: As Internet-based market-places are the major pacemakers in inter-company collaboration processes, the APS system must be able to communicate with both the leading general market-places (e.g. Transora, WWRE, or GlobalNetExchange) and retailer-operated mar-ketplaces such as the E-Contor of the German discounter Plus (Rode 2004a).

GEN6 – Flexible collaboration: Most German fresh food companies are of mid-size and deliver to several retailers. As the retailers are frequently the center in the SC, the fresh food manufacturer must be able to handle several SCs. The APS sys-tem must on the one hand enable ad hoc, new cooperation; while on the other hand it must be possible to terminate cooperation with only minor efforts. Friedrich (2002) suggests a “flexible click-in” into a SC using a “plug and play” SCM soft-ware as the most efficient way to cope with this issue. A flexible click-in leads to lower relationship-specific investments (asset-specificity) and hence to lower transaction costs (for a detailed analysis of transaction costs and asset-specificity see Rindfleisch and Heide 1997).

GEN7 - Support of POS data: POS data is direct input for all leading retailer’s SCs (Cap Gemini Ernst & Young 2002e). In case of SC collaboration, fresh food manufacturers are also provided with this POS data in order to level off high variations in the SC (“bullwhip-effect”). Hence, the APS system must be able to read and process POS data of the most common cash register systems.


GEN8 - Variable Units of Measure: The methods of measurement of fresh food materials differ between different companies and different stages in the SC even for the same material. For example, raw milk can be measured in liters or kg; yo-gurt as final product is measured in kg, packages or pallets. To be successful in fresh food industries, APS systems must allow the administration of different units of measure and to convert the volumes from one unit to the other (N.N. 2002b).

GEN9 - Country-specific characteristics: Since fresh food companies are be-coming increasingly international, the APS systems must support country-specific characteristics. A good first indicator of the internationality of an APS system is the number of supported languages (Kortmann and Lessing 2000). Other aspects include currencies, taxes, export/import duties and different legal requirements concerning food handling and safety.

GEN9 - Tracking of fresh food specific KPIs: The monitoring and controlling functions of the APS system must be adapted to the specifics of fresh food indus-tries and should also include other SC partners (Mebus 2004). Most commonly used KPIs for FFSCs are given below; for more SC performance measures, it is referred for example to van der Vorst (2000) and Sürie and Wagner (2002):

Forecasting accuracy, Stock-outs (central warehouse, DC, and shelf), On-time deliveries and service levels (a “must” according to Thonemann et al. 2003),Inventory obsolescence due to expiry of products, Inventory level in the chain (in production, distribution and in the shelf), orShelf life output (freshness).

Table 4.2 General requirements

N° General Requirements Relevance for production of

Yogurt Sausage Poultry

GEN1 Data exchange with fresh food ERP systems 5 5 5 GEN2 Support of EDI data 5 5 5 GEN3 Support of the EPC and RFID 4 4 4 GEN4 Support of Internet-interfaces / XML 5 5 5 GEN5 Support of Internet-based marketplaces 5 5 5 GEN6 Flexible collaboration 4 4 4 GEN7 Support of POS data 4 4 4 GEN8 Variable units of measure 5 5 5 GEN9 Country-specific characteristics 5 5 5 GEN10 Tracking of fresh food specific KPIs 5 5 5

4.3 Requirements for Strategic Network Design 93

4.3 Requirements for Strategic Network Design

The necessity to re-design the SC is of rising importance for fresh food industries due to its consolidation and internationalization (Hauptmann and Zeier 2001). A typical task performed by the SND module is the analysis of the consequences of a delivery agreement for private labels, which must be delivered in high volumes to many parts of a country and which greatly affects the SC.

SND1 - Supplier selection and assignment of suppliers to sites: Farmers are the main suppliers of fresh food manufacturers. Many small entities in sometimes very remote locations characterize the farming stage of the SC. The SND module must therefore support the decision as to which farmers to include in the SC and to which site to assign each farmer. The raw material collection costs are particularly important in the dairy industry, since fresh milk must be transported every day or every second day. In the meat industry, the transportation frequency is lower; nev-ertheless special constraints on transportation times for living animals must also be considered.

SND2 - Center-of-Gravity modeling: For distribution-intense industries such as fresh foods, the center-of gravity modeling allows to situate sites and warehouses by minimizing material flows or costs (Cap Gemini Ernst & Young 2002f). This method is useful for both the determination of the supply and the distribution net-work and provides the highest value in case of dairy networks with frequent trans-ports.

SND3 - Selection of processing technology: The selection of the processing technology is important as well since the technology to be implemented follows the volume that will be processed at a site. For example, in sausage production, a batch cutter is sufficient to process small volumes, while a continuous cutter automat is required for larger volumes. Moreover, with regard to the capacity de-termination of capital-intensive packaging lines, this function is important for all three sample industries.

SND4 - Service level by time and distance: The service level is an essential suc-cess factor in fresh food industries. Thus, Cap Gemini Ernst & Young (2002f) stress that an APS system should provide the capability to model these service levels by both time (e.g. 99.5% service level within one hour of transport) and dis-tance (e.g. 99.5% within 50 km). Time and distance determine the position of warehouses and the demarcation of delivery areas per site and per warehouse.

SND5 - Determination of inventory levels: Due to the high risk of inventory ob-solescence of fresh foods, the determination of inventory levels should already be considered in the SND module, especially in correlation with service levels. In-ventory levels are a required input figure to determine the size of warehousing fa-cilities. In the calculation of inventory carrying costs, the risk of inventory obso-lescence must be reflected as well.


SND6 - Determination of transportation strategy: The transportation strategy in fresh food industries is particularly concerned with lead-times and delivery in-tervals. The SND module should enable the planner to evaluate the costs and benefits of individual lead-times and delivery interval strategies, which have a high impact on warehouse sizes, truck characteristics, and the production set-up.

SND7 - Calculation of transportation rates: Usually, the rates of LSPs are used as input factors to the SC model. However, as these rates reflect the network struc-ture of the LSP and not that of the manufacturer, the transportation rates should be calculated bottom-up, including truck costs, driver costs and gasoline (Cap Gemini Ernst & Young 2002f).

SND8 - Integration of fresh food risk factors: The APS system should enable the modeling of fresh food industry specific risk factors and their effect on the SC (Beamon 1998). Examples are high fluctuations in demand for specific products or animal diseases in some areas leading to raw material shortages. As these factors can put the entire company at risk, the SND module must make the consequences of these effects visible in order to make the FFSC flexible and robust enough to resist these sudden shocks (Goetschalckx 2002).

SND9 - Support of use reusable packaging usage decision: Reusable packaging is common in some fresh food industries (e.g. dairy). Within the SND module, a support for the decision of which type of packaging to use should be integrated. An effective decision support must take into account the entire SC and environ-mental aspects in addition.

SND10 - Integration of shelf life: Shelf life as a key element should be integral part of all planning activities, from the supplier to the retail outlet. The shelf life of product puts a constraint on stock levels, production intervals or transportation times. Therefore, besides cost elements, time is another factor that must be inte-grated when defining the SC.

SND11 - Multi-objective optimization: As planning models and decisions in SND have both interrelated spatial and temporal characteristics (Goetschalckx 2002), the optimization algorithms incorporated in the SND module must be able to handle multi-objective optimization. For example, time factors such as shelf life or the responsiveness of a SC can become equally important to cost or profit as-pects. The system must enable the planner to assign individual weighting factors to each objective.

4.4 Requirements for Demand Planning 95

Table 4.3 Requirements for Strategic Network Planning

N° Requirements for Strategic Network Planning Relevance for production of


SND1 Supplier selection / assignment of suppliers to sites 4 2 3 SND2 Center-of-Gravity modeling 4 3 3 SND3 Selection of processing technology 4 4 4 SND4 Service level by time and distance 4 4 4 SND5 Determination of inventory levels 4 4 4 SND6 Determination of transportation strategy 4 4 4 SND7 Calculation of transportation rates 3 3 3 SND8 Integration of fresh food risk factors 5 5 5 SND9 Support reusable packaging usage decision 3 1 1 SND10 Integration of shelf life 4 4 4 SND11 Multi-objective optimization 4 4 4

4.4 Requirements for Demand Planning

The DP module is of outstanding importance for fresh food industries since the customer order decoupling point lies relatively late in the SC, the holding of in-ventories is expensive and stock-outs reduce the customer’s trust in the company (Fleischmann et al. 2002). Most decisions must be made before the actual order arrives (Hauptmann and Zeier 2001). The high importance of the DP module is confirmed by Cap Gemini Ernst& Yong’s study (2002e) stating that specialized demand management software providers make ca. 60% of their turnover in the Netherlands with CPG manufacturers and retailers (ca. 40% on an European scale). Forecasting is further complicated by the necessity to provide many prod-uct variants (van Beek et al. 1998). Finally, due to the limited product shelf life, the holding of inventories is only possible to a limited extent and relatively expen-sive due to decay of the products. On the other hand, low inventory levels run the risk of stock outs. To cope with all these requirements, the DP module must be ex-tremely powerful in order to effectively support forecasting in fresh food indus-tries.

DP1 - Multiple forecasting techniques: Fresh food products show an immense variety in terms of their sales volume development. All kinds of demand patterns such as trends, seasonalities, or periodic demand occur. Different types of influ-encing factors must be considered as well (e.g. the weather or the temperature, competitor actions, promotions and new product launches, effects of pricing etc.). Hence, the DP module of the APS system must provide a variety of forecasting techniques, from simple moving-average and exponential smoothing techniques to Croston methodology and advanced causal methods.

DP2 - Automated forecasting method selection: Based on the analysis of his-torical demand data, the APS systems should automatically generate a proposition for a forecasting method with the highest degree of accuracy (“pick-best method-ology”) and determine the weight for each influencing factor (Arminger 2002).


Typically, 2-3 complete seasonal cycles are required to initialize the forecast (Wagner 2002). The selected method must be explained in detail so that the plan-ner knows how the forecast has been calculated. However, frequent changes of the forecasting method must be avoided to limit the forecast nervousness.

DP3 - Combination of forecasting techniques: The APS system should further allow the use of several forecasting methods simultaneously. Bayesian Forecasting for example makes a weighted average of multiple forecasting methods based on the quality of each of the individual forecasting methods, thus providing a much more stable forecast as it avoids complete switching of the forecasting method (Davies et al. 2002).

DP4 - Pre-defined industry-specific demand patterns: As most APS providers have deep industry know-how in one or two specific industries, they also know the typical demand patterns of these industries. This know-how should be transformed into industry-specific demand patterns, which a manufacturer can adapt and cali-brate with his own data (Fraunhofer Gesellschaft 1999).

DP5 - Automated analysis of forecasting accuracy: The APS system should automatically analyze the forecasting accuracy and the applied forecasting meth-ods. If the forecast error exceeds a predefined threshold, the system should alert the planner (Zeier 2002b). An individualized feedback on their planning perform-ance should be given to the people involved in the forecasting process (Fraunhofer Gesellschaft 1999).

DP6 - Automated and self-adaptive forecasting: Due to the high number of products, customers and sales regions in fresh food industries, the planner does not have time to plan all these figures manually. Therefore, the APS system should automatically generate the forecasts; the planner should only pay attention to products or customers with the highest volumes or to exceptions (Zeier 2002b). Furthermore, the forecasting methods should be self-adaptive so that they can be operated after initial configuration without the manual intervention of specialists (Cap Gemini Ernst & Young 2002e).

DP7 - Top-down and bottom-up forecasting: Both top-down and bottom-up forecasting modes are required in order to integrate the top-down strategic direc-tion setting of the board and bottom-up planning based on for example key ac-counts or sales persons in order to have one figure throughout the company (Kortmann and Lessing 2000). It is crucial to clearly distinguish pure sales target setting from realistic forecasting on which the production should be based.

DP8 - Data aggregation and disaggregation: Any part of the forecast at any level of aggregation should be made available to the planner. Important data splits in fresh food industries are the splits by product, by product group, by product family, by customer, by customer group, by DC and by region. For example, the preferences for fresh food specialties (e.g. sausages) can vary extensively even by micro-region (Hauptmann and Zeier 2001).


DP9 - Forecasting of product variants: As the number of product variants can be very high in fresh food industries (e.g. due to individual packaging for each re-tail chain), the forecast is often generated on product group level and then broken-down into products and product variants. This breakdown should be performed automatically by the system relying on historical distributions of variants within a product group and actual parameters such as advertising campaigns.

DP10 - Cannibalization: The fact that the increase of the sales volume of one product can have a negative effect on the sales volume of other products is called cannibalization. Product cannibalization must particularly be taken into account in the case of new product introductions and promotions.

DP11 - Life cycle planning: The consideration of product life cycles within fore-casting is notably relevant for products with shorter life cycles (e.g. yogurt prod-ucts). Two stages of the life cycle are critical. For new products at the beginning of the cycle, there is no history available to apply statistical forecasting methods. At the end of the cycle, the risk of stock obsolescence makes inventory holding relatively expensive, especially for products with short shelf lifes. Therefore, spe-cial “in- and out-phasing curves” are required to accurately forecast these products (Hauptmann and Zeier 2001). Specific forecasting algorithms are necessary to cope with these highly volatile sales patterns (Cap Gemini Ernst & Young 2002e). The system should allow overtaking the parameters of a similar product to the new product. Moreover, supersession should be possible (phase-out of one or several old and phase-in of one or several new products in parallel).

DP12 - Forecasting of dependent demand: Fresh food production systems are in most cases characterized by divergent product structures, e.g. a bird is cut into many different products. As many final products rely on the same raw material, the DP module should derive the dependent demand of raw materials by summa-rizing the forecasts of the individual products for a first rough-cut capacity check.

DP13 - Extensive information basis: Numerous factors such as the temperature or competitor actions influence the development of demand in fresh food indus-tries. To ease the ex-post analysis of these factors and its consequences on the de-mand, historical demand data and information on all major important influencing factors should be stored in an extensive database (Hauptmann and Zeier 2000). Furthermore, numerous other kinds of information must be integrated to increase forecasting accuracy. Examples include market research information (in German fresh food industries in particular AC Nielsen and Gesellschaft für Konsumgüter-forschung), open customer orders or POS data from the retail. The APS system should provide a support to include this information.

DP14 - Integration of stock-out data: The primary source to forecast future de-mand is historical demand data. However, in most ERP systems, the actual de-mand is registered in the form of sales or warehouse exits. As stock-outs are not registered, the demand has been higher than the recorded number (Tempelmeier 1999a). Hence, it is necessary to provide the possibility to record the stock-out


volumes and un-fulfilled demand. It should be possible to gather a part of this data with the ATP module by registering unfulfilled customer orders.

DP15 - Price and revenue optimization: Price and revenue optimization aim at optimizing sales prices and revenues in situations of volatile supply and demand. In these situations, forecasting includes not only the sales per unit, but also the price. Examples from the meat industry are the volatile raw material markets for pork and beef. With regard to the three sample industries, this function is most important in poultry processing, as the shelf life for this product is very short and the meat can be bought and sold on spot markets. On the contrary, for the produc-tion of yogurt the function is only of minor importance as most products have a customer specific packaging and cannot be sold to other channels. A major chal-lenge for pricing optimization is that the source and the adjustments to price come from various departments within the organization and from SC partners (TEC Group N.N., SCM Software Selection Template).

DP16 - Analysis of terms and conditions: In general, relationships between manufacturer and retailer are regulated by a complex terms and conditions system. As up to 30-40 different types of discounts (for example volume discounts, reve-nue discounts, slotting allowances, payment agreements, contributions to promo-tions) can exist between a manufacturer and a retailer (Seifert 2001), current terms-and-conditions systems are not transparent and lead to high administrative costs. To manage these complicated systems, the APS should allow assessing the resulting costs and benefits of specific terms and conditions and their benefits on the demand.

DP17 - Deviation of shelf space requirements: The availability of shelf space in the retail outlet is a critical parameter for the forecasting of food products. Based on the forecasted volumes per article, the APS system should support the calcula-tion of the necessary shelf space in the retail outlet (Hauptmann and Zeier 2001) and consider possible limitations when generating the forecast.

DP18 - Forecasting of promotions and events: Promotional volumes should al-ways be forecasted separately from the standard volumes of a product because promotions can cause exorbitant variations in demand, especially with regard to price-promotions in markets with a high price-elasticity (Thonemann et al. 2003). Several functions are necessary to forecast promotional volumes accurately. A li-brary of past promotions and causal analysis methods help to forecast promotional demand, to calculate the price-demand-elasticity and to build a catalogue of pro-motion patterns (Zeier 2002c). Special factors such as cannibalization, smart-shoppers, and the planning of promotion sets, which incorporate several different products, must be taken into consideration (Zeier 2002d). For the assessment of possible competitor actions and the effects of the accompanying marketing cam-paign, the collaboration component of DP should be used intensively in order to integrate all departments within the company and all relevant trading partners. Upon completion, the system should be able to analyze the success of the promo-tion (e.g. in terms of contribution margin or ROI), since currently most promo-tions do not create real value for the companies.


DP19 - Consensus-based forecasting / intra-company collaboration: The more information is collected and processed during the forecasting process, the better the forecast (Zeier 2002b). With regard to fresh food industries, it is essential to integrate the perspectives of the marketing, sales and procurement departments in order to generate an accurate forecast. All departments should work out their plans, which must then be integrated and consolidated into one single, company-wide forecast so that only one plan exists within the company at the end. All em-ployees involved need an electronic access to the forecasts.

DP20 - Inter-company collaboration: Inter-company collaboration in demand planning and forecasting is the key element of the CPFR concept and has been the major development in demand planning in the recent years (Cap Gemini Ernst & Young 2002e). The Voluntary Interindustry Commerce Standards Association has established a standard for collaborative demand forecasting. Several activities must be supported by an APS system to enable efficient and effective collabora-tion in demand planning and to comply with the CPFR demand planning require-ments. First of all, external customer’s demand data must be captured by the APS system, either by allowing the customers to enter their demand information di-rectly and on-line into the system, or by defining an interface to the customer’s planning systems (Cap Gemini Ernst & Young 2002a). The system should offer multiple possibilities to integrate partner forecast data such as EDI, imports from Customer Relationship Management (CRM) systems, Excel spreadsheets, or di-rectly via the Internet (Davies et al. 2002). The DP module should then as a sec-ond step analyze the data in order to match the total forecasted demand and the sum of all customer-specific demand. In case of critical deviations (e.g. manufac-turer’s versus retailer’s forecast, actual forecast versus past forecast, or forecast versus actual POS sales), an escalation and clearing process must be implemented in order to finally agree on one number in the SC (Seifert 2002b).

DP21 - Integration with CRM systems: CRM systems have been implemented at many fresh food manufacturers in the last five years. Starting with the electronic recording of orders, these systems today cover a broad range of functions such as order workflow, document management, data integration between key account management, sales force and office support, assortment planning (for the assort-ment planning of the Metro Group see for example Wietschel 2004), or shelf op-timization (Bjorksten and Knopf 1999; Schwetz 2003). They provide a standard-ized support for the relationship with customers with regard to the marketing, sales and after-sales cycle and should allow presenting a uniform image to the cus-tomer (Knolmayer et al. 2002). CRM systems rely strongly on demand forecasts, making a tight integration between the DP module and the CRM system neces-sary. Moreover, most CRM systems contain basic forecasting functions as well (mainly aggregated manual forecasts of sales employees). Consequently, the inte-gration of both systems is also necessary in order to avoid several forecasts within a corporation (Cap Gemini Ernst & Young 2002e).


Table 4.4 Requirements for Demand Planning

N° Requirements for Demand Planning Relevance for production of


DP1 Multiple forecasting techniques 5 5 5 DP2 Automated forecasting method selection 4 4 4 DP3 Combination of forecasting techniques 4 4 4 DP4 Pre-defined industry-specific demand patterns 3 3 3 DP5 Automated analysis of forecasting accuracy 5 5 5 DP6 Automated and self-adaptive forecasting 5 5 5 DP7 Top-down and bottom-up forecasting 4 4 4 DP8 Data aggregation and disaggregation 5 5 5 DP9 Forecasting of product variants 4 4 4 DP10 Cannibalization 4 4 4 DP11 Life cycle planning 4 3 3 DP12 Forecasting of dependent demand 4 3 4 DP13 Extensive information basis 4 4 4 DP14 Integration of stock-out data 4 4 4 DP15 Price and revenue optimization 2 3 4 DP16 Analysis of terms and conditions 3 3 3 DP17 Deviation of shelf space requirements 3 3 3 DP18 Forecasting of promotions and events 5 5 5

DP19Consensus-based forecasting / intra-company collaboration

5 5 5

DP20 Inter-company collaboration 4 4 4 DP21 Integration with CRM systems 4 4 4

4.5 Requirements for Supply Network Planning

SNP has its highest importance in industries in which the production takes place before the customer order is known (Zeier 2002c), which is the case for most fresh food products. Moreover, the supply of raw material is often variable only in the mid-term (e.g. one growing cycle in poultry production can take several months). Hence, the major task of SNP for fresh food industries is to accurately derive the raw material supply volumes. Furthermore, it has to assure high service levels while keeping stock levels down (Hauptmann and Zeier 2001). Nevertheless, with regard to the SNP module, the requirements of fresh food industries do not differ considerably from the requirements of other industries.

SNP1 - Variable raw material characteristics: When planning the raw material supply in fresh food industries, it is indispensable to integrate factors that cause variation in the material characteristics. For example, the fat contents of raw milk or the weight of living birds fluctuates depending on the seasons. Planning has to consider these characteristics; otherwise the supply volume will not be accurate.

SNP2 - Integration of shelf life: As inventories of raw materials and finished products are perishable and subject to decay, the SNP module must consider these factors as constraints for the supply network plan. They lead to lower inventory

4.6 Requirements for Purchasing & Materials Requirements Planning 101

levels and higher set-up times and cost due to higher production frequencies, which will finally reduce the available capacities.

SNP3 - Make-or-buy decision support: In case of high capacity utilization, fresh food producers occasionally use third party production capacities in addition to their own capacities. On the one hand, fresh food manufacturers buy semi-finished products from other manufacturers (e.g. rough-cut poultry); on the other hand they even outsource the production of complete products. Thus, the APS must be able to consider third party production capacities in case of own capacity shortage (Tempelmeier 1999a). Furthermore, in case of raw material shortages, the soft-ware must consider buying raw materials on the spot market.

SNP4 - Determination of overtime requirements: Although a clear trend to-wards more working time flexibility can be observed, most food manufacturers still have a relatively low work time flexibility. Therefore, changes in the working time pattern must be announced in the mid-term and are a result of the supply network plan (Fleischmann et al. 2002).

Table 4.5 Requirements for Supply Network Planning

N° Requirements for Supply Network Planning Relevance for production of


SNP1 Variable raw material characteristics 5 2 5 SNP2 Integration of shelf life 5 5 5 SNP3 Make-or-buy decision support 4 4 4 SNP4 Determination of overtime requirements 4 4 4

4.6 Requirements for Purchasing & Materials Requirements Planning

Based on the determinations of the supply network plan, the MRP part of the module “explodes” the independent demand to derive the dependent demand that can be divided into raw materials and packaging/ingredients. While the sourcing of packaging material and ingredients is a “must” for almost all industries, fresh food industries have some specific characteristics concerning the raw materials.

P&MRP1 - Divergent BOM and recipes: The typical BOM in discrete manufac-turing are convergent and have a family tree structure, which means that many dif-ferent parts are assembled to one final product (e.g. to a car or to an aircraft). However, in fresh food industries, the BOM is rather inverted or divergent. For example, a piece of poultry is processed into many items such as breast or legs (N.N. 2002b). Moreover, many fresh food industries use recipes instead of BOMs. In contrast to a BOM, recipes include information on equipment, routing and tim-ing for processing as well. All this information must be covered by an APS system for fresh food industries.


P&MRP2 - Versions: Despite the high numbers of products in fresh food indus-tries, most products are relatively similar to other products. Frequently, the only differentiating factor between two products is the packaging material or a taste in-gredient. Thus, it is desirable that the APS system supports a simple generation of new BOMs/recipes by copying from other products, or by relying on a product family BOM/recipe with different versions.

P&MRP3 - Multiple raw material sourcing: Multiple raw material sourcing is relevant in fresh food industries for two reasons. First, many different farmers usually deliver to a manufacturer and therefore the manufacturer must be able to manage a broad supply network. Secondly, different supply channels must also be enabled by the system. For example, in the case of peak demand, raw materials are usually bought on the spot market, from other manufacturers, or are imported.

P&MRP4 - Integration of shelf life: Shelf Life must be considered as a con-straint on the supply side because most raw materials (e.g. raw milk or meat for sausage production) are highly perishable and must be processed within a rela-tively short period of time.

P&MRP5 - Raw material blocking for testing: Some raw materials can only be released to the production after having passed the obligatory food quality and safety tests (e.g. microbiological and chemical testing of raw milk). The raw mate-rial must therefore be blocked in the system until positive test results are available.

P&MRP6 - Livestock planning: From the supply network plan, the necessary raw material quantities per time period can be derived. Particularly in the meat in-dustry, the supply of raw material (living animals) must be tightly coordinated with the farmers due to high lead-times. As much of the total livestock production (in particular of white meats such as pork and poultry) is highly vertically inte-grated (Harvey 2004), an integrated planning concept can easily be applied. The APS system should support the planning of the livestock, based on the supply network plan and considering factors such as the gender and breed of the animals, the expected grade, seasonal effects and available capacities of farmers.

P&MRP7 - Batch traceability: With regard to food safety and quality, the trace-ability of batches and products will become an important issue in the near future (see Chapter 3.3.5). To fulfil the requirements of EU regulations, the traceability of products must be ensured throughout the entire chain. For the manufacturer this means that already the raw material supply has to be monitored and traced. More-over, all raw materials must be unambiguously identified and documented.

P&MRP8 - Management of reusable packaging: In the dairy industry, reusable packaging is relatively common for milk or yogurt products. When determining the supply volumes for reusable packaging, the APS system must consider how often this type of packaging will be used and order only that number of packaging units that fails the inspection and must be replaced. Bloemhof-Ruwaard et al. (2002) state that in the dairy industry a reusable bottle is used 27 times on average.

4.7 Requirements for Production Planning and Production Scheduling 103

P&MRP9 - Integration with SRM systems: In analogy to the CRM systems, SRM systems are a relatively new type of software to manage external collabora-tion with suppliers (TEC Group N.N., SCM Software Selection Template). In fresh food industries, SRM systems are relevant for most sourcing groups other than raw materials (e.g. ingredients or packaging material). Within SRM systems, both strategic sourcing and operative purchasing processes are supported. In the strategic sourcing part, the manufacturer can identify potential suppliers and nego-tiate agreements using a connection to e-marketplaces or the request-for-informa-tion, request-for-proposal and e-auction process in order to conclude an outline agreement. The operative purchasing part supports automated ordering within the terms and conditions of the outline agreement, as well as e-fulfilment functions. A tight integration of the APS and the SRM system is necessary since both strategic sourcing and operative purchasing rely on information generated by the planning process (e.g. volume forecast for strategic sourcing or order volumes per item after BOM/recipe explosion).

Table 4.6 Requirements for Purchasing & Materials Requirements Planning

N° Requirements for Purchasing & Materials Relevance for production of

Requirements Planning Yogurt Sausage Poultry

P&MRP1 Divergent BOM and Recipes 5 5 5 P&MRP2 Versions 4 4 4 P&MRP3 Multiple raw material sourcing 4 3 4 P&MRP4 Integration of shelf life 5 5 5 P&MRP5 Raw material blocking for testing 4 4 4 P&MRP6 Livestock planning 1 1 4 P&MRP7 Batch traceability 5 5 5 P&MRP8 Management of reusable packaging 3 1 1 P&MRP9 Integration with SRM systems 3 3 3

4.7 Requirements for Production Planning and Production Scheduling

As production systems differ between industries and even between companies and sites, the need for an individualized solution is the highest for PP and PS (Günther and Tempelmeier 2000; Cap Gemini Ernst & Young 2002a; Zeier 2002e), these modules are relatively difficult to implement. Accordingly, the list of the require-ments for PP and PS is the longest of all modules.

PP&PS1 - Push and pull production: In fresh food industries, it is common practice that push and pull production occur simultaneously. For example, raw milk supply or living animal input is pushed into production, while consumer de-mand triggers a pull production. Push production often leads to material surplus whose utilization has to be managed by the system (e.g. freezing in case of poultry meat or using raw milk for UHT-milk production in the dairy industry).


PP&PS2 - Multiple decoupling points: The use of different decoupling points is widespread in fresh food industries. Standard, high volume products are often produced according to MTS principles, while MTO production or “Packaging-to-order” dominates for low volume specialties. Hence, an APS system must manage several customer order decoupling points.

PP&PS3 - Block planning: The objective of the block production is the minimi-zation of set-up and cleaning times and costs by the subsequent production of similar products. Within a block, a “natural” order of batches often exists that al-lows avoiding set-up times completely (Günther and Neuhaus 2004). In fresh food industries, the batches of a block can for example change from the weak to the strong color (e.g. yogurt). The generation of blocks reduces the production plan-ning and scheduling decision problem significantly due to the decrease of the number of binary variables. An APS system should therefore support both the se-quencing of batches within a block and the scheduling of blocks on a production or packaging line.

PP&PS4 - Special process industry optimization algorithms: The resolution of planning problems in process industries by LP or MILP models requires special algorithms that lie beyond the range of functions offered by most APS systems. Hauptmann and Zeier (2001) recommend the use of component-ware algorithms because no APS system can provide algorithms for all industry-specific planning problems. For example, the company ILOG offers so-called “cartridges” that can be linked to the APS and which run the optimization for the APS when special op-timization problems occur. However, an APS system focusing on fresh food in-dustries should provide algorithms for the most frequently occurring optimization problems (e.g. scheduling of packaging lines, block planning support).

PP&PS5 - Multi-objective optimization: Similar to other modules, the APS sys-tem should take into account multi-objective optimization on the production plan-ning and scheduling level as well. Dickersbach (2003), for example, favors a mix of lateness, lead-time and set-up times for production scheduling. In fresh food in-dustries with high emphasis on shelf life, a higher weight will be given to produc-tion schedules that maximize tardiness meaning that the production takes place as close as possible to the delivery date in order to not lose shelf life days.

PP&PS6 - Integration of shelf life as a constraint: Perishable product planning is regarded as an important differentiating factor for APS systems because the ca-pabilities differ notably between providers (Cap Gemini Ernst & Young 2002a). Most APS systems solve problems of peak demand by building up stocks, using spare capacity from preceding days or weeks. However, this procedure is almost impossible for products with short shelf life as in fresh food industries (N.N. 2002b). Therefore, the APS system must consider the shelf life of products as a constraint for production planning.

PP&PS7 - Optimization of shelf life output: The shelf life output should be sub-ject to customer-specific optimization since a longer residual shelf life has a higher value for some customers than for others. Also with regard to products, the


APS system should allow the optimization of the shelf life output because a sup-plemental day of shelf life is more important for products with a short shelf life than for products with a longer shelf life.

PP&PS8 - Customer requirements on shelf life: The PP and the PS module must consider the customer requirements on shelf life as a constraint. As a conse-quence, in special cases the remaining time to produce and deliver the products becomes very short, leading to lower batch sizes and higher production frequency.

PP&PS9 - Variable pricing in batch sizing and scheduling: In fresh foods in-dustries, prices are not always fixed. On the contrary, prices can depend on factors such as the remaining shelf life of the product or the raw material supply. There-fore, for the short-term production planning, variable prices have to be integrated that should be linkable to specific factors. Variable pricing is notably important for fresh poultry, but also for yogurt and sausage.

PP&PS10 - Shrinkage calculation: Some fresh food products lose weight when being stored. Examples are the water evaporation of sausages in the intermediate storage before packaging or the loss of meat when being cooled from the slaughter temperature of around 37°C to the processing temperature of ca. 0-4°C. The APS must consider this loss with regard to further planning steps in order to plan with the accurate quantities.

PP&PS11 - Correction of material input and process output: Products and processes in fresh food industries are subject to high stochastic variations in terms of quality, yield and processing times. For example, the catchweights of ham and pork loins can vary to a certain degree (Morris 2000). Although the APS systems can rely on averages for long- and mid-term planning, on the short-term planning level the accurate information (e.g. on catchweights) must be reported back to the APS system after checking the quality of the materials or after processing in order to base the upcoming planning steps on actual information (N.N. 2002b). The use of tolerances allows limiting the re-planning to situations in which major devia-tions occur.

PP&PS12 - Optimization of recipes: Due to the variable quality characteristics and costs of raw materials, the application of a standard recipe does not always yield the same results. The recipe has to be optimized according to the individual characteristics of the delivered raw material. As the raw materials for sausages are particularly expensive, recipe optimization yields significant financial benefits in that industry. Jahn and Wäscher (1999) describe the recipe optimization of a sau-sage manufacturer in detail. The optimization algorithms in APS systems can in most cases easily solve the resulting recipe optimization decision problems.

PP&PS13 - Planning of co- and by-products: In some fresh food industries, in particular in the meat industry, the manufacturing process yields not only the de-sired product, but also co- and by-products (Morris 2000; Cap Gemini Ernst & Young 2002a; co- and by-products are explained in detail in Chapter 3.4.3). Ex-amples are offal or blood in poultry slaughtering, cream in dairy production or re-moved tendons in sausage production. Planning of these products is necessary to


determine for example the capacity requirements for the disposal of by-products or to define the further utilization of co-products.

PP&PS14 - Integration of grades: Both raw materials and final products can have variable quality characteristics. In order to assess the different qualities, grades are given to these products, which are usually based on internationally rec-ognized grading standards. Grading is notably important in the meat industry; the grading process for poultry is given for example in Parry (1989). Frequently, ship-ments to customers can be constrained on grades, e.g. customer A gets only B-grades (Cap Gemini Ernst & Young 2002a). Hence, it is essential that APS sys-tems support the attribution of different, pre-defined grades within production planning. In this context, the CSB system offers an optimization engine that sup-ports the division of raw materials into standardized groups (Schimitzek 2004).

PP&PS15 - Combined divergent, convergent and cyclic material flows: As de-scribed in Chapter 4.6, besides convergent BOMs (e.g. for the final product in-cluding packaging materials) the production of fresh food mainly relies on recipes (e.g. dairy) and inverted BOMs (e.g. poultry). Moreover, the material flow can be cyclic. In sausage production, for example, slicing waste is re-integrated in regular production.

PP&PS16 - Working with patterns: The application of cutting-patterns is typical for the carcass disaggregation in the meat industry. For example, a poultry breast can be sold as a whole or cut into schnitzel of different sizes. For each product type, one or more pre-defined cutting-patterns including the target quantities of the resulting intermediate or final products must be applied.

PP&PS17 - Working with yields: In meat processing, several factors influence the yield of a cutting process. Although the exact yield of the process can only be determined after the process is completed, some aspects must be taken into ac-count by the system when planning the expected outcome of a process. Examples include the degree of automation (a higher automation usually leads to a lower yield) or the experience and performance of the cutting team that can differ sig-nificantly between sites or even between shifts.

PP&PS18 – Simultaneous determination of lot-sizes and schedules: On the one hand, alterations in the sequence of batches can influence the lot sizing deci-sion due to high set-up times and costs. On the other hand, sequencing is based on given lot sizes, and therefore lot-sizes and schedules must be determined simulta-neously (Fleischmann et al. 2002).

PP&PS19 – Pre- and post set-up times and costs: Set-up times and costs are es-sential in fresh food industries. Pre set-ups include for instance the sterilization of a line before producing or the calibration of slicing equipment for sausages. Post set-up is mainly concerned with cleaning. Sometimes, it is also necessary to spec-ify a maximum amount of time for which the next set-up can be suspended in or-der to avoid a renewed sterilization of the line (N.N. 2002b).


PP&PS20 - Sequence-dependent changeovers: In industrial fresh food produc-tion, the set-up, sterilization and cleaning times and costs often depend on the se-quence of the batches on a specific line. For instance, if a strong taste is set-up be-fore a weak taste, intense cleaning is required. However, if the order is turned (weak taste before strong taste), the cleaning efforts can be reduced to a minimum or be neglected at all. An APS system should provide the possibility to incorporate these times and costs in form of a matrix. With regard to the three case study in-dustries, sequence-dependent changeovers are particularly relevant in yogurt and sausage production.

PP&PS21 - Time dependence of production steps: Minimal and maximal time periods between production steps are characteristic for process industries and must be considered in some fresh food industries (Hauptmann and Zeier 2001). With regard to yogurt, the fermented milk has to be filled within a pre-defined time pe-riod. In poultry processing, rough-cut and fine cut meat must be packed within a relatively short period of time.

PP&PS22 - Order splits: The splitting of large production orders is relevant in fresh food industries in particular with regard to shelf life issues. An APS system should support both order splits over time on the same line, and splits into multi-ple smaller orders to be processed on several lines (Cap Gemini Ernst & Young 2002a).

PP&PS23 - Planning of silos, tanks and vessels: In the industries considered here, silo-, tank- and vessel planning is primarily relevant in the dairy industry, but the ovens in sausage production may also be taken as an example. Planning of these equipment types is mandatory due to the fact that a utilization of these facili-ties of less than 100% makes the remaining space unavailable for other products. Furthermore, many tank resources are used for both production and storage. Hence, the available capacity is reduced noticeably (N.N. 2002b). Only a com-plete withdrawal of the product permits the silo or tank to be filled again.

PP&PS24 - Multi-purpose production units: Many units in fresh food indus-tries can be used for several purposes. Examples in the dairy industry are multi-purpose tanks that can be used for fermentation, pasteurization or storage of milk. The system must consider that these units cannot be used simultaneously for sev-eral processes.

PP&PS25 - Product-specific capacities: The utilization of capacities can vary according to the type of product. An example is the scalding chamber utilization in sausage production that differs according to the diameter of the sausages. Consid-ering the capacity utilization is essential in order to calculate the correct capaci-ties.

PP&PS26 - Traceability of batches: The traceability of batches and products must also be ensured in production. An APS system for fresh foods must offer this function to trace which raw material batch was transformed into which production batch and further into which products in order to comply with the upcoming EU legislation.


PP&PS27 - Integration of quality requirements: Fresh food industries are gen-erally subject to strict quality requirements (see Chapter 3.3.5). HACCP and GMP impose frequent quality measurements before, during and after the production. In some cases, a batch must be blocked by the system until the quality check has been performed. In case of longer-term product holds due to quality issues, the in-formation must be immediately considered in the APS system to be able to re-plan the production accordingly.

PP&PS28 - Management of reusable packaging: The returning packaging must be tested and cleaned before being reused again. The APS system must ensure that both sufficient testing and cleaning capacity is provided and enough bottles are available.

Table 4.7 Requirements for Production Planning and Production Scheduling

N° Requirements for Production Planning Relevance for production of

and Production Scheduling Yogurt Sausage Poultry

PP&PS1 Push and pull production 5 3 5 PP&PS2 Multiple decoupling points 4 5 4 PP&PS3 Block planning 5 4 2 PP&PS4 Special process industry optimization algorithms 4 4 2 PP&PS5 Multi-objective optimization 4 4 4 PP&PS6 Integration of shelf life as a constraint 5 5 5 PP&PS7 Optimization of shelf life output 4 4 4 PP&PS8 Customer requirements on shelf life 5 5 5 PP&PS9 Variable pricing in batch sizing and scheduling 2 2 4 PP&PS10 Shrinkage calculation 1 4 4 PP&PS11 Correction of material input and process output 5 5 5 PP&PS12 Optimization of recipes 2 5 2 PP&PS13 Planning of co- and by-products 4 3 5 PP&PS14 Integration of grades 2 4 4

PP&PS15Combined divergent, convergent and cyclic material flows

3 4 4

PP&PS16 Working with patterns 1 1 5 PP&PS17 Working with yields 3 3 5

PP&PS18Simultaneous determination of lot-sizes and schedules

5 5 2

PP&PS19 Pre- and post set-up times and costs 5 5 3 PP&PS20 Sequence-dependent changeovers 5 4 2 PP&PS21 Time dependence of production steps 5 4 5 PP&PS22 Order splits 4 4 3 PP&PS23 Planning of silos, tanks and vessels 5 5 1 PP&PS24 Multi-purpose production units 5 2 3 PP&PS25 Product-specific capacities 4 4 4 PP&PS26 Traceability of batches 5 5 5 PP&PS27 Integration of quality requirements 4 4 4 PP&PS28 Management of reusable packaging 3 1 1

4.8 Requirements for Distribution Planning 109

4.8 Requirements for Distribution Planning

In non-durable consumer industries such as fresh food industries, distribution costs constitute the second most important cost factor after the purchasing spend (Knolmayer et al. 2002). As a result, the accurate planning of distribution and transportation can yield considerable financial benefits for a fresh food producer. DisP and TP are very useful in multi-echelon distribution networks with tight de-livery plans and trucks that are only partially loaded (Zeier 2002c). For that rea-son, distribution-intense industries are expected to benefit most from this module (Berger 1999).

DIS1 - Safety stock calculation with seasonality: Based on the stock levels cal-culated in the SNP module, an APS system for fresh food industries must be able to calculate safety stock levels considering the variability of demand, the forecast accuracy, product lead-times and lead-time variations as well as customer service level requirements (Cap Gemini Ernst & Young 2002a). Furthermore, since the safety stocks can vary with the season, seasonality must also be considered when determining safety stocks. Separate formulas should be used for different product-location combinations, e.g. differentiated by the product inventory turns in a spe-cific location (Davies et al. 2002).

DIS2 – Allocation of safety stocks: In a multi-stage distribution environment as it can be found in FFSCs, the total amount of safety stocks must be allocated to sin-gle stages of the distribution system such as the central warehouse or the retail DC (Fleischmann et al. 2002).

DIS3 - Administration of service levels: The service level is one of the most im-portant KPIs in fresh food industries. In the service level driven inventory optimi-zation, the resulting inventory level is derived from the required service level per customer or product. As fluctuating future demand must be taken into account, the inventory level will also fluctuate accordingly (Cap Gemini Ernst & Young 2002b).

DIS4 - Product blocking and quarantine times: In fresh food industries, it is common that intermediate or final products have to spend at least a minimum time in the warehouse. This can be caused by obligatory quality checks (e.g. veterinary) after a certain time or by the fact that products have to mature or ferment before they can be sold. Therefore, blocking times must be taken into account as a restric-tion. Furthermore, it might be necessary to block products in case of quality con-cerns (quarantine time). The products can only be released after a resolution of the concern.

DIS5 - Product phase-in and phase-out: The introduction of new products into the retail is often handicapped by high inventories in the retail DC and outlet (Seifert 2001). Although this is only a minor a problem for products with a very short shelf life, it can already become an issue for products with a shelf life of some weeks or months. To effectively support new product introductions, the APS


system should be able to determine the optimal point in time for the product intro-duction and decrease the inventory levels of the old product accordingly.

DIS6 - Different temperature zones: Warehouses in fresh food industries are of-ten divided into a frozen, a chilled and an ambient temperature zone. When plan-ning the inventory levels in each warehouse, the system must consider capacity constraints by temperature zone.

DIS7 - Integration of shelf life: The integration of shelf life is a decisive factor when planning the distribution in fresh food industries. Due to decay, the value of the product strongly decreases once the minimum customer requirement on shelf life has been exceeded. After that, the product can usually only be sold with high price discounts or have even to be wasted. Thus, the development of the product value over time must be incorporated into the APS system.

DIS8 - Customer requirements on shelf life: Most customers impose minimum requirements on the remaining product shelf life, which must be respected by the APS system when planning the distribution. Products with less shelf life than the requirement cannot be shipped to customers.

DIS9 - Alert monitor for inventories with risk of obsolescence: As perishable products decrease markedly in value if they get older than the minimum customer requirement on shelf life, an alert should be given to the planner in order to make him actively sell these products.

DIS10 - Support of VMI: A key element of an effective VMI-support at cus-tomer warehouses is the capability to integrate retail data such as the demand forecast, current inventory levels, POS data, or in-transit inventories. Based on this data, the manufacturer determines the actual order size (Seifert 2001). The manufacturer must be enabled by the system to calculate this optimal order size and to release the order on behalf of the customer (Zeier 2002b).

DIS11 - Support of CPFR planning steps: In the CPFR concept, the agreed sales forecast (see Chapter 4.4) is transformed into a forecast of single orders con-sidering the constraints of the retailer and the manufacturer such as lead-times, ac-tual inventories, service levels, or delivery constraints (Bastok et al. 2002). Both manufacturer and retailer must agree upon these orders. In case of critical devia-tions, an escalation and clearing process is activated (in analogy to the forecasting process). Finally, these orders must be transformed into real orders, similar to the VMI concept (Busch et al. 2002).

DIS12 - Traceability of batches and products: The traceability of batches and products must further be ensured following the production step. Each batch and product, either in the warehouse, on a truck, or already delivered to the customer must be unambiguously identifiable by the system.

4.9 Requirements for Transport Planning 111

Table 4.8 Requirements for Distribution Planning

N° Requirements for Distribution Planning Relevance for production of


DIS1 Safety stock calculation with seasonality 5 5 5 DIS2 Allocation of safety stocks 3 3 2 DIS3 Administration of service levels 4 4 4 DIS4 Product blocking times 5 5 5 DIS5 Product phase-in and phase-out 3 3 2 DIS6 Different temperature zones 4 4 4 DIS7 Integration of shelf life 5 5 5 DIS8 Customer requirements on shelf life 5 5 5 DIS9 Alert monitor for inventories with obsolescence risk 4 4 4 DIS10 Support of VMI 4 4 4 DIS11 Support of CPFR planning steps 4 4 4 DIS12 Traceability of batches and products 5 5 5

4.9 Requirements for Transport Planning

The short-term TP module plans the demand for transportation services resulting from the distribution plan by determining the transportation mode, the load and the route (Zeier 2002b). Optimized transport planning (in cooperation with DisP) can have an increasingly important monetary effect in fresh food industries as the average transportation distances still continue to grow (Bernsmann and Bone 2003).

TP1 - Selection of transportation mode: The predominant transportation mode in fresh food industries is transportation by truck. Even with respect to the planned truck toll in Germany, rail is no real alternative (Bernsmann and Bone 2003). A decision has to be made as to which kind of truck to use (e.g. own fleet, rental truck, forwarder, or pick up by retailers).

TP2 - Route optimization: The generation of pick-up or milk-run tours can yield important benefits in fresh food industries. However, numerous constraints must be respected in route optimization, e.g. distances between locations, the average speed on different road types and different at times-of-day, real time traffic infor-mation and weather, or different loading times (Cap Gemini Ernst & Young 2002c). Moreover, linking the vehicle to the IT of the company permits dynamic route planning (Knolmayer et al. 2002) and real time control of the truck.

TP3: Pallet load optimization: In order to increase vehicle utilization, the system must analyze the physical product attributes to determine how to best pack pallets and vehicles. When looking specifically at fresh food industries, for the packing of the pallet the APS system must consider different product temperatures, different product and packaging stabilities and the picking order in the retail outlet. More-over, for the placement of the pallet in the truck, the most determining factor is the order in which the pallets are going to be unloaded. In addition, the loading of re-


turnable pallets must be considered as well as different temperature zones of a truck.

TP4 - Multi-objective optimization: In analogy to the optimization runs in the SND, SNP and PP&PS modules, several objectives must be considered simultane-ously when optimizing the transport plan in fresh food industries. Besides cost factors, the system should consider the time as a component in the optimization to respect shelf life, as well as service levels. Regarding time, it can be useful to serve outlets with a higher volume first to increase the average product freshness of all deliveries. With respect to the service level, it must be considered that uncer-tainties exist (e.g. regarding the traffic or the weather) and that many outlets can only be delivered within a limited delivery time window.

TP5 - Transport packaging planning: Numerous different transport packages are used in the retail environment. Seifert (2001), for example, remarks that over 50 different transport packages must be processed at a transshipment point. The TP module must be able to handle this variety and assign the correct transport package to a shipment. Furthermore, in case of returnable or pool pallets such as the Chep system (see for example Loderhose 2004), the pallets must be accounted for and the return of the pallets must be handled. Vis and Roodbergen (2002) de-scribe how the reverse flow of carts can be optimized in order to avoid out-of-stocks due to delivery incapability.

TP6 - Reusable product packaging planning: Although environmentally advan-tageous, the essential drawback of reusable packaging such as polycarbonate or glass bottles is that they use as much space filled on the outward journey as empty on the way back (Bloemhof-Ruwaard et al. 2002). Thus, the TP module must con-sider the inbound freight of used packages as well in setting up the transportation plan.

TP7 - Less than Truckload (LTL): Due to the high frequency of ordering, the average transportation load in fresh food industries is relatively small and only seldom fills a full truck. If a 3PL is used, it is common that only a part of the pos-sible total load is available for the fresh food manufacturer since the 3PL usually serves several customers with one truck. The TP module must consider these par-tial load constraints when optimizing the transport. Furthermore, the rates of the 3PL generally differ between LTL and Full Truck Load.

TP8 - Support of inbound logistics: Due to the agricultural character of the raw materials, the inbound transportation in fresh food industries can become very complex. This holds true especially in the dairy and in the meat industry when the raw materials are collected from farmers who are spread throughout the country. An effective planning support can yield comparable benefits for outbound logis-tics.

TP9 - Cross Docking: To enable cross docking, the manufacturer must pack the pallets so that they can be delivered directly to the retail outlet without further picking in the DC. Consequently, the retailer must split the orders by retail outlet.

4.9 Requirements for Transport Planning 113

The manufacturer’s APS system must then be able to handle these outlet-specific orders and plan the pallet loading accordingly.

TP10 - Roll cage sequencing: In the concept of roll cage sequencing, the prod-ucts are picked in the order in which the products are unpacked later in the outlet (Seifert 2001). To support roll cage sequencing, information on the unpacking or-der of products in the outlet must be integrated into the APS system so that the picking of the products can be performed accordingly.

TP11 - Multi-temperature trucks: In particular for the direct delivery to smaller outlets, the use of multi-temperature trucks is common (Seifert 2001). These trucks have different temperature zones, Smith and Sparks (2004) distinguish fro-zen chill (-25°C for ice cream and -18°C for other foods), cold chill (0°C to +1°C, e.g. for fresh meat, poultry and most dairy products), medium chill (+5°C, e.g. for pastry, butter and cheese) and exotic chill (+10°C to +15°C, e.g. for potatoes or eggs). By using multi-temperature trucks, a dairy manufacturer can for example deliver UHT-milk, yogurt and ice cream simultaneously. The TP module then has to consider the capacity constraints of each temperature zone when scheduling the trucks.

TP12 - Hazardous materials transports: The by-products in the meat industry are partly classified as hazardous materials and require special treatment and transportation (e.g. the transportation of offal or blood). In addition to the trans-portation of raw materials and final products, the APS system must be able to plan these transports as well, considering special legal restrictions on the disposal of hazardous materials.

TP13 - Transports of living animals: The transport of living animals has some restrictions on both the employed trucks and the transportation times which the APS system must respect in order to ensure animalfriendly transportation. The trucks must have specific configurations, in particular in the case of poultry trans-portation. Moreover, the transportation times for living animals are legally limited which reduces the catchment area of a slaughterhouse and the ability to coordinate round-trips.

TP14 - Blending in tanker transports: In general, raw milk is collected by tanker trucks, in which the milk supply of several farmers gets blended. To ensure traceability of the raw milk, the supplied volumes of the farmers must be attached to this newly generated batch.

TP15 - Tracking & tracing: According to Cap Gemini Ernst & Young (2002c), tracking is the maintenance of the status information of shipments and equipment, while tracing is checking the movement of shipments and equipment. In fresh food industries, the importance of these two functions stems from the need for tracing the products throughout the entire chain (e.g. for product recalls etc.), as well as from increasing customer requirements concerning delivery reliability. Real time tracking & tracing requires that the trucks be fitted with in-vehicle-computer systems that allow sending and receiving messages (Knolmayer et al. 2002).


TP16 - External collaboration for carrier selection: As transportation logistics are already outsourced to a high degree in fresh food industries (for example McKinnon 2004), an external collaboration helps to cover peak demands for transportation capacities (Cap Gemini Ernst & Young 2002a). In particular, this holds true for the trunk haulage activities (McKinnon 2004). To tender and finally to book a specific freight or a bundle of freights, several techniques such as EDI, hyperlinks to a website, or XML documents should be available (Kilger and Reuter 2002; Knolmayer et al. 2002).

Table 4.9 Requirements for Transport Planning

N° Requirements for Distribution Planning Relevance for production of


TP1 Selection of transportation mode 3 3 3 TP2 Route optimization 4 4 4 TP3 Pallet load optimization 4 4 4 TP4 Multi-objective optimization 4 4 4 TP5 Transport packaging planning 4 4 4 TP6 Reusable product packaging planning 3 1 1 TP7 Less than truckload 4 4 4 TP8 Support of inbound logistics 5 2 4 TP9 Cross Docking 4 4 4 TP10 Roll cage sequencing 3 3 3 TP11 Multi-temperature trucks 5 3 3 TP12 Hazardous materials transports 2 2 5 TP13 Transports of living animals 1 1 5 TP14 Blending in tanker transports 4 1 1 TP15 Tracking & tracing 5 5 5 TP16 External collaboration for carrier selection 3 3 3

4.10 Requirements for Demand Fulfilment and Available-to-Promise

In an MTS environment as it can be found in fresh food industries (see Chapter 2.2.9), ATP is generally represented at the finished goods level (Kilger and Schneeweiss 2002a). The large number of product variants makes the ATP proce-dures very difficult to execute (Knolmayer et al. 2002). As most product variants are customer-specific, the proposition of alternative variants is often impossible. Furthermore, satisfying the order from another location is also difficult due to shelf life constraints. Consequently, the re-planning of production commitments is of higher importance in order to avoid losing the order.

ATP1: Capable-to-Promise: If an order cannot be served from stock or by the al-ready planned volumes, the system has to check if the required volume can be produced. Therefore, the system must review production commitments on the in-termediate stock level or even on the raw material level and re-plan the production

4.10 Requirements for Demand Fulfilment and Available-to-Promise 115

(see Chapter 2.2.9). As fresh food customers generally order short-term, CTP is notably relevant in a production environment with short lead-times and relatively high production flexibility such as poultry production.

ATP2: Profitable-to-Promise: Furthermore, the APS systems should take into account the profitability of an order to support the fulfilment decision. This holds true especially for products with variable sales prices (e.g. products that approach the BBD). A prerequisite is that cost structures are available per product to deter-mine the contribution margins and the profitability. However, in a production en-vironment with many different product variants, clear product cost accounting is difficult, especially when considering inventory and variant-dependent fixed costs (Knolmayer et al. 2002). Therefore, a Profitable-to-Promise support is desirable.

ATP3: Available-to-Customer: Available-to-Customer can be seen as an ATP function for the entire chain down to the consumer, taking into account the time to deliver the products from a DC to the customer (Davies et al. 2002). Thus, it is a better tool to limit stock-outs than the simple ATP.

ATP4: Various volume allocation rules: Due to the decay of products, a manu-facturer endeavors to keep stock levels small; as a consequence product shortages are relatively common situations. Then, the available volume must be allocated according to pre-determined rules such as quotas or customer priorities (see Chap-ter 2.2.9). For fresh food industries, the APS system should provide different shortage allocation schemes. The most relevant are (Fischer 2001):

Allocation of volume proportionally to customer orders, Allocation of volume proportionally to target stock levels in case of VMI, Allocation of volume proportionally to customer priority or customer turnover,Allocation of volume according to case-specific management instruc-tions,Forecast consumption, Price increase, and Auctioning of the volume.

ATP5: Integration of shelf life: Most customers demand pure delivery batches (only one shelf life). In such cases, it is not possible to use volumes of two batches with different shelf lives to fulfil the delivery requirement (Zeier 2002c). Further-more, most customers impose a minimum remaining shelf life of the products, which has to be considered by the ATP module as well.

ATP6: Dynamic pricing: To fully support all requirements of fresh food indus-tries, the ATP functionality should not only investigate if and when an order can be served, but also give the corresponding price. This function is of high impor-tance for products that reach the BBD and that can only be sold with discounts. The underlying time-value function is generally not linear; the decrease in value starts slowly and increases drastically towards the end of the shelf life.


ATP7: Automated ATP: Retailers usually order smaller volumes, which leads to a high number of orders. In order to avoid costly manual handling, the ATP proc-ess should be automated as much as possible. Furthermore, the ATP module should be connected to the retailers ordering system to further increase ordering efficiency.

ATP8: Interface to CRM system: The confirmed and the rejected customer or-ders are indispensable information for the sales management. For this reason, the CRM system must be provided with this information via a real time interface.

Table 4.10 Requirements for Demand Fulfilment and Available-to-Promise

N° Requirements for Demand Fulfilment and Relevance for production of

Available-to-Promise Yogurt Sausage Poultry

ATP1 Capable-to-Promise 4 4 5 ATP2 Profitable-to-Promise 4 4 4 ATP3 Available-to-Customer 3 3 3 ATP4 Various volume allocation rules 4 4 5 ATP5 Integration of shelf life 5 5 5 ATP6 Dynamic pricing 2 3 4 ATP7 Automated ATP 5 5 5 ATP8 Interface to CRM system 4 4 4

4.11 Conclusion

For the implementation of APS systems in fresh food industries, many different industry-specific requirements must be respected and integrated. The highest need for customization arises for the DP, DisP, and PP/PS modules. Moreover, as indi-vidual fresh food industries vary considerably from others, the requirements re-garding APS systems also differ between these industries. Three sample industries (yogurt, sausage and fresh poultry) have been analyzed in detail and the impor-tance of the requirements for each of the three industries has been assessed.

Some requirements are important for all fresh food industries and concern sev-eral modules. For example, the necessity to collaborate intensively with SC part-ners and particularly with retailers requires collaboration components in various modules (e.g. DP, DisP, TP, or ATP). Furthermore, the integration of shelf life is decisive for almost all modules, from the long-term oriented SND over mid-term SNP and DisP to short-term PP/PS and TP. The consideration of the perishability of products is also a principal driver for the need for multi-objective optimization algorithms. Therefore, the following chapters concentrate on this most distinguish-ing characteristic of fresh food industries. The factors that cause and influence shelf life are analyzed in detail in Chapter 5. In Chapter 6, leading APS systems are assessed with regard to their coverage of shelf life in each of their models. Fi-nally, in Chapters 7 to 9, MILP models are developed that integrate shelf life at the operational production planning stage.

5 Shelf Life in Fresh Food Industries

5.1 Shelf Life of Food Products

5.1.1 Definition and Limiting Factors

Product freshness is one of the most important buying criteria for consumers (see also chapter 3.3.4). Sloan et al. (1986) argue that “freshness” has even replaced “price” as the primary concern of consumers regarding food. Reasons for this de-velopment are that today the consumer has a higher level of education, more knowledge about nutrition and more money available to spend for food products. For consumers, besides purity and safety, product freshness is a major part of product quality and will help to prevent health problems. Today, the consumer has two main sources of information concerning product freshness. On the one hand, he can use his senses to evaluate the sensory qualities of the product. This is pos-sible for all products that are only slightly packed, e.g. fruits or vegetables. How-ever, for fully packed products such as most dairy products or many meat prod-ucts, the consumer does not get in touch with the product itself. In that case, he must rely on the information provided by the shelf life date in order to assess the freshness of this product.

There is no generally accepted definition in literature for the term shelf life. For Hine (1987), shelf life refers to “the duration of that period, between packing a product and using it, for which the quality of the product remains acceptable to the product user”. Labuza and Taoukis (1990) define shelf life as the period in which the food “will retain an acceptable level of eating quality from a safety and or-ganoleptic point of view”. As consumers also demand consistently high food qual-ity between purchase and consumption, Kilcast and Subramaniam (2000) empha-size that the shelf life labeling should reflect these requirements by not only considering safety issues, but unwanted changes in sensory quality as well. Within their research, they use the shelf life definition issued by the Institute of Food Sci-ence and Technology according to which shelf life is defined as “the time during

which the food product will remain safe, be certain to retain the sensory, chemi-

cal, physical and microbiological characteristics, and comply with any label dec-laration of nutritional data”. The Food and Agricultural Organization of the United Nations and the World Health Organization provide more detailed and workable definitions, however, they do not refer explicitly to shelf life (Barberio 1986). Besides the Sell-By-Date (“the last date of offer for sale to the consumer

after which there remains a reasonable storage period in the home”) and the Use-

By-Date (“recommended last consumption date”); they introduce in particular the

118 5 Shelf Life in Fresh Food Industries

notion of the Date-of-Minimum-Durability or Best-Before-Date which is “theend of the period under any stated storage conditions during which the product

will remain fully marketable and retain any specific qualities for which tacit or

express claims have been made”. Nonetheless, also beyond the BBD the product may comply with all requirements. Within this research, the term shelf life refers generally to the BBD definition given above, which is comparable to the German “Mindesthaltbarkeitsdatum”.

In general, the manufacturer determines the shelf life of a product (Barberio 1986). The actual shelf life of a product depends on four major factors: formula-tion, processing, packaging and storage conditions. All of these factors are critical; however, their relative importance depends on the perishability of the product. With respect to formulation, it is decisive to choose the right raw materials and in-gredients and to ensure that they have not run out of their shelf life. Within proc-essing, the final characteristics of the product are achieved by changing the physi-cal and chemical characteristics of the raw materials. The packaging and storage conditions determine the shelf life of a product to a large extent as they have a big influence on the microenvironment of the product. Important dimensions are the gas composition, the relative humidity, pressure, light, and temperature (Labuza and Taoukis 1990). Besides the storage conditions at the manufacturer’s facilities, the conditions during transport and particularly in the retail outlet are important as well. For this reason, manufacturers increasingly aim at educating the retailers to be cautious when handling their products (Barberio 1986). On the other side, re-tailers increasingly demand lower delivery temperatures in order to avoid shelf life issues in the outlets. This, however, will require significant supplemental invest-ments by the manufacturers in the future (Murmann 2004c).

Nonetheless, even if properly handled and stored, fresh food products have generally a shelf life of less than 14 days (Labuza and Taoukis 1990). Several mechanisms cause the deterioration of the food product (Kilcast and Subramaniam 2000). In most cases, the shelf life of food products is determined by biochemical

or microbial decay (Labuza and Taoukis 1990). The growth of microorganisms is influenced by a variety of parameters such as the moisture content and the relative humidity, the pH, the gas composition in the packaging, the presence of preserva-tives, and also the initial spoilage of the product (Kilcast and Subramaniam 2000). Chemical changes of the product notably include oxidation, which occurs in vari-ous forms (e.g. as rancidity of fats or respiration and ripening of fresh fruits). Moreover, intensive light can cause negative effects on food products such as color fading or activity loss of vitamins (Hine 1987). Major reason for physical

deterioration is moisture migration, which is possible in both directions. On the one hand, fresh produce for example ages by losing water. On the other hand, other products can deteriorate by water uptake (Kilcast and Subramaniam 2000). Finally, the temperature has a principal influence on the rate and nature of other causes and is therefore decisive for the shelf life of most products (Hine 1987).

5.1 Shelf Life of Food Products 119

5.1.2 Determination of Shelf Life

As all stated factors and mechanisms operate interactively and often unpredictably (Kilcast and Subramaniam 2000), it is often very difficult or even impossible to predict shelf life of food products precisely. Due to the high diversity of food products, there is no standard method for determining shelf life. Most manufactur-ers have developed their own measuring procedures (Barberio 1986), which can be categorized into sensory and instrumental methods. On the one hand, sensory techniques are the backbone of shelf life assessments (Kilcast 2000). As trained panels usually perform the sensory evaluations, they give a good estimate on the overall state of a food (Labuza 2000), but they are relatively expensive and time consuming (Kilcast and Subramaniam 2000). On the other hand, instrumental methods cover a broad range of chemical, microbiological and physical tests and are widely used to study the quality of a food product. For enumerations of avail-able testing methods and further explanations, it is referred to Barberio (1986) and Labuza (2000). Kilcast and Subramaniam (2000) emphasize that all instrumental techniques must be validated against sensory measures; the instrumentally meas-ured parameters are only valid if they correlate with sensory test results.

Nonetheless, despite this broad range of available measurement procedures, for the prediction of shelf life of a new product most R&D departments rely on “edu-cated guesses”, supported by limited experiments and measurements, due to time constraints. Five principal approaches can be found in practice (according to La-buza and Taoukis 1990): First, the shelf life can be estimated based on literature values or shelf lives of similar products of a company. Secondly, the distribution times for similar products can be used as an indication for the determination of shelf life. However, if no testing is carried out, the manufacturer is taking a con-siderable risk. Thirdly, for distribution abuse tests products are collected in the re-tail and analyzed in the laboratory under at home conditions. Fourthly, consumer complaints can give an idea about the shelf life of a product knowing that for each complain there are 50-60 other spoiled food cases that are not indicated. Finally, Accelerated Shelf Life Tests have gained considerable importance in the recent decades. They refer to methods to evaluate product stability, based on data ob-tained in a significantly shorter period than the product shelf life. Precondition is that the deterioration process has a valid kinetic model (Mizrahi 2000). The fin-ished product is stored under some abuse conditions and periodically examined until the end of shelf life. A shelf life prediction for true distribution conditions is then based on the gained data (Labuza 2000).

Time-Temperature Indicators constitute a relatively new technology to moni-tor the shelf life of a product throughout the SC. According to Labuza (2000), a it is a “simple, inexpensive device that can show an easily measurable, time-

temperature dependent change that reflects the full or partial temperature history of a food product to which it is attached”. Time-Temperature Indicators are based on a mechanical, (electro-) chemical, enzymatic, or microbiological irreversible change that is cumulative in nature, while the rate of change depends on the tem-perature (Taoukis 2001). For that reason, they are able to indicate the remaining shelf life of a product as a function of time and temperature. As the labeled shelf


life on the packaging of a product supposes that the required cold chain is main-tained (Bhushan and Gummaraju N.N.), they are better suited to indicate the “real” shelf life of a food product. Bhushan and Gummaraju (N.N.) favor the combination of Time-Temperature Indicators with wireless technology in order to better manage inventory levels in the cold chain. Their importance within FFSCs is expected to increase with the rise of RFID technology in the forthcoming years.

5.1.3 Technological Shelf Life Extensions

The need to extend the shelf life of products arises from several reasons. Sloan et al. (1986) focus their research on the convenience aspect. As traditional family structures have changed (e.g. decreasing average household size, a larger portion of women working), eating patterns, food choices and preparation have changed accordingly. In particular, the time available for shopping decreased considerably. Therefore, it is more convenient to rely on products with longer shelf life that can be stocked at home, thus avoiding frequent shopping. Furthermore, additional re-quirements on shelf life are caused by the increasing transportation times of fresh products due to the growing importance of economies-of-scale in production and to the trend to more and more exotic products. In addition, consumers demand seasonal products to be available throughout the year. On the other side, consum-ers associate very long shelf lives with poor product quality (Kilcast and Subra-maniam 2000). They expect food products with more sensory appeal and less ad-ditives as well as optimized, minimal processing. To get as close as possible to achieving these almost irreconcilable objectives of longer shelf life and less proc-essing, optimization at all production and distribution stages is necessary (Labuza and Taoukis 1990).

The attempt to extend shelf life and to preserve foods is certainly not new. Techniques used in the past by ancient civilizations include drying in the sun, pre-serving with snow and ice, smoking, salting and fermenting, e.g. with regard to al-coholic beverages. The second half of the nineteenth century has brought a couple of new techniques such as food canning, refrigerated shipping or pasteurization (Barberio 1986).

When looking specifically at perishable products, their shelf life is mainly de-termined by the ability to control microbial growth. Microbial growth can be in-fluenced at several points in the FFSC, e.g. by the selection of raw materials or by optimizing storage and distribution conditions. However, the two most important areas to archive a longer product shelf life are certainly processing and packaging; intensive research is carried out in these fields. With regard to the processing stage, the product shelf life can be extended by either killing the microorganisms (e.g. by heat or radiation) or by limiting their growth. The latter can be achieved by reducing the temperature, by reducing water activity or by adding preservatives (Kilcast and Subramaniam 2000). Van Boekel (1998) provides a comprehensive overview of the latest developments in food stabilization techniques. He empha-sizes that heat treatment is the most widely used food stabilization technology. Nonetheless, heating can affect the quality of food as some chemical reactions are

5.2 Shelf Life Characteristics of Case Study Products 121

accelerated, causing loss of nutritional value and organoleptic changes. Therefore, new stabilization technologies focus on damaging the final product as little as pos-sible. Examples are high-pressure treatment, minimal processing, or sous-vide.

On the other side, advances in packaging materials and technologies allow to extend the shelf life of food products. For example, aseptic packaging plays an important role in the dairy industry. In that case, the product and the packaging are free from undesired microorganisms (van Boekel 1998). Moreover, modified-atmosphere packaging is another technology that has first been applied in the meat industry and which is now being extended to fruits and vegetables, bread, cheese and other fresh food products. The technology relies on altering the atmosphere around the product from air to one that limits microbial growth (e.g. carbon diox-ide or nitrogen). By modified-atmosphere packaging, the natural shelf life has been increased by two to ten times. For vacuum packaging as a special form of modified-atmosphere packaging, the atmosphere is removed fully (Emblem 2000). Finally, active packaging materials interactively influence the shelf life of foods. Examples are oxygen absorbers or scavengers, ethylene absorbers or carbon diox-ide emitters (Kilcast and Subramaniam 2000).

Despite all the attempts to extend the shelf life of a food product as carefully as possible to keep the nutritional and organoleptic properties and despite the pro-gress that has been made, consumers remain skeptical, particularly regarding the use of preservatives in food (Sloan et al. 1986). Special attention is given to the freshness of a product. Therefore, the remaining shelf life of the product for the consumer can be extended further by producing the product as close as possible to the consumer demand in the retail. This will increase both the product shelf life for the consumer and the freshness of the product. The consequences of producing closer to the demand on the FFSC are discussed in detail in Chapter 5.3.2.

5.2 Shelf Life Characteristics of Case Study Products

5.2.1 Case Study 1: Shelf Life of Yogurt

As all fresh products, yogurt has a relatively short shelf life (Menz 1986). Pro-duced under regular conditions, the yogurt shelf life is about 8-10 days at tempera-tures below 10°C (Spreer 1995). The application of different preservation tech-nologies allows extending the shelf life to three to four weeks (see below). The low temperature slows down the biological and biochemical reactions in the coag-ulum (Tamime and Robinson 1999). The changes that occur after manufacturing and packaging are mainly associated with the physical separation of phases and with the growth of microorganisms. The main limitation for the shelf life of yogurt products is the spoilage by bacteria, moulds and yeasts that grow even at refrigera-tion temperatures (Muir and Banks 2000). In addition, Kilcast and Subramaniam (2000) name syneresis (leading to serum separation) and oxidation (leading to rancidity) as major limiting factors for yogurt shelf life.


In order to ensure the labeled shelf life, special attention must be paid to vari-ous factors. First of all, the respect of the prescribed temperatures is of prime im-portance, not only at the manufacturer’s site, but also during the transport and in the retail outlet. Furthermore, the recommendations for hygienic manufacture and the implementation of an HACCP concept are critical elements to ensure shelf life (Tamime and Robinson 1999). Zikakis (1986) underlines that notably the prepara-tion of the starter culture demands a maximum of accuracy and hygiene.

Following the trend to concentrate production capacities, to extend the deliv-ered markets and to increase the product portfolio, many manufacturers have in-creased the shelf life of their products up to four weeks. Three main technologies are available to extend shelf life for yogurt products (Spreer 1995). Chemical

preservation (in most cases by adding sorbic acid) limits the growth of moulds and yeast; however, food legislations frequently prohibit these additives. Thermal

preservation relies on the principle of heating the yogurt in order to damage and kill the yogurt microorganisms. Yet, food legislations demand that living cells of typical yogurt microorganisms must be contained in the final product so that ther-mal preservation can be applied only to a limited extent. Finally, aseptic produc-

tion and packaging has gained significant importance in the recent years. It aims at avoiding a contamination of the yogurt with germs and spores, particularly moulds and yeast. Hermetically closed production and packaging units are neces-sary that must be kept sterile by cleaning and disinfection cycles.

The two major quality problems of yogurt products are that it becomes too acid due to continued souring after delivery to the retailer, and that it becomes bitter. It is difficult to cool down the yogurt fast enough in order to avoid acidification (Walstra et al. 1999). Other processes taking place after manufacturing and pack-aging include the oxidation of fat in the presence of oxygen, changes in the color of fruit additives, or the hydration of protein constituents (Tamime and Robinson 1999). The resulting defects are off-flavors or too little characteristic flavors, vis-cosity issues or problems with the granularity of the coagulum (Walstra et al. 1999). Tamime and Robinson (1999) outline the most common defects and rea-sons.

5.2.2 Case Study 2: Shelf Life of Sausages

The most important deteriorating factors for sausages are microbial growth and oxidation of fats, which leads to rancidity (Kilcast and Subramaniam 2000). Savic (1985) emphasizes that the microflora of sausages is fundamentally different from that of carcass meat. Typical spoilage microorganisms found in meat processing industries include bacteria, yeasts and moulds. All of them are able to destroy the color, the flavor and the structure of the product. As the shelf life of scalded sau-sages depends on many factors and parameters, there is no universally applicable shelf life for scalded sausages; each manufacturer determines the shelf life of its products based on the individual characteristics of its plant and its distribution sys-tem. For example, while Savic (1985) names an average shelf life for emulsion-type sausages of 1-3 days at 10-15°C and 3-5 days at 0-4°C, Wirth et al. (1990) as

5.2 Shelf Life Characteristics of Case Study Products 123

well as Radespiel (1996) mention 14 days for cold cuts packed under vacuum and stored at below 5°C, and Andrae (1996) gives a range between 1 and 4 weeks if stored below 10°C.

In order to guarantee the labeled shelf life, several factors must be considered during production and distribution. First of all, as many of the product deteriorat-ing biochemical and chemical reactions are temperature-dependent (Tändler 1984a), temperature is one of the most important parameters to ensure and extend shelf life. At the processing stage, heating is applied to achieve a solid protein scaffold, to kill microorganisms and to obtain the sensory characteristics of the sausage (Stiebing 1984). The heating temperature should not be under 72°C, better 75°C, to ensure that over 99% of all microorganisms are eliminated (Wirth et al. 1990). Low temperatures limit the growth of the remaining bacteria at the ware-housing, packaging and distribution stages. Although the EU legislation prescribes an upper limit of 7°C, Tändler (1984a) recommends stocking cold cuts at 0-4°C at maximum, as each degree lower can extend shelf life by one day. Many problems occur at the retail level, where stocking temperatures increase frequently to up to 10°C.

Furthermore, reducing the water content in the final product is another lever to extend shelf life as the amount of water has a direct effect on microbial, chemical and enzymatic reactions. The water content available for microbial growth (not chemically bounded) is usually measured by the water activity or aw-value (Savic 1985). Decreasing this value (e.g. by drying or by adding additives) has a consid-erable extension effect on shelf life. A further decisive parameter is the initial

spoilage of the sausage with bacteria. Tändler (1984a) distinguishes the initial spoilage of the raw material and secondary spoilage from manual handling, slicing machines, or conveyor belts. Very high hygiene standards, particularly in the packaging area, help to keep the labeled shelf life. Moreover, the influence of oxy-gen is twofold: it leads to changes in color and to rancidity when reacting with fats. Vacuum packaging or modified atmosphere packaging reduces these factors significantly. Finally, intensive light can accelerate oxidation as it supplies energy for these processes. Therefore, lighting should be reduced as far as possible (Tändler 1984a).

5.2.3 Case Study 3: Shelf Life of Fresh Poultry

Microbial growth is certainly the decisive factor for the shelf life of fresh poultry; consequences are microbial spoilage and off-odors (Kilcast and Subramaniam 2000). The various processing steps such as cutting, deboning, mixing and pack-aging, increase the likelihood of microbial contamination (Richardson 1989). This contamination is affected by several parameters; Wirth et al. (1990) give an over-view of the most important influencing factors. In addition to storage temperatures and the initial contaminating microflora, the water activity, the substrate pH and hygiene factors play a decisive role as well (Bell 2001). Besides microbial spoil-age, Bell (2001) cites three more reasons for the perishability of fresh meat. The loss of moisture or water has a negative affect on both the sales weight and on the


appearance of the product. Changes in color are mainly caused by oxidation of the muscle pigment myoglobin. It gives a brown color to the product, which the con-sumers associate with staleness and a loss of sensory and nutritional properties. Finally, the oxidation of fats leads to rancidity as well as to off-odors and flavors.

Several processing steps have a high impact on the shelf life of the final prod-uct. Temperature requirements must particularly be respected at the chilling and storage stages. In order to reduce or to prevent the development of microorgan-isms, the carcasses must be chilled down from above 30°C to 0-4°C as quickly as possible after slaughtering (Varnam and Sutherland 1995). The air temperature should be kept around 0°C, but should not decrease –1°C in order to avoid freez-ing damage on the surface of the carcass (Cano-Munoz 1991). Also with respect to the storage of the carcasses and of the final packaged products, a temperature of –1°C to 2°C should be respected. As for processed meats, the legislation requires a temperature below 7°C (Radespiel 1996). In addition to the temperature require-ments, hygiene standards must be kept carefully as the number of spores on the surface also considerably influences the degree of spoilage of meat products. Fre-quent cleaning and disinfections of equipment and staff are necessary to guarantee shelf life and food safety for the consumer (Wirth et al. 1990). Numerous legisla-tions and recommendations set hygiene standards; for details it is referred to Sie-laff (1996).

Packaging is undoubtedly one of the most important steps with regard to shelf life of fresh poultry products. According to Richardson (1989), three main types of fresh poultry meat packaging can be distinguished. The most popular method is using a rigid or plastic tray over-wrapped by a clear film of material with a high oxygen and low water-vapor permeability. Although vacuum packaging has gained significant importance in the red meat industry, it remained relatively un-successful for fresh poultry due to the unpleasant appearance of the packaged meat (Varnam and Sutherland 1995). Finally, modified atmosphere or gas packaging, combined with oxygen absorbers, is more popular for cut-up, diced or minced meat (Richardson 1989).

As packaging has a major influence on shelf life, most authors differentiate their shelf life specifications by the packaging format. For entire unpacked poultry carcasses, Radespiel (1996) names a shelf life of 8-12 days if stored at –1 to 0°C and 6 days if stored at 2-4°C. Cano-Munoz (1991) mentions 7-10 days if stored at –1 to 0°C. Russell (2001) reports that the shelf life of fresh carcasses decreases dramatically with increasing temperatures, from 6-8 days at 4.4°C to 2-3 days at 10.6°C. If the carcass is packed under vacuum, the shelf life can be extended to 25 days, and even to 70 days if packed in a carbon dioxide environment. The shelf life of portioned poultry is much shorter. If only over-wrapped, shelf life is limited to 2-4 days; an extension to 4-8 days is possible if a modified atmosphere is ap-plied (Bell 2001). Radespiel (1996) limits this shelf life to 3 days for over-wrapped packages and to 5 days for modified atmosphere packages. Nonetheless, as the shelf life of fresh-portioned poultry products is the shortest of all considered case study industries. Consequently, the integration of shelf life into production planning of poultry processing has the highest priority.

5.3 Shelf Life in Fresh Food Supply Chain Management 125

5.3 Shelf Life in Fresh Food Supply Chain Management

5.3.1 Literature Review

In the literature, most production planning approaches assume an unlimited stor-age of intermediate and finished products. Many authors concede the necessity to integrate the shelf life of products in production planning and scheduling (e.g. Kallrath 2002; Günther and Neuhaus 2004); nonetheless shelf life has only been considered explicitly in very few models. Blömer (1999) gives an overview of 31 different approaches to batch scheduling. Although most authors respected unlim-ited or zero storage times, no author considered finite shelf life. Within the avail-able approaches that consider the shelf life or the deterioration of products in OR-related literature, two main avenues can be distinguished. On the one hand, inven-tory models for deteriorating items have been studied for decades. On the other hand, a limited number of researchers have integrated shelf life aspects into pro-duction planning and scheduling approaches. Major contributions of both streams are presented in this chapter.

A vast body of literature exists on inventory management for perishable prod-ucts or deteriorating items. Beside perishable food products, perishable inventory theory covers also the behavior of radioactive materials, photographic film, pre-scription drugs, or blood conserves. Nahmias (1982) and Raafat (1991) give com-prehensive literature overviews and analyze the proposed inventory models for perishables. Nahmias (1982) distinguishes two classifications of perishability: fixed lifetime, if the shelf life of products is known a priori and independent of other system parameters, and random lifetime which includes exponential decay or lifetime as a variable with a specific probability distribution. Within the fixed lifetime category, he further distinguishes deterministic and stochastic demand. Raafat (1991) defines decay or deterioration as “any process that prevents an item from being used for its intended original use” and names the examples of spoilage (e.g. foodstuff), physical depletion (e.g. evaporation of volatile liquids) and decay (e.g. radioactive substances). He differentiates models in which all products be-come obsolete at the same time (e.g. fashion or newspapers) and models in which the products deteriorate throughout the planning horizon. The latter can further be divided into products with a fixed lifetime and products with a continuous decay. While Nahmias focused on the fixed lifetime models, Raafat pays more attention to the continuously deteriorating models.

Many authors extended the basic inventory management models by focusing on specific aspects. The model of Wee (1995) describes a situation with an exponen-tially declining demand, allowing complete and partial backordering. Wee and Shum (1999) integrate product deterioration into the Wagner-Whitin approach, to the Silver-Meal algorithm and to the Least-Unit-Cost heuristic. Teng et al. (1999) developed a model for fluctuating demand, thereby generalizing the cases of in-creasing, decreasing or constant demand. Manna and Chaudhuri (2001) present two models with time-dependent demand and deterioration rates, with and without inventory shortages. Yang and Wee (2001) integrate the present value concept into


the inventory model for deteriorating items. Among the very recent publications, Goyal and Giri (2003) study an inventory model in which demand, production and deterioration rates of a product vary with the time. Rau et al. (2003) as well as Yang and Wee (2003) extend the inventory models to a multi-echelon SC by inte-grating suppliers, producers and buyers. Two contributions that analyze the influ-ence of marketing and pricing in perishables inventory management are to be mentioned. Goyal and Gunasekaran (1995) underline that marketing policies have a significant influence on the demand of a product. Therefore, they analyze the impact of marketing instruments (e.g. price per unit or advertisement frequency) on the demand of a perishable product. The idea of optimal pricing is further de-veloped by Abad (2003). He considers customers to be impatient so that in case of backlogging the remaining demand is a function of the waiting time. As the de-mand is sensitive to the sales price, pricing, replenishment planning and backlog-ging must be considered simultaneously.

The major drawback of all models dealing with perishable inventory theory is that production issues are almost completely neglected despite the fact that the shelf life of products is actually determined by their time of production. Produc-tion capacities, sequence-dependent set-up times, and production on multiple units or lines are not considered. Moreover, most models are restricted to one single item although there is a dependency of several products in production if many products are manufactured based on the same production batch, only differentiated by the packaging material.

With regard to the integration of shelf life into production planning and sched-uling approaches, most research deals with adding a shelf life constraint to the Economic Lot Scheduling Problem (ELSP), which is concerned with generating a cyclic schedule for several products, based on a single resource and constant de-mand rate (see for example Elmaghraby 1978; Cooke et al. 2004). Soman et al. (2004) provide a review of the major contributions. As one of the first authors to consider shelf life in this context, Silver (1989) argued that reducing the produc-tion rate is more effective than reducing the cycle time in order not to violate shelf life constraints. However, he assumed that reducing the production rate does not cause additional costs. Sarker and Babu (1993) completed the model of Silver by adding production time related costs and demonstrated that the choice between re-ducing the cycle time or the production rate depends on the shelf life of the prod-ucts, machine and product set-up times, and unit costs. Goyal (1994) and Viswanathan (1995) elaborate on the idea to produce a product more than once in a cycle. Silver (1995) as well as Viswanathan and Goyal (1997) aim at optimizing the cycle time and the production rate simultaneously. Viswanathan and Goyal (2000) allow backordering within the ELSP with shelf life considerations. Among the recent publications, Chowdhury and Sarker (2001), Viswanathan and Goyal (2002), and Sarker and Chowdhury (2002) work on the three options “adjusting the production rate”, “adjusting the cycle time” and “adjusting production rate and cycle time simultaneously” with regard to production scheduling and raw material ordering. Finally, Soman et al. (2004) argue that in case of high capacity utiliza-tion as it can be found in the food industry the production rate cannot be reduced because quality problems can occur if the production rate is changed. Therefore,

5.3 Shelf Life in Fresh Food Supply Chain Management 127

they use a constant production rate in their model formulation. Furthermore, they do not allow backordering, which reflects the competitive environment in FFSCs with retailers having strong bargaining power.

Almost all mentioned ELSP models include assumptions that are seldom pre-sent in an industrial environment. First of all, a constant demand rate is supposed, which is not very realistic for fresh food industries with high seasonality and in-tense promotional activities. For example, Manna and Chaudhuri (2001) underline that the demand for deteriorating products may be for example time-dependent, stock dependent and price-dependent. Moreover, most models consider only one single facility and do not account for sequence-dependent set-up times. Finally, the most important criticism is that ELSP models aim at generating a production cycle which is repeated in certain intervals and which must not exceed the shortest product shelf life. Therefore, product freshness is not considered in the objective function, only as a constraint that must be respected.

5.3.2 Role of Shelf Life in Fresh Food Supply Chains

The short shelf life of fresh food products is an important limiting factor for FFSCs. It has an impact on several factors:

Due to the fact that fresh products cannot be stored infinitely, the produc-

tion frequency of fresh products is relatively high compared to other, non-perishable products. In case the product exceeds the labeled shelf life or the minimum cus-tomer requirement on shelf life, it can only be sold with significant price discounts or even not at all. In the latter case, besides the not realized revenues the costs of wastage must be considered in addition. The consumer’s requirement for fresh products necessitates frequent de-liveries to the retail outlets (see Chapter 3.3.4). Out-of-Stock rates present a severe problem for manufacturers and retail-ers (see Chapter 3.3.2). Higher stock levels in the outlets could help to decrease these Out-of-Stock rates. However, when determining the target stock levels for fresh products, a risk component must be included in or-der to account for stock obsolescence and to avoid product wastage due to the short shelf life.

Consumers judge the freshness of a product based on its remaining shelf life if the packaging prohibits to see, touch or smell the product. The BBD is the only in-formation provided to assess the freshness. Therefore, the product with the longer remaining shelf life is considered to be the fresher product, because the consumer cannot assess the total shelf life of a product. If technological possibilities to ex-tend shelf life are exhausted (see Chapter 5.1.3), the manufacturer can only try to produce as closely to the demand as possible in order to guarantee a longer shelf life to the consumer. This shelf life extension has advantages for the manufacturer, the retailer and the consumer:


First of all, product wastage can be decreased in the retail outlet as the tight shelf life deadlines are loosened to some extent. In addition, as a longer shelf life in the outlet diminishes the risk of stock obsolescence, the stock level of the product in the outlet can be increased. This will help to decrease the Out-of-Stock rates (see Chapter 3.3.2). Furthermore, the ability to offer fresher products than its competitors constitutes a pivotal competitive advantage for the manufacturer. In the current manufacturing environment, products are increasingly inter-changeable if they are not branded. For this reason, the guarantee of fresher products is a major differentiating factor, which supports the gen-eration of new business. Finally, the most important effect for the manufacturer is that the con-sumer’s choice which product to buy is influenced in a positive way. As freshness is a decisive buying criterion (see Chapter 5.1.1), the consumer is likely to be inclined to buy the product with the longer shelf life, which is particularly true in case of a short product shelf life. This consumer’s decision is related to both the possibility to store the product at home for a longer time and the health-related impression of fresher products.

Therefore, shelf life should be considered on all relevant levels in supply chain planning. Delivering products that are as fresh as possible should become one of the objective of production planning. As many of the benefits of fresher products are at the retailers side (e.g. less wastage, lower out-of-stock rates), however, al-most all additional costs arise at the manufacturer’s side (higher sterilization and cleaning cost), both parties must agree on how to share costs and benefits. To solve this conflict, one possibility is to integrate a shelf-life dependent pricing component into the terms and conditions system. In addition, it must be consid-ered that the importance of fresh products differs between retailers. A typical dis-counter with a lean assortment, high product volumes and inventory turns is usu-ally less concerned about shelf life than a smaller, more traditional retailer with a broad product range, but low inventory turns. Therefore, an approach is required that is differentiated by customer and product.

5.4 Conclusion

As shown in Chapter 4, the consideration of shelf life in fresh food industries is important at all planning levels, from strategic to operational planning and from procurement to sales and distribution. The shelf life of fresh products such as yo-gurt, sausages or fresh poultry represents a major constraint in production plan-ning. In addition, the retailers’ requirements concerning the remaining shelf life of a product further limits the operational flexibility of the manufacturer.

The residual shelf life of a product is one of the most decisive buying criteria for consumers. Two main options exist to extend the product’s shelf life. On the one hand, technological shelf life extensions have been very successful in the past.

5.4 Conclusion 129

Nonetheless, as today almost all manufacturers have implemented these new tech-nologies, it is not a differentiating factor any more. On the other hand, a manufac-turer can aim at delivering his products as freshly as possible to the retailer, thus extending the shelf life for the retailer and the consumer. An effective production and SC planning support, however, must support this second option. Therefore, the integration of shelf life in contemporary APS systems is analyzed in the next chapter.

6 Shelf Life Integration in APS-Systems

6.1 Introduction

For the analysis of the shelf life coverage of APS systems, three systems have been chosen that cover different software segments. Two of the packages are lead-ing APS systems: The Advanced Planner and Optimizer of SAP and the Supply Chain Planning suite of PeopleSoft’s EnterpriseOne software package. The third analyzed system (the CSB-System) is basically an ERP system with some APS functions. However, as it has very high implementation numbers in fresh food in-dustries (particularly in the meat and the dairy segment), it has been considered in addition.

The assessment of the systems is principally based on interviews with product managers and sales representatives of the providers. In addition, relevant literature on the systems, technical notes, brochures, and Internet sources have also been taken into consideration. Due to the different structures of the systems, the as-sessment of each system does not always cover all aspects for all systems. Empha-sis is given on the functions that distinguish a system from the others in order to derive possible areas of improvement. For each system, a brief overview of struc-ture and the major components of each system are given. Then, the offered shelf life support of each of the modules of an individual system is described. For the future development of the systems, several improvement potentials are highlighted in Chapter 5.4.

6.2 SAP APO

6.2.1 System Overview

SAP is the leading provider of enterprise software in the world, broadly recog-nized by its ERP system R/3. SAP is a relatively young player in the APS arena; the first release of its APS system SAP APO was in January 1999 (Davies et al. 2002). In the recent years, SAP enlarged and deepened the functions of the system so that today it is in a state to compete head-to-head with the other leading APS systems (Cap Gemini Ernst & Young 2002a). One of the main sales arguments for SAP APO is the easy integration with the SAP R/3 system. Even though SAP APO can also be integrated with any other ERP system or even run stand alone

132 6 Shelf Life Integration in APS-Systems

(Davies et al. 2002), almost all clients use the system as a layer on top of the SAP R/3.

Due to the widespread distribution of SAP R/3, SAP is expected to gain many new customers for its APO system and to increase its share in the APS market in the future (Cap Gemini Ernst & Young 2002a). The SAP APO system, together with the SAP R/3, constitutes the backbone of the so-called “mySAP Supply Chain Management” suite. This software suite aims at covering all aspects of managing the entire SC and integrates for example with the “mySAP Customer Relationship Management” suite, the “mySAP Supplier Relationship Manage-ment” suite, and the “mySAP Product Lifecycle Management” suite (Davies et al. 2002).

SAP APO consists of four main system components: the liveCache, a database, the solvers, and the applications (see Fig. 6.1). The liveCache technology is fun-damental to solve complex planning problems in real time. It serves to keep the data memory-resistent in order to avoid time-consuming reading and writing from and to a database. Within SAP APO, the liveCache orchestrates all data and proc-esses in the main memory. For the resolution of the planning problems, SAP APO applies a bundle of optimization algorithms and heuristics, which have been either developed in-house or delivered by ILOG and which are stored in the solver com-ponent. The solvers receive the data of the underlying planning problem from the liveCache and write the results of the optimization runs back to the liveCache. All data is finally stored in a relational database. SAP APO supports all common Da-tabase Management Systems such as SAP DB, IBM DB/2, MS SQL Server, or Oracle (Bartsch and Bickenbach 2002).

Fig. 6.1 SAP APO system architecture (based on Bartsch and Bickenbach 2002)

6.2 SAP APO 133

Furthermore, the Supply Chain Cockpit is the central entry point for the man-agement of the SC (Bartsch and Bickenbach 2002). It provides visualization capa-bilities for planning and controlling the network and contains a network design component for the modeling of the network, a navigation component and a moni-tor that sends out problem-related messages in case of pre-defined exceptions. A drill-down function allows investigating on the reasons (Knolmayer et al. 2002). The applications part contains the modules used to perform the actual planning tasks:

The DP module offers a relatively broad range of conventional statistical methods. Moreover, it provides several additional features covering the planning of promotions, new product introductions and phasing-out of products, product lifecycles, or collaboration support (Davies et al. 2002; Knolmayer et al. 2002). The SNP module integrates the areas of purchasing, production, distribu-tion and transport and enables the simulation and implementation of tac-tical planning decisions. This mid-term plan determines the volumes to be transported, produced or sourced per time period. Three different methods are available to generate this plan. On the one hand, heuristic approaches are rule-based methods that can handle complex planning problems but that do optimize an objective function. On the other hand, optimization approaches are available that are based on LP and MILP models, which use ILOG CPLEX for the solution process. Finally, the Capable-to-Match procedure aligns prioritized demand with available stocks. For the implementation of the generated plans, SNP additionally contains a Deployment function, which determines when and how stocks should be delivered to distribution centers or customers, and a Transport Load Builder, which aims at optimizing transport loads (Bartsch and Bickenbach 2002). On a short-term planning level, the Production Planning / Detailed Scheduling (PP/DS) module enables the planner to plan multi-site pro-duction while simultaneously taking into account product availability, sequence constraints and capacity (Davies et al. 2002). PP/DS applies constraint programming techniques and genetic algorithms to solve the planning problems (Knolmayer et al. 2002). In contrast to the SNP mod-ule that uses a time bucket approach, the PP/DS modules makes use of a continuous representation of time. Transport Planning / Vehicle Scheduling (TP/VS) is the short-term plan-ning application for transportation-related issues. It allows the simultane-ous consideration of constraints for in- and outbound transportation and for replenishment. The main functions are load consolidation and vehicle scheduling, route determination and carrier selection (Knolmayer et al. 2002). Optimal vehicle loadings and routings are derived using ILOG components, completed by genetic algorithms and heuristics (Meyr et al. 2002a).


The Global ATP is a rule-based, multi-level component and capacity check. In addition, it includes advanced techniques such as the determi-nation of alternative products or locations for production and purchasing. Moreover, the Global ATP module assigns resources to specific orders, regions or customers (Meyr et al. 2002a). Finally, the Supply Chain Network Design is an application that supports the strategic realignment of the supply network and helps to locate plants and facilities. The module is based on the demand data of the DP module and provides four different planning methods in order to, for example, assign distribution centers or locations to customers (Meyr et al. 2002a). Starting from version 3.0 it is only delivered based on specific customer requirements (information provided by Dr. G. Heisig, SAP AG, tele-phone interview on 26.05.2004).

Interfaces between SAP APO and other systems constitute a critical element for the success of the implementation. A tight integration with the company’s ERP systems is required as the underlying data for an APS system is usually stored in one or several ERP systems (see Chapter 2.1.3). Furthermore, in order to address all issues in SCM, other systems must be integrated such as CRM or SRM systems (see Fig. 6.1). Finally, the APS system must also address collaboration aspects with customers and suppliers because inter-company planning aspects get increas-ingly important. However, the integration of these different systems is a challeng-ing task if the IT landscape is very heterogeneous. Therefore, SAP APO offers a variety of interfaces within the Application Link Enabling part that allow to read and to write back data to other systems:

The most important interface connects SAP APO to SAP R/3 system. This plug-in is called Core-Interface (Bartsch and Bickenbach 2002). Business Application Programming Interfaces, which are open and pro-vider-independent interfaces to manipulate business objects in distributed systems, can connect other ERP systems (Bartsch and Bickenbach 2002). A connection via the Internet is supported by EDI and XML interfaces

(Knolmayer et al. 2002).

6.2.2 Shelf Life Integration

Within SAP APO, shelf life is considered in two of the mentioned modules, the SNP and the PP/DS module (if not indicated otherwise, this chapter is based on a telephone interview with Dr. G. Heisig, SAP AG, 26.05.2004). Other modules such as DP, Global ATP, or TP/VS do not take shelf life issues into account. Fur-thermore, the approach and scope of shelf life integration differs significantly be-tween the SNP and the PP/DS module. Therefore, both ways of shelf life integra-tion are described in the following paragraphs. Principally, the SAP APO system can administrate four shelf life related dates (SAP AG 2003c):

6.2 SAP APO 135

The shelf life of a product or batch refers to the time period after which the product or batch expires. It is maintained in the master data of the product. APO blocks products during maturation time. In this period they cannot serve to fill customer orders. The maturation time is part of the total shelf life. Hence, the total shelf life is composed of the maturation time and the availability time. The maturation time is also contained in the product master data. The required minimum shelf life stands for the minimum residual shelf life that a product has to have in order to be considered for a customer order. The required maximum shelf life is defined analogously. Both the re-quired minimum and the required maximum shelf life can either be main-tained in the product master data or refer to a specific customer order if a “characteristic dependent planning” approach is chosen (SAP AG 2003b). In the latter case, different customer requirements for the same products are possible.

This shelf life related data can be stored either directly in the SAP APO system or it can be imported from a third system, e.g. from an ERP system. Within SAP R/3 for example, the following shelf life related information can be maintained:

In the product master data, the total shelf life is stored. Batch-related information can contain the production, availability and expiry dates. The required minimum and maximum shelf life can be defined as charac-teristics of a customer order.

When planning with shelf life, it must be noticed that the shelf life settings are not location dependent, there are always valid for all locations (Dickersbach 2003). In case of location-specific shelf lives (e.g. caused by different manufactur-ing technologies at different sites), a new product has to be introduced for each shelf life. Furthermore, the shelf life cannot be propagated, it is not possible to pass the shelf life of raw materials or of intermediate products on to the shelf life of the corresponding final products. In addition, there is no possibility to assign two different shelf lives to the same production batch (e.g. in case of overnight production).

The PP/DS module can take all four shelf life related dates into account. The shelf life information can be respected within the planning process, but it may also be neglected. If the planner wants the shelf life to be considered, he must select the “Planning with shelf life” indicator. In that case, the shelf life information is used in pegging. The shelf life can either be determined on a day basis (e.g. pro-duction day + x shelf life days) or even on a point in time basis (e.g. 10.10.2004, 10:30, until 25.10.2004, 10:30). As depicted in Fig. 6.2, the shelf life and the ma-turity time of a supply element are checked with the required minimum and maximum shelf life of a demand element. Three additional constraints are neces-sary to consider shelf life information in pegging (Dickersbach 2003):


Requirement Date Availability Date + Maturity (6.1)

Requirement Date + Required Max. Shelf Life Availability Date + Shelf Life

(6.2)

Requirement Date + Required Min. Shelf Life Availability Date + Shelf Life

(6.3)

Fig. 6.2 Shelf life parameters in SAP APO (based on Dickersbach 2003)

Therefore, in Fig. 6.2 only in case 3 all constraints are met. In case 1, Equations 6.1 and 6.2 are violated as the requirement date is prior to the end of the matura-tion phase and the shelf life of the product exceeds the required maximum shelf life. In case 2, Equation 6.2 is violated, again due to the required maximum shelf life. In case 4, Equations 6.2 and 6.3 are violated as both the required minimum and maximum shelf life constraints are violated. In these three latter cases, the demand element cannot be pegged as shelf life, maturation times and the required shelf lives are considered as hard constraints in PP/DS. Therefore, taking into ac-count the shelf life information can complicate the search for a solution of the planning problem. Starting from version 3.1 of APO, SAP provides two heuristics for PP/DS which support shelf life (SAP_PP_SL001 and SAP_PP_SL002; Dick-ersbach 2003).

6.3 PeopleSoft EnterpriseOne 137

Regarding alerts, SAP APO provides four different types. The most important one indicates the unpegged receipts with the corresponding expiration dates. In addition, three alerts are made available for fixed pegging edges violating shelf life constraints. These are “shelf life too short”, “shelf life too long”, and “fall be-low maturation time” (SAP AG 2003b). All other alerts (e.g. based on pre-defined alert thresholds) can be generated in the underlying ERP system.

With respect to the SNP module, the relevant planning process does not cover shelf life. However, regarding the different algorithms of the SNP module, some approaches are available in the SNP optimizer (SAP AG 2003d). Neither the Ca-pable-to-Match procedure nor the Deployment function nor the Transport Load Builder take shelf life information into consideration (SAP AG 2003b). Within the SNP optimizer, the only information taken into account is the total shelf life of a product; maturation times as well as required minimum and maximum shelf lives are not considered. There are two options in the SNP optimizer to integrate shelf life: the product can either be disposed after the expiry date, or it can be allowed to further use the product. In the latter case, penalty costs can be specified which can be either the maintained procurement cost of the product or cost elements that are not dependent on the product (SAP AG 2002). As batches cannot be used in SNP, the shelf life of stocks transferred from an ERP system to SAP APO is not available (SAP AG 2003d).

It is important to notice that the shelf life determined in the product master is not passed on to other locations during transport. This leads to the fact that the de-termined shelf life period starts again if the product is transported from one loca-tion to another. In case of cycle-free networks (no transport of a product from one location back to the start location), this limitation might be managed by consider-ing only the last stage in the distribution chain (e.g. the DC) with the correct shelf life and deduct the average inventory times of the other SC stages from the total shelf life. However, even then the consideration of shelf life will be inaccurate. Moreover, if a product can be transported back to the start location, the SNP opti-mizer will likely use this option to virtually extend the shelf life. Therefore, this type of inventory movements must be monitored closely (SAP AG 2003a).

6.3 PeopleSoft EnterpriseOne


With a turnover of US$ 2.3 bn. and 12,000 employees, PeopleSoft is the second largest provider of enterprise application software (PeopleSoft Inc. 2004a). Throughout the recent years, PeopleSoft acquired several smaller companies that were specialized in Advanced Planning and Scheduling (APS) software (e.g. Red Pepper in October 1996 or Advance Planning Solutions in May 2000). However, by acquiring J.D. Edwards in August 2000, PeopleSoft became one of the leading providers of APS systems worldwide (PeopleSoft Inc. 2004b). Although J.D. Ed-wards traditionally offered ERP software, it had strengthened its APS capabilities


by the acquisition of Numetrix in 1999, a specialized APS provider with more than 20 years of experience (Meyr et al. 2002a). The APS capabilities of People-Soft are part of the EnterpriseOne software package, which are modular pre-integrated industry-specific business applications based on a pure Internet archi-tecture. The APS functions belong to the Supply Chain Management suite; other suites are for example CRM, SRM, Financial Management or Human Capital Management (PeopleSoft Inc. 2004c). The Supply Chain Management suite in-cludes four main product solutions: Customer Order Management, Logistics, Manufacturing, and Supply Chain Planning, the last covering the following APS modules (PeopleSoft Inc. 2004e):

The Strategic Network Optimization module allows modeling the entire supply chain and supports decision making on a strategic level. The op-timization methods applied include LP and MILP models as well as heu-ristics (Davies et al. 2002) in order to find optimal configurations and flows. Meyr et al. (2002a) underline the powerful visualization and mod-eling capabilities. A special heuristic is provided for capital asset deci-sions, which proposes not only the best assets to open or close, but also the transition plan (Davies et al. 2002). Two modules support the demand planning process: the Demand Fore-casting and the Demand Consensus module. Demand Forecasting allows performing statistical forecasting based on historical sales data and causal factors, relying on multiple statistical methods (e.g. time series methods, causal methods). Statistical planning of promotions and events is inte-grated as well (PeopleSoft Inc. 2004h). In addition, the Demand Consensus module addresses collaborative ac-tivities in the forecasting process. It supports internet-based collaboration between internal users, partners, and customers in order to reconcile mul-tiple forecasts into one single forecast valid for the entire corporation (PeopleSoft Inc. 2004g). The Production and Distribution Planning module supports mid-term and also short-term planning decisions. On the mid-term planning level, the module generates procurement, production and distribution plans for the supply network. In addition, the sub-module Vehicle Loading covers short-term transportation planning decisions (Davies et al. 2002). On the short-term production planning level, two distinct modules are available. On the one hand, Production Scheduling – Process focuses on process industries such as food and beverages, consumer goods, chemi-cals, and pharmaceuticals (PeopleSoft Inc. 2004k). It aims at creating feasible schedules while respecting constraints such as machine and labor capacities, product sequencing, or changeovers. The module is especially suited for parallel continuous production lines with up to two stages of production in a make-and-pack environment. In contrast to many other scheduling applications, the module considers cost-based objectives in addition to the usual time-based objectives (Meyr et al. 2002a).

6.3 PeopleSoft EnterpriseOne 139

On the other hand, Production Scheduling – Discrete is particularly de-signed to support scheduling decisions in production environments with complex products, deep bill-of-material structures and floating bottle-necks (PeopleSoft Inc. 2004j). Finally, the Order Promising module includes three methods to deter-mine the order delivery date. Besides the common Available-to-Promise (ATP) and Capable-to-Promise (CTP) procedures, the module offers in addition a Profitable-to-Promise (PTP) function that indicates the best way to satisfy customer demand while maintaining expected margin (PeopleSoft Inc. 2004i). Two operating modes can be distinguished: While the Autopromise function automatically generates promise date proposals, the Scenario Management function allows the user to evaluate different promising alternatives (Davies et al. 2002).

Starting with version 8.9, EnterpriseOne applies a pure Internet technology, thus avoiding the installation of client software. The communication with the Internet application server is ensured by standard internet technologies such as HTTP, HTML, or XML (PeopleSoft Inc. 2004d). Horizontal and vertical integra-tion is done by the Supply Chain Business Modeler, which relies on XML-based input and output as well as an Object Oriented database and which offers func-tionalities for data aggregation into higher planning levels. It integrates APS with PeopleSoft’s EnterpriseOne and has the capabilities to integrate with every system that can provide the needed data (and with different systems in parallel; informa-tion provided by Dr. H.-H. Schulz, PeopleSoft, 28.09.2004). Other databases or ERP systems can be integrated by the “eXtended Business Processes” framework that executes the business logic and data transformations (PeopleSoft Inc. 2004f).


The support of shelf life functions within the solution of PeopleSoft differs sig-nificantly between the Production and Distribution Planning and the Order Prom-ising module on the one hand and the Production Scheduling - Process module on the other hand (if not indicated otherwise, this chapter is based on a questionnaire and a telephone interview with Dr. H.-H. Schulz, PeopleSoft, 15.07.2004). Other modules do not offer a shelf life support. The Production and Distribution Plan-ning module considers two shelf life related parameters, the shelf life of a product and the minimum customer requirement on shelf life. While the shelf life of the products is available per batch, the customer requirements on shelf life are origi-nally alert parameters which can however be used to cover the customer require-ment functionality. The customer requirement is product-dependent, not customer-dependent. As the shelf life is stored in the product master data, only one shelf life is available per product. The shelf life data in the Production and Distribution Planning module is measured in days since a day constitutes the lowest possible planning granularity. With regard to optimization, the shelf life and the minimum customer requirement on shelf life are modeled as soft constraints. In case of a


violation of the constraints, the procurement costs of a new product are applied as penalties. The penalty costs cannot be individualized (e.g. application of other costs than the procurement costs for a new product or penalties depending on the number of days the time constraint is exceeded by).

The Order Promising module applies the same data model as the Production and Distribution Planning module. Therefore, the shelf life of a product and the minimum customer requirement on shelf life are respected in the ATP, CTP and PTP procedures. Again, the customer requirement on shelf life is product-dependent, not customer-dependent. A prioritization of customer orders with re-gard to shelf life can be performed by manually assigning orders and batches.

In view of the Production Scheduling modules, shelf life functions are only provided in the Production Scheduling – Process module; the Production Schedul-ing for Discrete Industries module incorporates no shelf life functions. The sup-ported shelf life related parameters include the shelf life of a product, the mini-mum customer requirement on shelf life as well as minimum and maximum maturation time. In contrast to the Production and Distribution Planning module, the customer requirements on shelf life can also be customer-dependent, not only product-dependent.

Furthermore, different shelf lives for a single product depending on the produc-tion site are possible as the module can be implemented on a site-specific basis. In addition, the shelf life of raw and intermediate products can be considered as well. The shelf life of products can be represented in both ways, in days or exact. The assignment of two shelf lives for the same production batch (e.g. in case of a pro-duction of a batch over several days) is not possible, in this case the batch must be split. Moreover, the shelf life cannot be propagated (e.g. from intermediate prod-uct to final product). With respect to the optimization, all constraints (e.g. shelf life, customer requirements and maturation times) are soft constraints; a bonus for the delivery of fresher products is not provided. Several alerts are available, e.g. for the violation of shelf lives constraints, of customer requirements on shelf life or of maturation time constraints. The alerts can be distinguished into yellow alerts that are activated if a deadline (e.g. the expiry date) is approached, and red alerts if the deadline is exceeded.

6.4 CSB-System


In contrast to the presented APS systems, the CSB system is not a pure APS sys-tem; it is rather an integrated solution for specific industries. It has s strong focus on batch-oriented process industries and is particularly suited for food and bever-ages, for chemicals and paints, for pharmaceuticals and cosmetics and for retail and logistics. The system has been considered in addition to the pure APS systems as it has a relatively high number of implementations in fresh food industries (es-pecially in the meat, the dairy and the bakery industry; CSB-System AG 2004a).

6.4 CSB-System 141

Fig. 6.3 Structure of the CSB system (based on CSB-System AG 2004b)

The structure of the system is given in Fig. 6.3. On the one hand, the CSB con-tains modules that are usually part of an ERP system. Examples are the accounting and finance module, the management and controlling module, the human re-sources module or the coverage of the value chain from purchasing to sales and distribution (see Chapter 2.1.1). However, on the other hand the system incorpo-rates a variety of functions that are very specific for batch-oriented process indus-tries and especially for the food industry. Some examples are (CSB-System AG 2004a):

Integrated quality management system in compliance with the ISO 9000 or HACCP standards (see Chapter 3.3.5), Batch traceability (see Chapter 3.3.5), Integrated laboratory information and management system, A hazardous materials management component, A livestock management module, A producer clearance system for purchasing agricultural products, or A nutritional value calculation module.

One of the major differentiating factors of the CSB system is the availability of industry-specific optimization algorithms; three different types are offered. First, a “cutting process and joint package acquisition” component helps to choose the right cutting variation for the cutting of joint product packages into individual products.


Fig. 6.4 Advanced production scheduling process in the CSB system (figure provided by S. Schiller, CSB System AG)

Second, the “standardization procedures and optimization” component can be applied in order to divide raw materials into standardized groups. Third, the “rec-ipe organization and optimization” component allows to generate cost-optimal recipes while complying with food regulations and sensory requirements (Schimitzek 2004). However, the system does not support other optimization algo-rithms as they are incorporated in the two systems presented previously (e.g. algo-rithms for Supply Network Planning or Scheduling). The reason for this limita-tion is “that only those models should be supported that include transparent approaches (or objectives) and, at the same time, allow the user to intervene in

process details” (Schimitzek 2004). The “APS” module covers the production planning processes (see Fig. 6.4). In

the notation of the CSB system, APS stands for Advanced Production Scheduling and resembles the MRP II process (see Chapter 2.1.1). The production planning process of the CSB system is given in detail in Fig. 6.4. At the beginning of the

6.5 Summary and Conclusion 143

process, sales forecasts can be generated by means of forecasting models. A pe-riod overview is generated by checking the long-term planning data against four types of capacities (labor, machines, materials and financial). As a result, “planned production orders” are generated by the forecasting system. For the establishment of final production orders, the forecasted data is updated with current sales order data. Their feasibility is again checked against capacity and material requirements. Occurring bottlenecks can be resolved by adjusting the intensity, the time, or the quantity (Schimitzek 2004).


The CSB system supports more shelf life related parameters than the APS systems presented previously (if not indicated otherwise, this chapter is based on a per-sonal interview with S. Schiller, CSB-System AG, 05.07.2004):

the shelf life of product (Best-Before-Date, “Mindesthaltbarkeitsdatum”), the “Use-by-Date” (after which the product perishes), an alert date, the customer requirement on shelf life, and a maturation time.

In addition, the system distinguishes a “blocking time” which is the time be-tween the testing of a product and the result of the test) and a “quarantine time” which stands for the time between a negative result of testing to the resolution of conflict. In both cases, the products cannot be used to satisfy customer orders. Dif-ferent shelf lives (e.g. in case of different sites of different raw material batches) and different minimum customer requirements for the same product can be inte-grated via product variants. Furthermore, the shelf lives of raw and intermediate products are considered by the system. The system allows also propagating the shelf life, e.g. from intermediate products to the final product. As for the APS sys-tems presented before, both a daily and an exact representation of shelf life are en-abled by the CSB system. In addition, several alerts are supported (e.g. violating shelf life constraints, minimum customer requirement, or maturation time con-straints). The assignment of two shelf lives for the same batch is not possible. In that case, the batch must be split. As the system offers optimization algorithms only for very specific problems (see Chapter 6.4.1), shelf life is not part of the op-timization. For this reason, no penalty costs are applied for exceeding shelf life and no bonus is applied for delivering fresher products.

6.5 Summary and Conclusion

In summary, all analyzed systems cover some basic shelf life functions. Concern-ing the available data fields, the total product shelf life, a minimum maturation time and a minimum customer requirement on shelf life is usually supported. The


shelf life is not only considered for final products, but also for raw materials and intermediate products. With regard to the PS modules, both a daily and an exact representation of shelf life are normally available; however, most manufacturers use the daily representation of shelf life. For the optimization, shelf life constraints are frequently modeled as soft constraints in order to increase the solvability of the models. The procurement costs of a new product at a specific location are often applied as penalty costs. Finally, many different alerts are usually supported (e.g. exceeding of shelf life, of the minimum customer requirements, or of the matura-tion times). However, several shelf life functions that are relatively important for fresh foods are currently only available in selected systems or not available at all. Hence, the APS providers should address the following improvement potential in order to increase the dissemination of their software in fresh food industries:

Currently, the providers focus their shelf life functions particularly on the PP / PS modules. However, shelf life is not only an issue for short-term production planning, but should also be integrated into other planning tasks. Shelf life is very important for SNP and ATP and it can become important for SND and TP. For example, SAP-APO provides no shelf life support for the ATP module and the CTM, Deployment, or TLB pro-cedures of the SNP module. In addition to the consideration of shelf life functions within other mod-ules, the providers should aim at completing the supported parameters.The most important are the total shelf life, a minimum and a maximum maturation time, a minimum and a maximum customer requirement on shelf life, a use-by-date, as well as blocking and quarantine times. The shelf life of a product is the result of many different processing steps (see Chapter 5.1.1), which can differ between for example the sites of a company or even within a specific site. Hence, it may become necessary to attach different shelf lives to the same product, depending for instance on the production location, the used raw materials or the applied process-ing steps. In the considered systems, the shelf life of a product is either fixed globally for all sites or locally for one site. A differentiation by raw material or processing step is not possible. In many fresh food industries, the shelf life of a product depends on the shelf life of the raw materials (e.g. fruits and vegetables) or of the inter-mediate products (e.g. sausages). Therefore, the propagation of the shelf life of these products and material on to the final products is a prerequi-site for these industries, which should be supported by the software. Be-sides the simple one-to-one propagation of the shelf life, it can also be-come necessary to propagate the shelf life by means of a formula (for example shelf life of final product = shelf life of intermediate product + 10 days due to modified atmosphere packaging). In some systems, the determined shelf life period starts again if the prod-uct is transported from one location to another. In case of cyclic net-works, the optimizer will likely use this option to even virtually extend

6.5 Summary and Conclusion 145

the shelf life of the products. In order to avoid these problems, the shelf life of a product should also be propagated during transport.No system supports currently the assignment of two different shelf lives

to the same production batch. This can be the case, for example, in the dairy industry when high volumes of the same product are produced based on several batches of intermediate products or over several days (e.g. yogurt for the discount channel or private labels). The batch must always be split; yet if sequence-dependent set-up times are not consid-ered, the set-up times of the second batch must be corrected manually. As a longer shelf life of products is more important for some retail cus-tomers than for others, the system should enable an automatic prioritiza-

tion of the customer orders with regard to shelf life. For example, a longer shelf life of the products is more important for a traditional grocer with a complex assortment and low turn rates than for a discounter with very high turn rates. Currently, this prioritization can only be performed manually, if at all. With regard to the optimization, shelf life constraints are usually modeled as soft constraints, especially in the SNP module. The penalty costs ap-plied if the constraint is violated are generally the procurement cost of the product. However, as the products have a value even after having ex-ceeded the minimum customer requirement, it is desirable to assign indi-vidual and also time-dependent penalty costs.Similar to the penalties for exceeding the shelf life constraints, the system should reward the delivery of fresher products, as fresher products are of a higher value for the retail customers regarding wastage and out-of-stock rates (see Chapter 5.3.2). Currently, no system can offer this function.

As shelf life is one of the most important factors to consider in fresh food in-dustries, the mentioned deficits in the coverage of shelf life functions are one ma-jor reason for the relatively low implementation numbers in these industries. The CSB system shows that concentration on the requirements of specific industries can result in a high number of implementations. Therefore, in the following three chapters it is shown how shelf life can be integrated into production planning for fresh foods, based on the examples of yogurt, sausage and poultry production. Since the shelf life of the products is determined to a large extent on the short-term planning level, the case studies focus on this planning level with a planning horizon of one week. The shelf life data required for the support the proposed models (e.g. the total product shelf life, a minimum maturation time and a mini-mum customer requirement on shelf life) is usually incorporated into the PP and PS modules of the systems. The models particularly address the deficits related to assigning two shelf lives to one batch (see Chapter 7), the prioritization of cus-tomer demand and the reward the delivery of fresher products, since currently no system offers a support for these issues.

7 Shelf Life Integration in Yogurt Production

7.1 Problem Demarcation and Modeling Approach

This first case study aims at integrating shelf life into the short-term planning of yogurt production (for the yogurt production models, see also Seiler 2004). For a detailed description of the production processes, it is referred to Chapter 3.5. Within this case study, particular emphasis is given on the filling and packaging stage (see Fig. 7.1). The production system in the case study is composed of four packaging lines (three different line types), on which 30 different products are produced. A product can only be produced on a specific line type. The supply processes for raw milk are neglected since the incoming amount of raw milk is relatively fixed over a certain period of time due to the longer-term contracts with the farmers. Although there is a possibility to purchase raw milk on the spot mar-ket, most dairy factories generally face a raw milk surplus which is usually used for the production of milk with a very long shelf life (UHT-milk). However, this type of milk yields only very low margins. Therefore, all adjustments of the raw milk supply should be performed on a tactical planning level.

Fig. 7.1 Focus of yogurt production model

148 7 Shelf Life Integration in Yogurt Production

Furthermore, the delivery of products is not regarded within the models be-cause distribution logistics is often a task performed by retail organizations (N.N. 2003a; Biehl 2004). A comparison of the processes of fermentation and packaging shows that the production facilities required to add fruits and to wrap and seal the final products are far more capital intensive than those needed for fermentation purposes. In addition, manpower requirements of the packaging facilities are higher. Due to the use of multipurpose tanks for the fermentation (which can also be used for example for other dairy products), the fermentation processes are only considered by a capacity restriction and by imposing minimum batch sizes for the packaging lines, thus ensuring a minimum filling level of the fermentation tanks.

Although many parameters in fresh food production show some variations (e.g. processing times and yields, demand, equipment reliability), the proposed models are based on a deterministic modeling approach and do not include stochastic fea-tures, which is due to several reasons: First, as the deterministic models are al-ready relatively complex, especially with respect to the number of binary vari-ables, stochastic models would drastically increase the model complexity and lead to very long computational times. Second, on a weekly planning level many pa-rameters can relatively well be represented by long-term averages (e.g. processing times and yields) which are then corrected with the actual weights or times after the processing. If major deviations occur, the production system generally offers a sufficient degree of flexibility to cover these deviations. Finally, the results of sto-chastic modeling are very difficult to analyze for planners in an industrial envi-ronment which usually did not receive extensive education and training in OR methods.

The scheduling horizon of the planning problem is part of the short-term op-erational planning. It is aimed at solving problems concerning lot sizing and se-quencing decisions and to achieve a near-optimum planning solution as a recom-mendation for the facility according to the chosen target criteria (Meyr 1999). The time horizon of short term planning for yogurt manufacturing is usually one week (Nakhla 1995). In the example discussed here, the planning horizon is divided into macro-periods as indicated in Fig. 7.2. The regular production time during the week is from Monday 0:00 (day 6) until Friday 24:00 (day 10). If necessary, the production time can be extended from Sunday (earliest at 0:00, day 5) till Saturday (latest at 24:00, day 11).

Fig. 7.2 Planning horizon for yogurt production

7.1 Problem Demarcation and Modeling Approach 149

The planning of the packaging lines is always performed for one entire week taking into consideration the stored products from the previous week. All accumu-lated demand has to be met from Monday 0:00 (day 6) to Tuesday 24:00 (day 14) of the following week. The demand data is based on already confirmed customer orders as well as on forecasts. Due to the various types of packaging for different retailers, it is relatively simple to assign the products to their customers.

Batch production in the chemical industry can be taken as a reference for the production of yogurt as both must consider numerous variants, which are based on few product types or recipes. The variants are caused by the different types of packaging as well as by different tastes that are obtained by adding fruit prepara-tions and other ingredients at the filling and packaging stage. In the literature, a production environment that is characterized by a single production stage and a subsequent packaging stage is named “make and pack production” (e.g. Méndez and Cerdá 2002; Günther and Neuhaus 2004). Major issues of operational produc-tion planning in this environment are lot sizing and scheduling, which can be per-formed in one single or two separate planning steps. As lot sizing usually aims at balancing set-up costs on the one hand and inventory holding costs on the other hand, the set-up costs of each single set-up operation must be known in advance. However, as set-up times and costs in yogurt production are sequence-dependent,their exact values can only be determined after the sequencing of the orders. On the other hand, sequencing is dependent on lot sizing so that both tasks must be performed simultaneously (Sikora 1996; Stadtler 2002).

The available approaches can be classified according to their representation of time: Discrete and continuous time representations can be distinguished. For a dis-crete representation of time, the planning horizon is divided into a certain number of periods. These usually have the same length; all material inflows take place at the beginning of a period, all outflows at the end. Examples of problems and re-lated modeling approaches with discrete time representation are the Capacitated Lot size Problem (Günther 1987), the Discrete Lot-sizing and Scheduling Problem (Fleischmann 1990), the Continuous Set-up and Lot-sizing Problem (Karmarkar and Schrage 1985), the Proportional Lot-sizing and Scheduling Problem (Haase 1994), and the Capacitated Lot-sizing Problem with Sequence-dependent Setup Costs (Haase 1996). Fleischmann and Meyr (1997) integrate all mentioned models within the General Lot-sizing and Scheduling Problem. All models have in com-mon that setup times can only be considered if they do not exceed the length of a period. However, Koçlar and Süral (2005) show that, through a simple modifica-tion of the General Lot-sizing and Scheduling Problem, set-up times exceeding the length of a period can also be incorporated.

Nevertheless, choosing the length of a period becomes a crucial aspect of mod-eling. Especially because a high number of relatively small periods is required for an exact representation of the production, the number of periods can considerably increase the size of the model (Méndez and Cerdá 2002; Günther and Neuhaus 2004). Stadtler (2002) emphasizes that particularly sequence-dependent setup times cannot be represented properly within a model that uses large time periods. In addition, the approximation or “rounding” of processing and setup times to fit them into the fixed periods can lead to overproduction, idle time, or infeasibility


(Ierapetritou and Floudas 1998; Günther and Neuhaus 2004). Due to the high complexity of these models, they are suitable to practical purposes only to a lim-ited extent (Meyr 1999; Burkard et al. 2002). The application of time periods of different length can be useful to avoid the stated problems. Burkard et al. (2002) introduce the notion of the ”Event-Driven Model”, in which only such points in time are considered, at which a process is allowed to start. In this way, the model is more easily solvable than models using a uniform time grid.

Alternatively, it is possible to use a continuous representation of time, which al-lows precise planning of the beginnings and ends of events or processes on a con-tinuous time line. In particular, infeasibilities due to “fitting” setup times into a discrete time grid can be avoided (Ierapetritou and Floudas 1998; Günther and Neuhaus 2004). Sahinidis and Grossmann (1991) propose a Mixed Integer NonLinear Programming model that uses a continuous representation of time and that considers explicitly sequence-dependent setup times. They apply a position-based model and assume constant demand patterns to generate a cyclic schedule. Among the latest publications, the approach of Ierapetritou and Floudas (1998) is to be mentioned that considers sequence-dependent setup times for both batch production and continuous production. They report that model sizes of practical relevance have been solved within reasonable time. Méndez and Cerdá (2002) suggest a similar model formulation that is characterized by a lower number of bi-nary variables and hence is computationally more efficient. Nevertheless, none of the named models supports lot sizing Nevertheless, none of the mentioned models supports lot sizing and it is not possible to integrate demand data of a single final product for every single production day, only for aggregate demand of the entire planning horizon.

Günther and Neuhaus (2004) present an approach that is based on the principle of block planning and that simultaneously considers lot sizing and scheduling (see Fig. 7.3; with dejd indicating demand element of product j assigned to the end of demand period d). By integrating several variants of a product type or recipe into a “block”, the complexity of the model is significantly reduced. Within a block, a “natural” sequence of batches often exists, for example from the lower taste to the stronger or from the brighter color to the darker. The product sequence within a block is determined by increasing color intensity of the products, e.g. banana be-fore cherry taste (visualized by increasing darkness of the bars).

Fig. 7.3 Block planning approach (based on Günther and Neuhaus 2004)


Fig. 7.4 Yogurt production product tree

In order to guarantee the compactness and computability of the models, this block planning approach is chosen as a basis for the model development. A block covers all products based on the same recipe (see Fig. 7.4). As packaging facilities are multi-functional and can process several recipes, it is always necessary to per-form a set-up operation when changing the production between two products that are not based upon the same recipe. Only when changing the production between two product variants of the same recipe, the cleaning and sterilizing of the produc-tion facilities can be neglected.

Hence, not only the sequence of products within a block is fixed but also the sequence of recipes/blocks can be fixed within the production day. In that case, the different recipes are enumerated according to their position within the day. Forming a block and producing the various product variants of a single recipe fa-cilitates the planning and scheduling of the fermentation tanks since only the scheduling results for the blocks must be considered.

Like in the approach of Günther and Neuhaus (2004), the developed models are based on a continuous representation of time. For the consideration of shelf life, it is necessary to employ a discrete, uniform time grid in addition (Günther and Neuhaus 2004). A period refers to a day as shelf life for yogurt is usually given in days. For the application of block planning in fresh food production, it is not nec-essary to employ a stock balance, because only by conserving the information on the production- and demand-date, a reliable calculation of the shelf life is possible. Therefore, a variable is introduced which assigns production lots to demand dates.

Regarding the objective function, time-based objectives seem to be of little relevance for the regarded problem since additional time needed for production due to an extension of shelf life has to be weighed up against the utilization of ad-ditional capacity on the weekend. On the other hand, cost objectives fail to con-


sider additional revenue due to an extended shelf life. Therefore, the chosen objec-tive function maximizes the contribution margin by considering revenues and variable cost elements. As the entire demand has to be met, regular revenues are not considered. Furthermore, fixed costs such as line depreciation are not taken into account. Shelf life is considered in the objective function by a shelf life de-pendent pricing component supposing that the manufacturer can yield a financial benefit if the products have a longer shelf life when being delivered. This pricing component may also include holding costs. The shelf life dependent benefit in-creases linearly between the minimum customer requirement on shelf life and the maximum possible shelf life.

In the following chapter, three different model formulations for the yogurt pro-duction planning problem described above are proposed. In the first model, the production of a product or block over midnight (one batch in two macro-periods) is not possible. This limitation is lifted in the second model by introducing an ad-ditional binary variable. The third model is a completely different model formula-tion, which relies on a position-based method. In this model, overnight production is possible as well. In Chapter 7.4, some computational results are given. The per-formance of the models is analyzed and their suitability for different planning situations is assessed.

7.2 Model Formulations

7.2.1 Model 1: Model with Day Bounds

This “Model with Day Bounds” (MDB) allows every product to be produced on every production day. The recipe sequence on a line is the same on every produc-tion day. As depicted in Fig. 7.5, the sterilization of the line (sterl) is required at the beginning of each block. After having produced all products of the block (Xjpl),the line must be cleaned (cll). The cleaning process causes a loss of fermented milk that is still in the tubes and units at the end of production. In this model, a production of a product over midnight is not possible. Between two successive days, a line always has to be cleaned at the end of the day and sterilized at the be-ginning of the following day.

Precondition for the production of a product on production day p and on pack-aging line l is that the corresponding block r is set up on production day p and on packaging line l (Srpl = 1). In order to cover the demand of product j of demand day d (dejd), all quantities (Xjpl) can be considered that have been produced on pro-duction days prior to the demand day (p < d) or products from stock of the previ-ous week.

7.2 Model Formulations 153

Fig. 7.5 Variables of the MDB model for yogurt production

Indices

j,k J products l L lines s S days p P S production days d D S demand days r R recipes, blocks j J(r) products based on recipe r

l LR(r) lines that can process recipe r

l LJ(j) lines that can process product j d D(s) demand days (to meet the demand on these days, lots produced

on day s can be considered) s S(d) production days (the lots produced on these days can be con-

sidered to meet the demand of demand day d)

Parameters

varcj variable costs for the production of one unit of product jslj maximum shelf life of product j in days B1, B2 sufficiently large numbers capl capacity of line l, in units per day sterl sterilization time of line lcll cleaning time of line lbenj maximum additional benefit when meeting the maximum shelf

life of product j, in € per kg


lossr loss of fermented plain yogurt of recipe r when cleaning the line, in kg

clossr costs for the cleaning loss of plain yogurt of recipe r, in € per kg dejd demand of product j on demand day dsjs inventory of product j, produced on day scl costs of utilization of line l, in € per daymb minimum batch size to be processed, in kgpsj packaging size of product j, in kg per unitftr fermentation time for recipe r, in hours fc fermentation capacity, in kg-hours per day crj minimum shelf life of product j required by the customer (as a

fraction of maximum shelf life, applied to multiply the shelf life of product j)

qj maturation time of product j osl overtime supplement for weekend production on line l, in € per

dayadj percentage of plain yogurt contained in one unit of product jfdp start of the first production day within the week (Sunday) ldp start of the last production day within the week (Saturday)

Decision Variables

Srpl =1, if recipe r is set-up on production day p on line l (0, other-wise)

Xjpl units of product j produced on line l on production day pZjds units of product j produced on production day s that is used to

meet the demand of demand day dLrpl duration of recipe/block r on production day p on line lENDrpl end time of recipe/block r on production day p on line lESTl start time of line lLFTl end time of line lSAOl overtime at the end of scheduling horizon (Saturday) on line lSUOl overtime at the beginning of the scheduling horizon (Sunday) on

line l

Objective Function

Pp Rr rLRl

rrrpl

Ll

lll

Ll

lll

Jj Pp jLJl

jjpl

Jj Dd dSs jj

jj

jjds

closslossSosSUOSAO

cESTLFTvarcX

slcr

sdslcrbenZ

)(

)(

)( 1

1max

(7.1)


Fig. 7.6 Shelf life and shelf life dependent pricing component

The objective function aims at maximizing the contribution margin. It contains the profit from the shelf life depending pricing component (benj). This benefit in-creases linearly between the minimum customer requirement on shelf life (crj) and the maximum possible shelf life (slj) since the benefits for the retailer increase with every additional day of residual shelf life (see Fig. 7.6). As an example, sup-posing a product j with a total shelf life of 30 days (slj = 30) and a customer re-quirement on the minimum residual shelf life when being delivered of 66% of the total shelf life (crj = 0.66). Supposing further that the shelf life of the product starts on day 6 (s = 6), the product is delivered to the retailer on day 10 (d = 10) and the maximum benefit for meeting the maximum shelf life of product j, benj is€ 0.30 per kg. In this case, the manufacturer yields a financial benefit of € 0.18 per kg of product j (60% of the maximum benefit).

This financial benefit can be justified by several factors. On the one hand, the manufacturer will probably yield a higher turnover as consumers tend to buy the product with a longer remaining shelf life (see Chapter 5.3.2). On the other hand, the retailer will yield financial benefits as well (e.g. less write-offs due to stock obsolescence or less stock-outs) and parts of the benefits are transferred to the manufacturer by means of the shelf life dependent pricing component in order to cover its higher production costs.

Costs in the objective function include the variable costs for the input factors (varcj), the costs of the utilization of the packaging lines in regular (cl) and over-time (osl) mode and the set-up costs for cleaning losses of plain yogurt (lossr *

clossr).

Constraints to be considered are the following.

jj

jj

jslcr

sdslcrben

1

1


Set-Up

1)(

BSXrpl

rJjjpl

p P; r R; l LR(r) (7.2)

Product j can only be produced on production day p and on packaging line l if the line is set-up for the corresponding recipe (Srpl = 1). A production day is one of the days within the planning period from Sunday to the following Saturday. The large number B1 has the function to allow the production of product j on produc-tion day p and on packaging line l if the corresponding block r is set-up on pro-duction day p and on packaging line l. Therefore, this number should be as high as the highest possible output of a product on a line per day. Choosing a higher num-ber than the highest possible output of a product per day will not improve the value of the objective function, however this might impair the computational per-formance of the model. On the other hand, choosing a lower number than the highest possible output of a product per day can have a negative effect on the Ob-jective Value (OV) as this reduces the available capacities.

Output Quantities

rpl

rJj l

jpl

llrpl Lcap

XclsterS

)(

)( p P; r R; l L(r) (7.3)

The duration of block r on line l (Lrpl) contains the sterilization (sterl) and cleaning (cll) times as well as the production time of the products based on this recipe j J(r). If a line is not prepared for a recipe (Srpl = 0), the duration may be-come zero and hence the start and end time of this block are the same.

Sequencing

lprrplrpl ENDLEND ,,1 p P; r R: r>1; l LR(r) (7.4)

Due to the fixed sequence of recipes within a day, this constraint ensures that block r may not start before the end of its predecessor r-1. This avoids an overlap-ping of blocks on packaging line l.

Day boundaries

1pENDrpl p P; r R; l LR(r) (7.5)

pLEND rplrpl p P; r R; l LR(r) (7.6)

These constraints ensure that every production lot is assigned to one distinct production day. Equation 7.5 makes sure that every block ends before the end of the corresponding day, Equation 7.6 that every block starts after the beginning of this day. The feasible interval for the float variables indicating the end of a block is set by the interval derived from the integer day numbers. For instance, accord-ing to Equations 7.5 and 7.6, any block r produced on day p=5, must be completed


by 6 and started before 5. Thus, ENDrpl will assume a value between 5.00 + Lrlp

and 6.00.

Stock balance

)()( sDd

jds

jLJl

jpl ZX j J; s S: ldp s fdp; p P: p=s (7.7)

This constraint allows the produced lots of product j on different packaging lines to be considered in order to meet the demands of different demand days dand makes sure that the volume is sufficient.

)(sDd

jdsjs Zs j J; s S: s<fdp (7.8)

This constraint assigns the inventory to demand days. This does not account for inventory built up in the current planning period.

Utilization of packaging lines

21 BSENDLFTrplrpll

p P; r R; l LR(r) (7.9)

The end time of packaging line l within the planning horizon is the same as the end of the last block produced on this line. Yet, because even those blocks that do not have any production output have an end time (which is equal to the begin-ning), it is necessary to ensure the line was actually set-up for the block. If this is not the case, the constraint is still valid for all values of LFTl due to the subtraction of a large number (B2). This number must be larger than the number of the last day on which overtime is allowed. For example, if the last day considered is day 10, than B2 must at least take a value of 11. As explained for Equation 7.2, this large number should be determined as small as possible in order not to worsen computa-tional times.

21 BSLENDESTrplrplrpll

p P; r R; l LR(r) (7.10)

Similar to Equation 7.9, the start time of packaging line l within the planning horizon is the same as the start time of the first block produced on this line.

ll ESTfdpSUO 1 l L (7.11)

The overtime at the beginning of the week (Sunday) can be calculated by sub-tracting the start time of the line from the end of the first day (fdp+1). In case Sun-day overtime is not utilized, SUOl will become zero due to Equation 7.21.

ldpLFTSAO ll l L (7.12)

The overtime at the end of the week (Saturday) is determined by subtracting the end time of the regular working period (ldp) from the end time of the last block on


the packaging line. If Saturday overtime is not utilized within a planning period, SAOl will become zero due to Equation 7.21.

Meeting demand

)()1()(:)( dSs

jds

slcrsddqsdSs

jdsjd ZZdejjj

j J; d D(7.13)

The demand of product j on demand day d can only be filled using lots that comply with the requirements concerning shelf life and maturation time. A quar-antine time is precondition for the achievement of the desired sensory qualities. Therefore, the products can only be delivered after the quarantine time has passed (s+qj < d). Moreover, customers demand a minimum residual shelf life (crj), which must be respected (d-s (1-crj)*slj). For the validity of the objective func-tion (Equation 7.1) it is necessary that the volume meeting these requirements be the only one considered to meet the demand on the corresponding demand days. In addition, due to the powerful retail customers who do not allow backordering, the demand must be completely fulfilled.

Minimum batch sizes

rpl

rJj

jjjpl SmbadpsX)(

r R; p P; l LR(r) (7.14)

Since the production utilizes fermentation capacity, it is necessary to guarantee a certain filling level of the fermentation tanks, which is ensured by minimum batch sizes.

Fermentation capacity

fcftlossSftadpsXRr rLRl

rrrplr

r rJj rLRl

jjjpl

)()( )(

p P (7.15)

The daily fermentation volume must not exceed the fermentation capacity. Due to different fermentation times of recipes, the fermentation capacity is given in kg-hours.

Variable domains

1;0rplS p P; r R; l LR(r) (7.16)

0jplX j J; p P; l LJ(j) (7.17)

0jdsZ j J; s S; d D(s) (7.18)

0rplL r R; p P; l LR(r) (7.19)


fdpENDldp rpl1 r R; p P; l LR(r) (7.20)

0,,, llll SUOSAOLFTEST l L (7.21)

This model is suitable to integrate shelf life aspects into the planning and scheduling decisions in yogurt production, mainly because of the assignment of lots to specific production days. It does not support the conservation of the set-up state in order to allow production over midnight. It is possible to use this model for facilities that interrupt production during nighttime. However, high system utilization is desirable because packaging lines used in yogurt production are rela-tively capital intensive. Therefore, the model does not necessarily meet all re-quirements of the yogurt production. Yet it can be used for estimating the costs, the set-up frequency or the profit of a planning week.

7.2.2 Model 2: Model with Set-up Conservation

In order to allow overnight production, the model MDB is extended to the “Model with Set-up Conservation” (MSC) by adding a binary variable that conserves the set-up state over midnight. Furthermore, additional information about the position of the products within a block is necessary, in order to determine the exact product start and end times for overnight production.

Additional Parameters

bpj Position of product j in the corresponding block

Additional Variables

Cjpl =1, if product j is produced on line l until the end of production day p-1 and at the beginning of day p (0, otherwise)

Objective Function

The objective function of the model MDB (Equation 7.1) is replaced by the fol-lowing:

Pp Rr rLRl

rr

rJj

jplrpl

Ll

lll

Ll

lll

Jj Pp jLJl

jjpl

Jj Dd dSs jj

jj

jjds

closslossCS

osSUOSAOcESTLFTvarcX

slcr

sdslcrbenZ

)( )(

)(

)( 1

1max

(7.22)

The conservation of the set-up state for a day p has the effect that the cleaning loss is only applicable for day p-1. Although the cleaning itself takes place at the end of a block on a line, the loss is considered at the beginning of each block.


Output quantities

Furthermore, the constraint concerning the production output (Equation 7.3) has to be replaced by Equation 7.23 and 7.24.

rpl

rJj l

jpl

l

rJj

lpjrpll

rJj

jplrpl Lcap

XclCSsterCS

)()(,1,

)(

p P: p < ldp; r R; l LR(r)

(7.23)

As in Equation 7.3, the duration Lrpl includes the effective production time and the time for sterilization and cleaning. If the set-up state for product j J(r) is conserved from the previous day (Cjpl = 1), the sterilization time can be neglected. Furthermore, cleaning can be neglected, if the set-up state for product j J(r) is conserved for the following day (Cj,p+1,l = 1).

rpl

rJj l

jpl

lrpll

rJj

jplrpl Lcap

XclSsterCS

)()(

p P: p = ldp; r R; l LR(r)

(7.24)

A conservation of the set-up state is not possible for the last day of production (p = ldp), from Saturday to Sunday. The variable Cj,p+1,l is not defined for this case. On that day (Saturday; p = ldp), the transfer of the set-up state to the follow-ing day (Cj,p+1,l = 1) is not possible, since this day is not part of the planning hori-zon and the variable Cj,p+1,l is therefore not defined. Consequently, the packaging line has to be cleaned on this day, no matter which recipe was produced last. However, at the beginning of this day a conservation of the set-up state from the previous day is possible; in this case (Cjpl = 1) the time for sterilization can be ne-glected.

Minimum batch size

The constraints 7.25-7.27 replace Equation 7.14 and guarantee the respect of the minimum batch size for the fermentation facilities.

)( )(,1,,1, 1

rJj rJjlpjjjlpjjpl

mbmbCadpsXX

r R; p P: p < ldp; l LR(r)

(7.25)

If the set-up state for product j J(r) is transferred from production day p top+1, the amount of plain yogurt of the following day can be added in order to meet the minimum batch size (mb). Even in case the conservation of the set-up state for a recipe does not take place, the validity of the constraint is still guaran-teed. The right side of the Equation 7.25 does not need to be multiplied by the set-up variable Srpl, as the equation is only relevant for 1

)(,1,

rJj

lpjC .

In that case, Srpl is always equal to one because the set-up state for a product jof block r (j J(r)) can only be established if the corresponding block r is set up (Equation 7.35). Since a conservation of the set-up state from the last day of pro-


duction (p = ldp) to the successive day is not possible, this constraint is valid for all days of production but the last.

rplrJj

lpjrJj

jpljrJj

jjplSmbmbCBCadpsX

)(,1,

)()(

r R; p P: p < ldp; l LR(r)

(7.26)

In case the set-up state of a recipe is neither transferred from day p-1 to day pnor from day p to day p+1 (Cjpl = 0 and Cj,p+1,l = 0 for all j J(r)), the amount of plain yogurt of this block has to meet or exceed the minimum batch size.

rplrJj

jpljrJj

jjplSmbmbCadpsX

)()(

r R; p P: p = ldp; l LR(r)

(7.27)

Since the conservation of the set-up state from the last day of production (p=ldp) to its successor is not possible, the amount of plain yogurt for a block has to either meet or exceed the minimum batch size, except the set-up state is trans-ferred from the previous day.


fcftlossCSftadpsX rr

Rr rLRl rJj

jplrplr

Rr rJj rLRl

jjjpl

)( )()( )(

p P

(7.28)

The conservation of the set-up state for the day p has the effect that the cleaning loss can be neglected on production day p (Srpl – Cjpl = 0). Therefore the constraint concerning the fermentation capacity (Equation 7.15) has to be altered as indicated in Equation 7.28.

The following constraints (Equations 7.29 - 7.37) have to be considered in ad-dition to the ones formulated in Chapter 7.2.1:

Conservation of the set-up state

1,1,:)(

1 BCX lpj

bpbprJk

kpl

jk

r R; j J(r); p P: p < ldp; l LR(r) (7.29)

If product j J(r) is produced until the end of production day p and the set-up state is conserved for the following day (Cj,p+1,l = 1), the output quantities of the following products within this block (products with a higher block position than product j; bpk > bpj) have to be zero for that day.

lpjrpl CpEND ,1, r R; j J(r); p P: p < ldp; l LR(r) (7.30)


The variable Cj,p+1,l can only take the value of 1 if the production of product j J(r) on production day p runs right until the end of day p. Production of prod-uct j may then continue without interruption on day p+1 (see following con-straints).

lpjjpl CX ,1, j J; p P: p < ldp; l LJ(j) (7.31)

For those days, for which a transfer of the set-up state is possible (p < ldp), the conservation variable Cj,p+1,l can only take the value one if there has been a pro-duction output of product j at the end of day p.

)(

1rJj

jplrplrpl CpLEND r R; p P; l LR(r) (7.32)

If the set-up state is conserved from the day p-1 to the day p for product jJ(r), the production of this block r has to continue directly at the beginning of the day p. The expression ENDrpl – Lrpl – p can only take a value of zero if the start of the production (ENDrpl – Lrpl) is equal to the beginning of the production day the p. For example, if p represents day 5, then the start of the production of block r(ENDrpl – Lrpl) must be exactly at 00:00 on day 5 (=5.00) in order to set the left hand side of the constraint to zero. This value of zero on the left hand side of the constraint is necessary for the binary variable Cjpl to take a value of one.

1:)(

1 BCXjk bpbprJk

kpljpl r R; j J(r); p P; l LR(r) (7.33)

In case the set-up state is conserved for product j J(r) from production day p-1 to day p, production on this day may only continue with a product from this block. For all predecessors the output on production day p has to be zero. The large number B has the function to allow the production of product j in case the set-up state is not conserved for product k with a higher block position than prod-uct j (for the determination of this large number B1, it is referred to Equation 7.2).

1Jj

jplC p P; l L (7.34)

For every day and line, the set-up state for only one product can be conserved.

0)(rJj

jplrpl CS r R; p P; l LR(r) (7.35)

Only if a line is set-up for block r, the set-up state can be conserved for product j J(r) of this block. Cjpl can only take a value of one if Srpl takes a value of one as well.

0jplC j J; l LJ(j); p = fdp (7.36)


At the beginning of the week (p = fdp) a conservation of the set-up state from the preceding day is not possible.

Variable domains

}1;0{jplC p P; j J; l LJ(j) (7.37)

This model is capable of conserving the set-up state while meeting the strict as-signment of production lots and days. Therefore, it is not only possible to produce a product overnight but also for more than two successive days. Model MSC is an extension of the model MDB with a higher degree of flexibility. Yet the fixed se-quence of recipes on every packaging line is no longer random under shelf life as-pects. The order of the recipes influences the conservation of the set-up state.

7.2.3 Model 3: Position Based Model

Driven by the analysis of the regular permutation flowshop problem performed by Tseng et al. (2004), a third model formulation has been developed that relies on an assignment approach. Although assignment model formulations are more complex in terms of the number of binary variables involved, Tseng et al. (2004) report that assignment models perform significantly better than dichotomous models, espe-cially in case of large model instances. Hence, for the so-called “Position Based Model” (PBM), the planning horizon is split into consecutively enumerated posi-tions i I. A block can be assigned to every position (see Fig. 7.7). The beginning and the end of every block have to be on a given day the position is assigned to by the parameter estartli. Precondition for the set up of a block is that it is assigned to a specific position (Birl = 1).

Fig. 7.7 Variables of the PBM model for yogurt production


In case of a production over midnight, the set-up state of the previous days can be conserved to the following day (Cjol = 1), however only for the first position of a day (o O I). Then, the cleaning of the line on the production day and the sterilization of the line on the following day can be neglected. Since the model formulation is considerably different from the previous ones, the entire model PBM is presented. The performance of all model formulations is compared in Chapter 7.3.

Indices

j,k J products l L lines s S days p P S production days d D S demand days r,r' R recipes, blocks i I positions i,o O I positions at the beginning of a day, without first day of production j J(r) products based on recipe rl LR(r) lines that can process recipe rl LJ(j) lines that can process product jd D(s) demand days (to meet the demand of these days, lots produced

on the day s can be considered) s S(d) production days (the lots produced on these days can be con-

sidered to meet the demand of demand day d)

Parameters

estartli earliest starting time of a block at position i on line lvarcj variable costs for the production of one unit of product jslj maximum shelf life of product jB1, B2 sufficiently large number capl capacity of line l in units per day sterl sterilization time of line lcll cleaning time of line lbenj maximum additional benefit when meeting the maximum shelf

life of product j, in € per kg lossr loss of fermented plain yogurt of recipe r when cleaning the line,

in kg clossr costs for the cleaning loss of plain yogurt of recipe r, in € per kg dejd demand of product j on demand day dsjs inventory of product j, produced on day scl costs of utilization of line l, in € per day mb minimum batch size to be processed, in kg psj packaging size of product j, in kg per unit ftr fermentation time for recipe r, in hours


fc fermentation capacity in kg-hours per day crj minimum shelf life of product j required by the customer (as a

fraction of maximum shelf life, applied to multiply the shelf life of product j)

bpj position of product j in the corresponding block qj maturation time of product josl overtime supplement for weekend production on line l, in € per

dayadj percentage of plain yogurt contained in one unit of product jfdp start of the first production day within the week (Sunday) ldp start of the last production day within the week (Saturday)

Decision Variables

Bril =1, if recipe r is set-up on position i on line l (0, otherwise)Cjol =1, if product j is produced on position o-1 until the end of the

production day p-1 (estartlo = p) and on position o on packaging line l (0, otherwise)

Yjil units of product j produced on line l on position iZjds units of product j produced on production day s that is used to

meet the demand of demand day d Lril duration of recipe/block r on position i on line lENDril end time of recipe r on position i on line lESTl start time of line lLFTl end time of line lSAOl overtime at the end of the scheduling horizon (Saturday) on line

l

SUOl overtime at the beginning of the scheduling horizon (Sunday) on line l

Objective function

rrRr rLRl Oo rJj

jolrolOiIi Rr rLRl

rrril

Lllll

Lllll

Jj Ii jLJljjil

Jj Dd dSsjj

jj

jjds

closslossCBclosslossB

osSUOSAOcESTLFTvarcY

slcr

sdslcrbenZ

)( )(: )(

)(

)( 1

1max

(7.38)

This objective function is similar to Equation 7.22, with the production days pbeing replaced by positions i.

Constraints to be considered are the following.


Set-up

1)(

BBYril

rJjjil

i I; r R; l LR(r) (7.39)

1Rr

rilB i I; l L (7.40)

Similar to Equation 7.2, the set-up constraint for the production output has to be met. Additionally, a position may not be assigned to more than one block. For the determination of the large number B1, it is referred to Equation 7.2.

Output quantities

ril

rJj l

jil

llril Lcap

YclsterB

)(

)(

r R; l LR(r); i I: i O & (i+1) O

(7.41)

For those positions that can neither be scheduled at the beginning nor at the end of a day (i I: i O & i+1 O), the full sterilization and cleaning time must be considered within the duration Lril.

rol

rJj l

jol

lroll

rJj

jolrol Lcap

YclBsterCB

)()(

r R; l LR(r); o O

(7.42)

For a position that can be scheduled at the beginning of a day (i O), the set-up state for a product produced on the previous day can be conserved. Then, the sterilization process at the beginning of the day does not need to take place.

ril

rJj l

jil

lrill

rJj

lijril Lcap

YsterBclCB

)()(,1,

r R; l LR(r); i I: (i+1) O

(7.43)

Similarly, the set-up state of a position that can be scheduled at the end of a production day (i+1 O) can be transferred to the following day. Then, the clean-ing of the packaging line does not need to be considered.

Sequencing

lilirlir ENDLEND ,,r',1,,1,i I: i+1 I; r,r' R; l LR(r) LR(r') (7.44)

Similar to Equation 7.4, a block r' scheduled for position i must be finished be-fore block r at position i+1 is allowed to start.


Day bounds

1lirilrilli estartENDLestart r R, l LR(r), i I (7.45)

For any block r is assigned to position i, the start (ENDril – Lril) and the end (ENDril) of block r may only take place on the day for which the position i has been defined. For example, a value of estartli = 5.00 means that position i is car-ried out on day 5. ENDril has to take a value between 5.00 + Lril and 5.00 + 1.

Stock Balance

)(: )( sDd

jds

sestartIi jLJl

jil ZYli

j J; s S: ldp s fdp (7.46)

In analogy to Equation 7.7, the output of product j on different lines may be added to meet the demand of different demand days. Furthermore, the production quantities need to be sufficient.

)(sDd

jdsjs Zs j J; s S (7.47)

Inventory can be used to meet the demand of different demand days, yet the de-mand may not exceed the volume stored. As this does not account for inventory built up in the current planning period, the parameter sjs is zero for all days s > fdp.

Utilization of the packaging lines

2)1( BBLENDESTrilrilrill

r R; i I; l LR(r) (7.48)

As in Equation 7.10, this constraint sets the value of the variable ESTl (the start of production on packaging line l) equal to the starting time of the first block pro-duced on this line in the planning period.

2)1( BBENDLFTrilrill

r R; i I; l LR(r) (7.49)

Equation 7.49 sets the value of variable LFTl (the end time of production on packaging line l) equal to the finishing time of the last block produced on this line.

ll ESTfdpSUO 1 l L (7.50)

The overtime at the beginning of the week (Sunday) can be calculated by sub-tracting the start of the packaging line from the beginning of the regular working time (fdp+1).

ldpLFTSAO ll l L (7.51)

The overtime at the end of the week (Saturday) is determined by subtracting the end of the regular working period (ldp) from the ending time of the packaging line.


Meeting demand

)()1()(:)( dSs

jds

slcrsddqsdSs

jdsjd ZZdejjj

j J; d D(7.52)

In analogy to Equation 7.12, this constraint ensures that the external demand is met. Only those lots may be considered that fulfil the requirements of maturation time and minimum shelf life.

Minimum batch size

ril

rJj

jjjil BmbadpsY)(

r R; l LR(r); i I: i O & (i+1) O

(7.53)

For those positions that can neither be scheduled at the beginning of a day, nor at the end of a day (i I: i O & i+1 O), the volume of plain yogurt required for production of the block assigned to that position has to meet or exceed the minimum batch size.

)( )(,1,,1,

rJj rJj

jillirriljjlijjil CBBmbadpsYY

r R; l LR(r); i I: i O

(7.54)

For the positions that can be scheduled at the beginning of a production day, the required quantities of plain yogurt may be added to meet the minimum batch size if the set-up state is conserved from position i-1 to position 1.

rolrJj

joljjrJj

jolBmbmbCadpsY

)()(

r R; l LR(r); o O

(7.55)

If the set-up state is not conserved for a block scheduled at a position at the be-ginning of a production day, the amount of plain yogurt required for this block has to respect the minimum batch size.

rilrJj

lijjjrJj

jilBmbmbCadpsY

)(,1,

)(

r R; l LR(r); i I: (i+1) O

(7.56)

If the set-up state is not conserved for a block scheduled at a position at the end of a production day (i+1 O), the volume of plain yogurt required for this block has to meet or exceed the minimum batch size.



fcftrlossCB

ftrlossB

ftadpsY

rr

Rr rLRl pestarOo rJj

jolrol

Rr rLRl OipestarIi

rrril

r

r rJj rLRl pestartIi

jjjil

lo

li

li

)( : )(

)( :

)( )( :

p P

(7.57)

The utilized fermentation capacity must not exceed the installed capacity.

Conservation of the set-up state

lijli

bpbprJk l

killrilril Cestart

cap

YsterLEND

jk

,1,:)(

r R; j J(r); l LR(r); i I: (i +1) O

(7.58)

This constraint ensures that for positions, which can be scheduled at the end of a day (i+1 O), the conservation variable Cjol can only take the value of one for a product j J(r) if this product is produced until the end of that production day. For this purpose, the start time of the block on position i (ENDril – Lril) plus the time required to sterilize the line (sterl) plus the production time of all products with a block position lower than or equal to product j (bpk bpj) must be equal to the end of the production day, otherwise the binary variable Cj,i+1,l cannot take the value of one.

lijjil CY ,1, j J; l LJ(j); o O; i I: (i + 1) O (7.59)

The set-up state can only be preserved for a product j assigned to a position i at the end of a day (i+1 O), if the output of that product takes a positive value.

jollorolrol CestartLEND 1 r R; j J(r); l LR(r); o O (7.60)

If the set-up state is conserved for product j J(r) from position i-1 to position i O, the production of the block the product is based on has to continue directly after the beginning of the day the set-up state is transferred to. The expression ENDrol – Lrol – estartlo can only take a value of zero if the start of the production (ENDrol – Lrol) is equal to the beginning of the production day the position o is as-signed to (estartlo). A value of zero is necessary for the binary variable Cjol to take a value of one.


1:)(

1 BCYjk bpbprJk

koljol r R; j J(r); l LJ(j); o O (7.61)

As in Equation 7.33, in case the set-up state is conserved for a product j J(r)

from the position i-1 to the position i, production on this day may only continue with a product based on this block. Furthermore, production may only start with the product j J(r) or a successor of this product; the output on production day phas to be zero for all predecessors. The logic of this constraint is the same as for Equation 7.33. For the determination of the number B1, it is referred to Equation 7.2.

Jj

jolC 1 o O; l L (7.62)

Equation 7.62 ensures that the preservation of a set-up state can only take place for one product for every packaging line l L and every position that can be scheduled at the beginning of a day (o O).

0)(rJj

jolrol CB r R; l LR(r); o O (7.63)

Similar to Equation 7.35, the set-up state can only be preserved if the line is set-up for the block. Cjol can only take a value of one if Brol has taken a value of one as well.

Variable domains

1;0rilB i I; r R; l LR(r) (7.64)

1;0jolC j J; o O; l LJ(j) (7.65)

0jilY j J; i I; l LJ(j) (7.66)

0jdsZ j J; s S; d D(s) (7.67)

0rilL r R; i I; l LR(r) (7.68)

fdpENDldp ril1 r R; i I; l LR(r) (7.69)

0,,, llll SUOSAOLFTEST l L (7.70)

7.3 Computational Results 171

The solvability of this MILP model mainly depends on the number of binary variables and on the number of positions defined. This number must be deter-mined before solving the model. The number of positions can be restricted since the cleaning and sterilization of a packaging line is always necessary when pro-ceeding with the sequence of positions except when conserving the set-up state. For example, the number of possible positions per day multiplied by the set-up time of the considered packaging line should not exceed 24 hours. In order to keep the number of binary variables small, the conservation of the set-up state is only possible for positions at the beginning of a day (o O). Therefore, it is impossible for models that require more than one possible position per day to produce one block for more than 48 hours. To determine the number of positions required for a certain scheduling problem, the expected system utilization, the number of recipes to be processed on a line and the set-up time should be considered. Furthermore, it is possible to increase the number of positions iteratively and break-off if an im-provement of the OV can no longer be observed.

7.3 Computational Results

7.3.1 Simultaneous Optimization of All Lines

The purpose of the numerical investigation is to assess the suitability of the mod-els for specific planning problems in the industry. The different models can then be transformed into a tool-kit for the planner. To determine an acceptable MIP-Gap, the interpretability of the objective function has to be taken into account. The MIP-Gap is the difference in percent between the actual OV and a theoretical up-

per bound for the optimal OV which is obtained from a LP relaxation of the prob-lem. In their search for the optimal solution, branch-and-bound procedures sequen-tially add the missing integrality constraints to the LP relaxation and thus reduce the MIP-Gap to the best known feasible solution. Hence, the MIP-Gap represents the maximum remaining improvement potential of the OV. Since the revenues in the dairy industry are very small (Murmann and Wolfskeil 2004), the MIP-Gap should be less than 1 %. Therefore the objective functions of the presented models are altered for better interpretation by adding the following term:

Jj Ddjjd revd

The parameter revj represents the fixed pricing component per unit of product jwhen selling it to the retail organizations, which represents the price when the products are provided with the lowest possible shelf life. By adding this term, the objective function value can be interpreted as the contribution margin after sub-traction of all variable costs. The data set used to demonstrate the practical appli-cability of the proposed models consists of 30 products based on 11 recipes that can be processed on 4 packaging lines, which represents a mid- to large-scale yo-gurt production environment. In addition, the demand quantities differ signifi-


cantly per product, ranging from high-volume bulk products for the discount channel to the production of specialties. Two of the packaging lines process the same range of products (packaging line type c, lines 3 and 4); lines 1 and 2 serve for packaging a specific range of products each. Table 7.1 indicates the number of variables for the model presented in the previous section (figures indicated for PBM are based on 21 positions per week, i.e. 3 positions per day).

Table 7.1 Number and type of variables of the yogurt production models

Model No. of binary variables No. of continuous variables Total no. of variables

MDB 112 2,641 2,753

MSC 413 2,641 3,054

PBM 594 3,691 4,285

The numerical investigation was performed on a PC (MS Windows XP Profes-sional) with an AMD XP 2600+ CPU, 1 GB RAM. The models were implemented using ILOG’s OPL Studio 3.6.1 as a modeling environment and its incorporated standard optimization software CPLEX 8.1. The models were first examined based on the complete configuration of four packaging lines. The number of posi-tions used for model PBM was determined in advance to be 21 (3 positions per day).

The performance of the models is assessed along the dimensions OV, MIP-Gap, and CPU time (see Table 7.2 and Table 7.3). The OV is the value of the ob-jective function at the moment at which the optimization run is stopped. The CPU time is the time limit set for the optimization run or – in case the optimal value has been obtained (MIP-Gap = 0%) – the time required to obtain the OV. In order to generate the final solutions within short computational time (which is an important requirement for industry applications), two different time limits have been set, at 300s and 1,800s. In the first investigation all packaging lines are considered simul-taneously. The results of this first investigation show that the model MDB per-forms well regarding the MIP-Gap and the computational time. The additional de-grees of flexibility, introduced with the models MSC and PBM, do not result in a higher value within the time limits of 300s and 1,800s. However, the MIP-GAP of around 2% for models MSC and PBM indicates further potential for the OV. To realize these potentials, it is necessary to reduce the model complexity, e.g. by line decomposition or by planner’s experience.

Table 7.2 Optimizing all yogurt packaging lines simultaneously (t 300 s)

MDB MSC PBM

OV [€] 1,429,272 1,416,089 1,420,907

MIP-Gap 0.08% 2.37% 2.38%

t [s] 300.00 300.00 300.00


Table 7.3 Optimizing all yogurt packaging lines simultaneously (t 1,800 s)

MDB MSC PBM

OV [€] 1,429,498 1,416,089 1,428,973

MIP-Gap 0.00% 2.34% 1.78%

t [s] 1,167.14 1,800.00 1,800.00

7.3.2 Line Decomposition Approach

In the line decomposition approach, the different types of packaging lines are looked at separately. An independent optimization of the different packaging lines is possible, if the fermentation capacity is not limiting. Therefore, the feasibility of the solution obtained via this decomposition procedure has to be checked against the installed fermentation capacity. The optimization took place separately for the different packaging line types (type a: line 1; type b: line 2; type c: lines 3 and 4).

Table 7.4 Optimization of yogurt packaging line 1 (t 300 s)

MDB MSC PBM

OV [€] 383,773 384,806 384,752

MIP-Gap 0.00% 0.00% 0.13%

t [s] 0.54 29.48 300.00

Table 7.5 Optimization of yogurt packaging line 2 (t 300 s)

MDB MSC PBM

OV [€] 528,411 530,885 531,996

MIP-Gap 0.00% 0.00% 0.00%

t [s] 0.08 0.62 1.47

Table 7.6 Optimization of yogurt packaging lines 3 and 4 (t 300 s)

MDB MSC PBM

OV [€] 517,314 516,238 515,723

MIP-Gap 0.00% 1.58% 2.04%

t [s] 32.20 300.00 300.00

Table 7.7 Aggregate results by combining the line-specific optimization runs

MDB MSC PBM

OV [€] 1,429,498 1,431,929 1,432,470

MIP-Gap 0.00% 0.57% 0.76%

t [s] 32.82 330.10 601.47


The number of positions used for model PBM is 21 for types a and c, and 14 for type b. The number of positions to use per line type has been derived from the number of recipes and the volume of final products to produce on each line. In our example, line type b has to handle fewer recipes than the other line types. Hence the corresponding number of positions can be lower. Numerical results for opti-mizing the various lines separately are shown in Tables 7.4 to 7.6. Aggregate re-sults obtained from combining the three line-specific optimization runs are indi-cated in Table 7.7.

Results of the basic MDB model show that optimization runs are very fast and exact optimal solutions are obtained. In case all four packaging lines are optimized simultaneously, the optimal solution with an objective function value of € 1,429,498 is determined within 1,167 s (see Table 7.3). Optimizing the different types of packaging lines separately leads to a solution with almost the same objec-tive function value in a considerably shorter CPU time of 33 s (see Table 7.7). Therefore, decomposing the problem has shown to be effective in reducing com-putational times without impairing the overall quality of the production schedule. Also the other models benefit from the separate optimization of the packaging lines.

7.3.3 Model Combination and “Pick-the-Best” Approach

Still, the question is which model is the right choice for a specific problem. Using the planner’s experience can be another valuable approach to decrease model complexity. In particular for packaging lines 3 and 4, a combination of different models is recommendable. The distinct assignment of recipes to one or both lines can be derived from the results of the optimization run using model MDB (see Ta-ble 7.8).

Under these conditions, the objective function values as shown in Table 7.9 were obtained. For the two models with assignment of recipes to lines, the 32 s of the optimization run of model MDB have been added to the 300s computational time as model MDB must be run beforehand. With regard to OV and MIP-Gap it is noteworthy that particularly the results of model PBM have been improved by the assignment. On the other hand, model MSC improved only slightly.

Table 7.8 Assignment of recipes to yogurt packaging lines 3 and 4

Recipe Packaging Line 3 Packaging Line 4

1 + +

2 + +

3 - +

4 + -

5 + +


Table 7.9 Optimization for yogurt packaging line 3 and 4 (t 300 s)

MDB MSC MSC with recipe-

line assignment

PBM PBM with recipe-

line assignment

OV [€] 517,314 516,238 516,974 515,723 518,388

MIP-Gap 0.00% 1.58% 1.26% 2.04% 1.20%

t [s] 32.20 300.00 300.00 300.00 300.00

A suitable strategy for the decomposition of the entire optimization problem and the choice of adequate MILP models allows the planner in practice to deter-mine a suitable schedule within reasonable CPU time. Hence, for the case study considered, it is recommended to combine the presented models by taking the model that performs best per line type. This straightforward approach can easily performed by the planners in industry, without in-depth know-how of the model structures. In Table 7.10, such a possible combination is given. The solution is characterized by an MIP-Gap of approx. 0.5 %. The OV is almost 0.4 % higher than the best solution listed in Table 7.3. This result is obtained within only approx. 6 min. of computational time. The Gantt chart, created from these sched-uling results, is shown in Fig. 7.8.

Table 7.10 Combination of the different yogurt models

Line Model OV [€] t [s]

Line 1 MSC 384,806 29.48

Line 2 PBM 531,996 1.47

Line 3 & 4 MDB & PBM mZ 518,388 332.20

Total 1,435,190 363.15

Fig. 7.8 Scheduling results for all four packaging lines


Our numerical experiments revealed that the application of the MILP models represents a suitable approach for optimizing the production of yogurt in a case study taken from industry. The approach suggested appears to be very satisfactory both concerning the quality of the solutions as well as computational time. In or-der to benefit from the full potential of the presented models, practical experience and knowledge about the applicability of each of the models for specific problem settings is of major importance.

For an industrial application, intense education and training of the planners dealing with these models is required in order to ensure the correct use of the models. For instance, model MDB is suitable if only very short computational times are allowed or for determining the basic recipe-line assignment. Further-more, it should be used for more complex problems (e.g. longer planning horizons or greater variety of feasible recipe-line assignments). Model MSC is particularly appropriate for high-volume production with a limited variety of recipes (bulk), which requires the conservation of the set-up state for several consecutive days. For increased flexibility, desired for production systems covering a high variety of products (specialties), model PBM is more suitable, although this approach in-volves higher computational effort. Table 7.11 summarizes the main recommenda-tions for the different models.

Further numerical investigations were performed to analyze the effect of the shelf life dependent pricing component (benj), Its value has an influence on the number of set-up procedures performed and on the actual product shelf life. This effect is shown in Fig. 7.9 for the lines 3 and 4. The factor value of 1 represents the shelf life dependent pricing component as it was used in the above-mentioned computation (approx. 10% of the regular revenues).

In the further experiment, this pricing component was varied by using different factor values, ranging from 0.1 to 10. Starting with a factor value of 2 (meaning a shelf life dependent pricing component of approx. 20% of the regular revenues), the impact of this component increases significantly, leading to both fresher prod-ucts and more frequent set-ups.

Table 7.11 Suitability of the different yogurt models

Model Suitability

MDB Short computational time Complex planning problems due to long planning horizon Complex planning problems due to amount of recipes

MSC Large-scale bulk-production Production of one product over several days Many products based on very few recipes

PBM Small-scale specialty-production Planning problems for which computational time is not crucial

7.4 Conclusion 177

Fig. 7.9 Variation of the shelf life dependent pricing component (lines 3 and 4; run-time: 5 minutes)

7.4 Conclusion

Three different MILP model formulations for scheduling problems in fresh food industry have been presented. They have been shown to be suitable to generate near-optimal solutions for a planning and scheduling problem from industry. As the shelf life of the products has been explicitly considered, the use of the pro-posed planning tools promises improved product freshness in many industrial ap-plications. Different formulations have been presented for specific production en-vironments.

With respect to the planner in industry that often lack deep know-how in OR and mathematics, it is useful to provide clear guidelines on how to handle complex model instances in order to increase the solvability of the model. In the presented case study, a straightforward decomposition method based on line types has been proposed which can easily performed by the planner in industry. As next step, In order to further facilitate the planning task, an automated selection of the appro-priate model for the individual planning problem is desirable. In this case, a soft-ware will apply all three models to the planning problem and propose the one which is best suited with regard to OV and computational times. The selection cri-teria (e.g. importance of OV and computational times must be predefined. In addi-tion, the software should also support the determination of the right number of po-sitions required for the PBM model.

With regard to further extensions of the MILP models, one possibility is the ex-tension of model PBM, in order to allow that the production of a particular prod-uct may last several days. However, due to the small improvement, which was gained using model PBM instead of model MSC, this extension was not realized in the investigation. Since the complexity of such a model would still increase and

89,0%

89,5%

90,0%

90,5%

91,0%

91,5%

92,0%

92,5%

93,0%

93,5%

0,1 1 10

Factor for the Shelf Life Dependent Pricing Component

Avera

ge S

helf

Lif

e

[Sh

elf

Lif

e i

n %

of

Maxim

um

Sh

elf

Lif

e]

0

5

10

15

20

25

Nu

mb

er

of

Set-

Up

Op

era

tio

ns P

erf

orm

ed

Shelf Life in % of Maximum Shelf Life Number of Set-Up Operations Performed


the computational times would therefore be even higher, the practicability of this approach is questionable. Another option to extend the proposed models is the in-tegration of the fermentation processes into the planning procedure. Whether this can take place simultaneously to the scheduling of the packaging facilities is de-batable because of the multi-functional tanks often employed in industry. A se-quential planning procedure, which performs the planning and scheduling of the fermentation facilities based on the scheduling results of the packaging lines, seems to be more promising.

8 Shelf Life Integration in Sausage Production


Objective of the second case study is the integration of shelf life into the opera-tional production planning for scalded sausages (for the sausage production model, see also Zhang 2004). A description of the production processes is pro-vided in Chapter 3.6. The model scope comprises the production steps from scald-ing to storage and delivery, with special attention being paid to the scalding and slicing and packaging steps (see Fig. 8.1). All prior production steps from ingredi-ents input to stuffing and tying have been neglected for two reasons. First, the scalding step determines the shelf life of the final sausage product. In order to im-prove the freshness of the final product, it is sufficient to consider this and the fol-lowing processes. With respect to the meat raw materials, it must only be assured that they are processed within their shelf life period, which is different and much shorter than that of the final products. Secondly, scalding as well as slicing and packaging are typical bottlenecks of the entire production process. Therefore, the model will focus on ensuring a high utilization of these units. The storage and de-livery stage is modeled in analogy to the yogurt case study.

Fig. 8.1 Focus of sausage production model

180 8 Shelf Life Integration in Sausage Production

The production of sausages can be modeled as an extension of the models de-veloped for the production of yogurt. In contrast to the yogurt case study that comprises only one major production step (flavoring / packaging), the sausage model must cover two major production steps (scalding and slicing / packaging) within one model since the shelf life of the final products is already determined at the scalding step. Furthermore, the intermediate storage has to be integrated into the model additionally due to important time-dependent weight losses that occur at this stage. Thus, the modeling approach applied for this case study combines the strengths of the different models developed for the production of yogurt and inte-grates the principles of two of these models into a two-stage production model.

For the scalding stage, the PBM model is applied as in Chapter 7.2.3. Eight scalding chambers are considered with different costs per hour and different ca-pacities (see Fig. 8.1). After having brought the filled casings into the chamber, they are subject to several processes that include drying, smoking, scalding and cooling. All these activities are fully automated and take place within the chamber. The time required for this program is fixed and differs by the type of sausage pro-duced. However, if two intermediate sausages are based on the same program, they can be put together into the chamber (see Fig. 8.2). Set-up and cleaning times are only of minor importance and are part of the scalding program. The position-based modeling approach has been chosen to limit the size of the model. For ex-ample, if the shortest scalding program lasts about four hours, the number of posi-tions which determines how many batches can be processed during one day can be limited to six (24 hours divided by four hours). If the shortest scalding program takes more time, the number of positions can be further reduced to keep the num-ber of variables small. In contrast to the PBM model in the yogurt case study, the products are not processed successively (as on the packaging lines). The interme-diate products are processed simultaneously in the chambers. Therefore, the model formulation must be adapted accordingly.

Fig. 8.2 Sausage production product tree


Mainly due to the evaporation of water, important time dependent losses occur in the intermediate storage. The losses must be considered in the model to obtain correct sausage volumes for the packaging step. Furthermore, as most products are sold on a weight-basis, reducing the storage times of the intermediate products helps to increase the volume available for sale. Although it is desirable to keep the stock levels in the intermediate storage as low as possible, a minimum time is im-posed to avoid a later sticking of the slices at the packaging stage that would lead to a lower slicing quality.

The packaging stage is modeled relying on the MDB model (Chapter 7.2.1). At the packaging step, intense cleaning which takes several hours is required once a day. This cleaning, which usually takes place during the night, clearly separates two days so that the conservation of the set-up state as in the MSC or the PBM model (see Chapter 7.2.2 and 7.2.3) is not necessary. Products that can be proc-essed within the same block do not need to be based on the same intermediate product, since besides the raw material it is also the diameter of the intermediate product that determines the assignment of a product to a block (see Fig. 8.2). Hence, an additional index is applied in order to enumerate the different blocks. Again, for the sequence of the products within a block a “natural order” exists which is mainly based on microbiological concerns (weaker microbiological con-tamination before the stronger). Product wastage is also an issue at the packaging step. However, as most of this wastage can be captured, this volume can partly be reintegrated into the process at the chopping and emulsifying step in order to re-place raw meat. As in the yogurt models, the objective function aims at maximiz-ing the contribution margin. Shelf life is again considered by a shelf life dependent pricing component. However, in contrast to the yogurt case study, the shelf life of the final product starts already at the smoking and scalding stage. The cost ele-ments considered in the objective function include the variable cost of the product, the variable costs of the packaging lines and the chambers, and the variable cost of the intermediate products (particularly raw meat).

The planning horizon (see Fig. 8.3) covers one production week from Monday 00:00 to Friday 24:00 (days 7 to 11). In the packaging department, the Sunday be-fore (day 6) and the Saturday after (day 12) the production week can be used as overtime. The demand of final products is again based on forecasts and on already arrived customer orders and comprises the actual production week (days 7-12) and Monday, Tuesday and Wednesday of the following week (days 14-16). Further-more, two types of stock levels with their corresponding shelf life are taken into account. The stock level of intermediate products at the beginning of the planning period (day 6, Sunday 00:00) contains intermediate products that have been pro-duced on Tuesday, Wednesday, Thursday and Friday of the previous week (days 1 to 4). The stock level of finished products comprises only products that are based on intermediates of Tuesday, Wednesday and Thursday of the previous week (days 1-3). Due to the required minimum time in the intermediate storage, the in-termediate products that have been produced on Friday of the previous week (day 4) cannot be packed on the same day. Hence, at the beginning of the planning week there are no final products in the warehouse that are based on intermediate products that have been produced on Friday (day 4, see Fig. 8.3).


Fig. 8.3 Planning horizon for sausage production

In order to be able to start packaging on Monday of the following week (day 14), the corresponding volume of intermediate products has to be produced within the planning horizon in addition. This stock level of intermediate products at the end of the production week is assumed to be the same as the stock level at the be-ginning of the production week in terms of both volume and age. The stock level of final products at the end of the production week does not need to be determined in advance as it can be derived from the model results.

An overview of the major variables of the model is provided in Fig. 8.4. Pre-condition for a scalding program g to be set up at a position k in scalding chamber o is that the corresponding set up variable Tgko takes a value of one. The scalded volume of an intermediate product i at position k in chamber o is described by the variable Viko. The variable Wis summarizes all volumes of an intermediate product i scalded on shelf life day s. The variable Xjpsl determines the volume of a final product j scalded on shelf life day s that is sliced and packed on packaging day pon packaging line l.

Fig. 8.4 Variables of the sausage production model

8.2 Model Formulation 183

Prerequisite is that the corresponding block b is set up on that day p and on that packaging line l (Spbl = 1). The variable Zjpsd links the packed volumes of final product j to an demand element djd which describes the demand of final product jon the demand day d. This demand can also be satisfied using final products from the previous week (Yjsd).

8.2 Model Formulation

Indices

i I intermediate products j J final products o O chambers l L packaging lines s, so S production days (= shelf life days) p P packaging days d D demand days k K positions g G scalding programs b B blocks i I(g) intermediate products based on scalding program gj JB(b) final products based on block bj JI(i) final products based on intermediate product il LJ(j) packaging lines that can process final product jl LB(b) packaging lines that can process block b

Parameters

ici cost of intermediate product i in € per kg oui chamber utilization factor of intermediate product iocapo capacity of chamber o in kg oco cost of chamber o in € per day imb minimum batch size for a scalding program in kg scaltimeg processing time of scalding program gweightlossi loss of weight of intermediate product i per day in the

intermediate warehouse in % of the initial weight estartk earliest possible starting time of a block at position isis inventory of intermediate product i produced on day s in kgidis demand of intermediate product i to be produced on day s (to

start packaging at the beginning of the following week) iscap capacity of inventory for intermediate products fp final position varcj variable cost of final product j in € pro kg revj revenue for selling one kg of final product j in € slj maximum shelf life of final product j in days


crj minimum shelf life of product j required by the customer (as a fraction of maximum shelf life, applied to multiply the shelf life of product j)

luj factor indicating the packaging line utilization of final product jcapl capacity of packaging line l in kg per day cl cost of packaging line l in € per day stb set-up time for block b in days ct daily cleaning time for packaging lines benj maximum additional benefit when meeting the maximum shelf

life of final product j, in € per kg qj maturation time for final product j in days mb minimum batch size of a block in kg dejd demand of final product j of demand day d in kg fixdem demand day up to which the demand must be fully satisfied sjs inventory of final product j produced on day s in kg osl weekend overtime supplement on packaging line lplossj slicing loss of final product j in % of sliced volume rqj share of slicing losses of final product j in % that can be reinte-

grated into the production process ldp start of the last packaging day of the planning week (Saturday) fdp start of the first packaging day of the planning week (Sunday)

Decision Variables

Tgko =1, if scalding program g is set-up at position k in chamber o (0, otherwise)

Spbl =1, if block b is set-up on packaging day p on packaging line l(0, otherwise)

Viko volume of intermediate product i at position k in chamber o in kg Wis volume of intermediate product i produced on day s in kg STARTko start of production at position k in chamber oXjpsl volume of final product j produced on production day s and

packed on packaging day p on packaging line l in kg Zjpsd volume of final product j produced on production day s and

packed on packaging day p in kg that is used to meet the demand of demand day d

Yjsd inventory of final product j produced on production day s in kg that is used to meet the demand of demand day d

Lpbl duration of block b on packaging day p on packaging line lENDpbl end time of block b on packaging day p on packaging line lESTpl start time of packaging line l on packaging day pLFTpl end time of packaging line l on packaging day pSUOl overtime at the beginning of the planning week (Sunday) on

packaging line lSAOl overtime at the end of the planning week (Saturday) on packag-

ing line l


Objective Function

Lllll

Ll Pplplpl

Jj Pp Ss Lljjjpsl

gGg Kk Oo

ogkoIi Kk Oo

iiko

Pp Ss Ll Ii iJIjijjjpsl

jJj Ss Dd

jj

jj

jjsdPp

jpsd

osSUOSAOcESTLFT

varcplossX

scaltimeocTicV

icrqplossX

revslcr

sdslcrbenYZ

)1(

)1

1()(max

)(

(8.1)

As in the MILP models for yogurt production, the objective function aims at maximizing the contribution margin. Regarding revenues, it considers the regular revenues of the sold products and – in analogy to the yogurt models (see Chapter 7.2.1) – a shelf life dependent pricing component. It is again supposed that the manufacturer yields a financial benefit if the products have a longer residual shelf life when being delivered. Moreover, as slicing losses can partly be reintegrated into the process, they must be respected in addition (Xjpsl * plossj * rqj * ici). The considered cost elements of the smoking and scalding step include the variable costs for the raw meat and for the ingredients of the intermediate products (ici) as well as the processing costs of the scalding chambers (oco). In packaging, the vari-able costs of packaging are taken into consideration. These costs do not occur for the slicing losses (plossj). Furthermore, the costs of utilizing the packaging lines in regular and overtime mode are considered. Again, as in the models for yogurt pro-duction, fixed cost elements have been neglected, as there are not relevant on a weekly planning level.

Constraints to be respected are the following.

Set-up and capacity of chambers

ogko

gIi

iiko ocapTouV)(

)/( k K; g G; o O (8.2)

1Gg

gkoT k K; o O (8.3)

According to Equation 8.2, intermediate product i can only be produced in chamber o if the chamber is set-up (Tgko = 1) for the corresponding scalding pro-gram g. In addition, the volume of all intermediate products i that are based on the same scalding program g (i I(g)) and that are produced at the same position k in chamber o must not exceed the available capacity of the chamber (ocapo). Since smaller diameter sausages cannot fill the chamber to the same extent as large-caliber sausages, the fill grade of the chamber by different types of intermediate


products is taken into account by the factor oui. Moreover, a position k of scalding chamber o can only be used to start one scalding program (Equation 8.3).

Sequencing scalding

okggokko TscaltimeSTARTSTART ,1,,1

k K: k > 1; o O; g G

(8.4)

Equation 8.4 ensures that scalding program g may not start at position k before the end of its predecessor k-1 in order to avoid overlapping of scalding programs in a chamber.

Day bounds scalding

1kkok estartSTARTestart k K; o O (8.5)

1kgkogko estartTscaltimeSTART k K: k = fp; o O (8.6)

For the consideration of shelf life, each position k is assigned to a specific pro-duction day by the parameter estartk. Accordingly, the start of a scalding program at the position k must take place on the production day the position is assigned to (Equation 8.5). Since intermediate products can only be produced from Monday to Friday, the last position of the planning horizon (k = fp) must be finished at Friday midnight (= estartfp + 1) at the latest (Equation 8.6).

Production volume scalding

Oo sestartKk

ikois

k

VW:

s S; i I (8.7)

The produced volume of intermediate product i on production day s (Wis) must not exceed the sum of all volumes of that intermediate product i that have been produced at any position that is assigned to this production day s (estartk = s) and in any scalding chamber o.

Warehouse capacity for intermediate products

iscapouspweightlossX

ouisW

spPp Ss Ll Ii iJIj

iijpsl

Ii ssoSs

iisis

: )(

:

1/

/)(

s S

(8.8)

The warehouse capacity (iscap) for intermediate products must not be exceeded on any production day s (Equation 8.8). The inventory level on production day s is determined by first summing up the inventories of the previous week (isis) and the produced volumes of intermediate products of the previous days of the current


planning week (so s). Then, the volume of intermediate products that has al-ready been used for packaging is deducted (Xjpsl, for all packaging days p s). The capacity utilization of different intermediate products i is respected by the factor oui. The weight loss of intermediate products in the warehouse (weightlossi) is also taken into account. Although Equation 8.8 is only valid at the end of each day, it can nevertheless adequately represent the course of a day as all inventory inputs and outputs are relatively equally distributed over the day.

Minimum batch size scalding

gko

gIi

iko TimbV)(

k K; o O (8.9)

A minimum filling level of the chambers is ensured by Equation 8.9. If a scald-ing program g is set-up at position k in chamber o (Tgko > 0), the volumes of the in-termediate products being part of the production batch must meet or exceed the minimum batch size (imb).

Weight losses in intermediate storage and meeting packaging demand

is

iJIj Ll Pp

ijpslisis idspweightlossXisW)(

1/)(

s S; i I

(8.10)

Equation 8.10 constitutes the link between the scalding and the packaging part. For all intermediate products i of production day s, the produced quantities (Wis)and the inventory from the previous week (isis) must cover the demand of the packaging department (Xjpsl) and the demand of intermediate products for the start of packaging at the beginning of the following week (idis). The weight loss of the intermediate products in the warehouse must also be considered in order to pro-vide the accurate quantities of intermediate products for the packaging depart-ment. The weight loss that is mainly due to water evaporation is given in % per day in the warehouse (weightlossi) and multiplied by the days passed between scalding and packaging (p-s). Although in reality, the weight loss follows a more exponential function, a linear modeling approach for the weight losses has been chosen for two reasons. On the one hand, the solvability of the model is increased by choosing a linear approach. On the other hand, scalded sausages stay usually only a couple of days in the warehouse. The losses during these first days can rela-tively well be reflected by a linear function as the water loss rate remains rela-tively stable during these first days. However, in case of fermented sausages (e.g. salami) with longer maturing times (up to weeks and months), a more detailed modeling approach would be required.

Set-up packaging

)(

/bJBj Ss

lpbljjpsl capSluX p P; b B; l LB(b) (8.11)


Similar to Equation 8.2, the final product j can only be packed on packaging day p on packaging line l if the corresponding block b of the final product jJB(b) is set-up on that packaging line and on that day.

Output quantities packaging

pbl

bJBj Ss jl

jjpsl

bpbl Llucap

plossXstS

)(

)1(

p P; b B; l LB(b)

(8.12)

The duration of block b on packaging line l on packaging day p (Lpbl) includes the set-up time of the block b (stb) and the actual production times of all final products j JB(b) that are produced within the block b. As for intermediate prod-ucts in the scalding chambers, the final products also differ with regard to their degree of utilization of the capacity of the packaging lines. This fact is considered by the factor luj. If packaging line l is not set-up for the block b (Spbl = 0), the out-put quantities of all final products of this block become zero (Xjpsl = 0) according to Equation 8.11. Consequently, the duration of this block (Lpbl) becomes zero as well so that the start time and the end time of this block are the same. Since the capacities of the packaging lines (capl) are given in terms of output-kg, the slicing loss (plossj) must be subtracted from the quantities given into the process (Xjpsl).The slicing loss is partly re-integrated in the production process which is respected in the objective function (Equation 8.1).

Sequencing packaging

lbppblpbl ENDLEND ,1, p P; b B: b > 1; l LB(b) (8.13)

Equation 8.13 guarantees that an overlapping of blocks on a packaging line lcannot take place since block b may not start before the end of its predecessor b-1.It is supposed that not only the sequence of final products within a block is fixed, but also the sequence of blocks within a day.

Day bounds packaging

1pENDpbl b B; p P; l LB(b) (8.14)

ctpLEND pblpbl b B; p P; l LB(b) (8.15)

Equations 8.14 and 8.15 assign each block to a specific packaging day p. A block b on packaging line l on packaging day p must be finished before the end of production day p. Furthermore, the beginning of this block (ENDpbl - Lpbl) must take place after the beginning of the corresponding packaging day p. In addition, a general cleaning time (ct) must be respected for all lines, which takes place every day at midnight.


Stock balance final products

Dd

jpsdj

Ll

jpsl ZplossX )1( j J; s S; p P (8.16)

Dd

jsdjs Ys j J; s S (8.17)

For the coverage of the demand (Equation 8.22), final products from stock (sjs), produced in the previous week, or final products that have been produced within the actual planning week (Xjpsl) can be taken into account. Due to the fact that the slicing of the intermediate products takes place within the packaging step, the re-sulting slicing losses (plossj) must be substracted in order to provide accurate quantities. The slicing losses can partly be reintegrated into the production of in-termediate products (see Equation 8.1).

Utilization of packaging lines

pblpl ENDLFT b B; p P; l LB(b) (8.18)

pblpblpl LENDEST b B; p P; l LB(b) (8.19)

plpll ESTLFTSUO l L; p = fdp (8.20)

plpll ESTLFTSAO l L; p = ldp (8.21)

The variables LFTpl (Equation 8.18) and ESTpl (Equation 8.19) determine the finishing time and the start time of packaging line l on packaging day p. The dif-ference (LFTpl - ESTpl) describes the actual run time of the packaging line l on that packaging day p which is used in the objective function (Equation 8.1) in order to determine the cost of the line utilization. The variables SUOl and SAOl in Equa-tions 8.20 and 8.21 determine the overtime required on packaging line l at the be-ginning (Sunday, see Fig. 8.3) and at the end (Saturday) of the planning period.

Meeting demand

jjjj slcrsdSsjsd

Pp slcrsdSsjpsdjd

YZde)1()(:)1()(:

j J; d D: d fixdem(8.22)

jjjj slcrsdSsjsd

Pp slcrsdSsjpsdjd

YZde)1()(:)1()(:

j J; d D: d > fixdem (8.23)

The demand of final product j of demand day d can be covered by volumes produced within the planning week (Zjpsd) or from stock (Yjsd). The customer re-quirements regarding the minimum residual shelf life (crj) have to be respected for both products from stock and products produced within the planning week. For all


days up to the day fixdem, a full satisfaction of the demand is required, reflecting the very competitive retail environment. Fixdem is usually set to Saturday of the planning week. For the remaining days (especially the demand of the following week), the satisfaction of the demand is optional. In urgent cases, these products can be sliced and packed on Sunday of the following planning week.

Minimum batch size packaging

pbl

bJBj Ss

jjpsl SmbplossX)(

)1( b B; p P; l LB(b) (8.24)

Similar to the scalding chambers, a minimum batch size for block b for every packaging day p and on every packaging line l is required in order to justify the set-up times. A block consists of all production volumes of all final products of this block (j JB(b)).

Maturation time for intermediate products

0jpslX j J; p P; s S: s+qj p; l LJ(j) (8.25)

A maturation time for intermediate products in the warehouse (qj) is consid-ered. Products cannot be packed if the maturation time has not passed. Although the maturation time concerns intermediate products, the time is modeled for each product j in order to allow prolonging or shortening the maturation times for spe-cific customers.

Variable domains

1;0gkoT k K; g G; o O (8.26)

0ikoV k K; i I; o O (8.27)

0isW i I; s S (8.28)

ldpSTARTfdp xo1 k K; o O (8.29)

1;0pblS p P; b B; l LB(b) (8.30)

0jpslX j J; p P; s S: s< p; l LJ(j) (8.31)

0jpsdZ j J; p P; s S: s<p; d D: s<d & p<d (8.32)

0jsdY j J; s S; d D: s < d (8.33)


0pblL p P; b B; l LB(b) (8.34)

1pENDp pbl p P; b B; l LB(b) (8.35)

1pESTp lp p P; l L (8.36)

1pLFTp lp p P; l L (8.37)

0, ll SUOSAO l L (8.38)

Two variables (Tgko and Sbpl) are binary; all other variables are continuous. The variables ENDpbl, ESTlp and LFTlp can only assume values of the corresponding packaging day p. As the intermediate products must be produced before being sliced and packed (s < p) and the final products must be sliced and packed before being delivered (p < d), this sequence is guaranteed by the definition of the ac-cording variables.


In order to assess the applicability of the proposed model for a real-life planning problem, first the number of positions for the scalding chambers must be deter-mined. A low number of positions will substantially reduce the size of the model, especially with regard to the number of binary variables of the type Tgko. On the other side, a too low number of positions will also reduce the available capacity. Considering the duration of the scalding programs in the data set (4, 6 and 8 hours), 6 positions per day can be considered as upper bound, as the shortest scalding program (4 hours) can be run 6 times per day at the maximum. If the number of available positions per day is reduced to four, the capacity can still be utilized by 100% by 6- and 8-hour scalding programs; however, if only 4-hour scalding programs are required, the capacity can only be utilized for 16 hours per day, resulting in a capacity shortage of 33%. Some numerical analysis revealed that with 5 positions per day and chamber, high OVs and low MIP-gaps can be ob-tained within relatively short computational times which cannot be topped by a 6-positions-per-day model after one hour. Therefore, all following analysis has been performed using 5 positions per production day. For the numerical validation of the model, the following data set is applied, which represents a mid- to large-size production environment. The resulting model size is depicted in Table 8.1. It should be noted that not all variables are defined for all elements of the indices (e.g. Xjpsl is not defined for p s).


25 positions (5 positions per production day)15 intermediate products3 scalding programs 30 final products 7 blocks 8 chambers 6 packaging lines 5 production days (Monday to Friday) 7 packaging days (Sunday to Saturday) 11 shelf life days (from Wednesday of the previous week to Friday of the planning week) 10 demand days (from Monday to Wednesday of the following week)

Table 8.1 Number and type of variables in sausage production model

Type Symbol Number Sum

Binary Tgko 600

Binary Spbl 98 698 Binary

Continuous Viko 3,000

Continuous Wis 165

Continuous STARTko 200

Continuous Xjpsl 3,360

Continuous Zjpsd 10,920

Continuous Yjsd 2,850

Continuous Lpbl 98

Continuous ENDpbl 98

Continuous ESTpl 42

Continuous LFTpl 42

Continuous SAOl 6

Continuous SUOl 6 20,787 Continuous

In spite of the relatively high number of binary variables (698), the model shows a relatively good solvability. Since the performance of the model regarding solvability, computational times and MIP-gaps can vary depending on the problem instance, the demand has been varied in order to assess the stability of the model (see Table 8.2).

Table 8.2 Performance of sausage production model (5 positions per chamber and day)

Demand

[% of original demand]

t[s] for first solution Objective value [T€] MIP-gap

80% 45 sec 313.4 2,8%

100% 85 sec 390.8 3,0%

120% 280 sec 454.7 2,0%


Fig. 8.5 Scheduling results for sausage production (scalding chamber 1)

In all cases, a result has been obtained after less than 5 minutes. Nonetheless, in none of the three demand scenarios the first obtained solution cannot be improved even after several hours of computational time. Yet, MIP-gaps between 2% to 3% are still satisfactory considering the two-stage planning problem including scald-ing and packaging. The most important results of the model are schedules for the scalding chambers and the packaging lines. Two sample schedules are provided below, one for a scalding chamber (see Fig. 8.5) and one for a packaging line (see Fig. 8.6). The schedule for the scalding chamber indicates which intermediates should be processed at which point in time. Intermediates that require the same scalding program can be processed simultaneously. Due to different chamber utilization coefficients, the volumes to produce do not always add up to the theo-retical maximum capacity of the chamber.

With regard to packaging, the schedule indicates which volume of which prod-uct when to pack. According to the modeling approach, the products are integrated into blocks; within a block, the products follow a specific order. At set-up time is considered at the beginning of each block. In addition, a daily general cleaning time is integrated at the beginning of each day.

Further analysis has again been performed regarding the impact of the shelf life dependent pricing component on the freshness of the products and the production costs per kg of the final product. Like in the analysis of the yogurt production model (see Chapter 7.3.3), a factor of 1 represents shelf life dependent benefits as used in the data set (ca. 3.5% of the total revenues) – a factor of 0.1 means 10% of the initial value, a factor of 10 means 10 times the initial value. As depicted in Fig. 8.7, the freshness of the products - measured as the weighted average of the re-maining shelf life on the total shelf life of the products - increases if the shelf life factor is increased; however only to a limited extent (from 79% to ca. 81.5%). This is mainly due to the fact that the shelf life of the products is determined at the


scalding step; however packaging must also be performed with lower batch sizes. In the yogurt production model, the shelf life is determined at the packaging step. Therefore, the higher shelf life benefits at a factor of 5 or 10 are leveled off against higher scalding and packaging costs, which include not only a higher number of set-ups, but also the use of more expensive machines and of overtime capacity. Hence, the average production costs per kg of final product increased from € 1.27 at a shelf life factor of 0.1 to € 1.35 at a shelf life factor of 10.

Fig. 8.6 Scheduling results for sausage production (packaging line 1)

Fig. 8.7 Variation of the shelf life dependent pricing component (MIP-gap: 2%)

Factor for the shelf life dependent pricing component

Av

era

ge

sh

elf

lif

e [

in %

of

ma

xim

um

sh

elf

lif

e]

Pro

du

cti

on

co

sts

in

€ p

er

kg

of

fin

al

pro

du

ct

Shelf life in % of maximum shelf life Production costs in €

6 7 8 9 10 11 12 13

8.4 Conclusion 195

8.4 Conclusion

In this chapter, an MILP approach has been presented that integrates shelf life into the weekly planning of the production of scalded sausages. The model covers the steps of scalding and packaging as well as the intermediate warehouse in-between. For the scheduling of the scalding chambers, a position-based approach has been chosen while a block-planning approach has been selected for the scheduling of the packaging lines. In the numerical analysis, first the appropriate number of po-sitions to be used in then model has been derived. Then, a further evaluation has been conducted to determine adequate run-times for the model. Sample schedules have been provided for both a scalding chamber and a packaging line. Finally, the impact of the shelf life dependent pricing component has been assessed.

The model has shown to be able to deliver scheduling results with acceptable MIP-gaps for this two-stage planning problem in a relatively short time. By apply-ing a data set of practical relevance, the freshness of the products can now be in-fluenced explicitly in the weekly production planning. However, as the run-time of the model and the MIP-gap depend on the specific data set, the number of posi-tions to use and also the appropriate run-times must first be newly evaluated if ma-jor parameters of the planning problem change (e.g. number and capacity of cham-bers and packaging lines or number and structure of intermediates, blocks and products).

If very large model instances (due to for example a high number of intermedi-ate or final products or a high number of different scalding programs) must be re-solved, a decomposition approach as in the yogurt production example can be promising. However, in that case the model has to be reformulated in order to get a two stage planning procedure. At the beginning, the packaging lines that can process the same range of final products are optimized individually based on a first model formulation. The result of this first model are the required volumes of intermediate products per day. A second model must then summarize all required quantities of intermediate products and must schedule the scalding chambers ac-cordingly. The described procedure will help to reduce computational times. Nonetheless the overall solution quality will probably decrease since no overall optimization is carried out, local optima are obtained instead. Moreover, as the planner in industry will probably not be able to perform the decomposition on its own, the two new models must be developed by OR experts and delivered to the planner, along with clear guidelines on how to use the models.

9 Shelf Life Integration in Poultry Processing


The third case study is concerned with the integration of shelf life into production planning of poultry processing. The production process is given in detail in Chap-ter 3.7. This case study focuses on the fine-cutting and the packaging step (see Fig. 9.1). Most prior steps, from stunning and bleeding to rough-cutting, are usu-ally organized as a flow line so that only limited planning effort is required for these steps. Although the shelf life of the products starts at the stunning and bleed-ing stage, it can only to a minor extent be influenced within these first process steps. The only possibility is to alter the number of animals in the process. How-ever, this number is fixed in the short term, as mid- to long-term supply contracts exist for the living animals.

The freshness of the products can be influenced significantly in the fine-cutting step. At this stage, it is determined which final products are produced on a specific day taking into account actual orders and forecasted demand, set-up costs and available capacities. Packaging usually takes place on the same day as fine-cutting, since fine-cut products are relatively sensitive to microbial spoilage due to their bigger surface.

Fig. 9.1 Focus and product tree of poultry processing model

198 9 Shelf Life Integration in Poultry Processing

The shelf life of the products is integrated into the model in a similar way as in the two prior case studies, by applying a shelf life dependent revenue component in the objective function. However, the integration of shelf life into production planning is even more important in this case study because the shelf life of fresh meat is much shorter than that of yogurt or sausage products. Since overnight pro-duction does not occur, each production batch can easily be attributed to a specific production day.

The problem is formulated as a variant of the so-called cutting-stock problem in which specific cutting-patterns are applied to a base material. The objective of these models is generally to minimize the scrap. An example of this problem type is given in Günther and Tempelmeier (1995). Specifically in the meat industry, Whitaker and Cammell (1990) apply cutting-patterns to animal carcasses in order to maximize the yield. In poultry processing, two kinds of cutting-patterns can be distinguished: On the one hand, a rough-cut cutting-pattern is applied to the “ready-to-eat” bird after the chilling process. This cutting-pattern is almost the same for all birds; as result, rough-cut parts such as breasts or wings are obtained. Therefore, the rough-cut cutting-pattern is not included as a decision variable; the volume of rough-cuts obtained from one bird (differentiated by the gender of the bird) is calculated based on the bird input.

On the other hand, a fine-cut cutting-pattern is applied to the rough-cut prod-ucts (see also Fig. 9.1). For example, a breast can be sold as a whole or cut into schnitzel of different sizes. The effort required to set-up a cutting-pattern can vary considerably, between the simple arrangement and cleaning of tables to the time-consuming calibration and cleaning of a schnitzel-cutting machine. Outcome of a cutting-pattern are the final products as well as co-products (e.g. smaller pieces of meat that are used in sausage production) and by-products (e.g. bones). Hence, the volume to produce per day according to a specific cutting-pattern is the major de-cision variable in the model. At the packaging step, the fine-cuts are packed into different packaging materials and sizes (Packaging 1 and Packaging 2 in Fig. 9.1). The product-specific set-up times required to calibrate and clean the lines are con-sidered as well.

The planning period for production comprises one week from Monday to Fri-day (see Fig. 9.2). Each production day has eight regular working hours and four possible overtime hours per worker. As overnight work does not occur, it is not necessary to conserve the set-up state of a cutting-pattern or a packaging line over midnight as in the MSC and PBM models for yogurt production (see Chapter 7.2.2 and 7.2.3). In addition, the factory is closed on the weekend. Demand data is given for all five days of the planning week (days 4 to 8) as well as for the Monday of the following week (day 11). This day must be included since not all products re-quired to cover the demand of Monday can be produced on that day. In addition, stock levels of rough-cuts and finished products of Friday of the previous week (day 1) are considered. Older stocks are not integrated in the model due to the very high perishability of fresh meat. In contrast to the intermediate products in the sausage production model, at the end of the production week (Friday, day 8) no minimum inventory level for rough-cuts is imposed, as all rough-cuts required to start fine-cutting on Monday can be produced on Monday morning.


Fig. 9.2 Planning horizon for poultry processing

An overview of the major variables of the model is given in Fig. 9.3. The vol-ume of rough-cut r of gender g processed on day a (Urga) is further processed into fine-cuts on fine-cut day p. Prerequisite is the set up of the corresponding cutting pattern c on that day (Scp = 1). The variable Vcpa describes the volume of rough-cuts of age a that is processed according to cutting-pattern c on fine-cut day p. For the subsequent packaging step, which takes place on the same day as fine-cutting due to the high perishability of the products, it is necessary to set up a packaging line for the final product j (Pjp = 1). The variable Xjpa stands for the volume of fi-nal product j packed on production day p that is based on rough-cuts of day a. Fi-nally, the packaged volumes are assigned to demand elements (dejd) by the vari-able Zjdpa. In addition, volumes from stock at the beginning of the planning week can be considered to satisfy the demand (stockprodj)

Fig. 9.3 Variables of the poultry processing model


9.2 Model Formulation

Indices

a A rough-cut days c C cutting-pattern d D demand daysf F fine-cuts g G gender of poultry j J final products p P fine-cut days / packaging days r R rough-cuts c C(r,g) cutting-patterns based on rough cut r and gender gj J(f) final products based on fine-cut f

Parameters

revj revenue of selling one kg of final product j in € slj maximum shelf life of final product j in days crj minimum shelf life of product j required by the customer (as

a fraction of maximum shelf life, applied to multiply the shelf life of product j)

benj maximum additional benefit when meeting the maximum shelf life of product j, in € per kg

mfreezef maximum freezing volume of fine-cut f per day in kg freezerevf revenue for frozen fine-cut f in € per kg ipoultryag input of living bird per gender g on day arcpoultrygr volume of rough-cut r out of one bird of gender g in kg rctofcfc factor indicating how many kg of fine-cut f can be gained out

of one kg of cutting-pattern cB sufficiently large number normcap regular fine-cutting capacity in man-hours per day addcap additional capacity for fine-cutting in man-hours per day packnormcap regular packing capacity in machine-hours per day packaddcap additional packing capacity in machine-hours per day cnormcap cost of regular capacity for fine-cutting in € per man-hour caddcap cost of additional capacity for fine-cutting in € per man-hour cpacknormcap cost of regular packaging capacity in € per machine-hour cpackaddcap cost of additional packaging capacity in € per machine-hour setupc set-up time of cutting-pattern c in man-hours packsetupj packaging set-up time of final product j in machine-hours mbatchc minimum batch size of cutting-pattern c in kg cmachinec machine cost of cutting-pattern c in € per kg cpackj cost of packaging material in € per kg of final product jthroughputc throughput of cutting-pattern c in kg per man-hour packthroughputj packaging line throughput for final product j in kg per hour


dejd demand of final product j of demand day d in kg stockprodj stock of final product j in kg (Friday of the previous week) stockrcrg stock of rough cut r of gender g in kg (Friday of the previous

week)

Decision Variables

Scp =1, if cutting-pattern c is set-up on production day p (0, oth-erwise)

Pjp =1, if final product j is packed on production day p (0, oth-erwise)

Urga volume of rough-cut r of gender g of rough-cut day aVcpa volume of cutting-pattern c based on rough-cuts of day a

produced on production day pWfpa volume of fine-cut f produced on production day p based on

rough-cuts of day aXjpa volume of final product j packed on production day p based

on rough-cuts of day aYfpa volume of fine-cut f based on rough-cuts of day a, frozen on

production day pZjdpa volume of product j based on rough-cuts of day a used to

cover the demand of demand-day d produced on production day p

Normhoursp regular capacity man-hours on production day pAddhoursp additional capacity man-hours on production day pPacknormhoursp regular capacity man-hours on production day pPackaddhoursp additional capacity man-hours on production day p

Objective Function

Jj Pp Aa

jjpa

Cc Pp Aa

ccpa

Pp

pp

Pp

pp

Ff Pp Aa

ffpa

Jj Dd Pp Aa jj

jj

jjjdpa

cpackXthroughputcmachineV

pcpackaddcarsPackaddhouapcpacknormcursPacknormho

caddcapAddhourscnormcapNormhours

freezerevY

slcr

sdslcrbenrevZ

/

)1

1(max

(9.1)

The objective function aims at optimizing the total profit by considering reve-nues and variable costs. The total revenue for the considered planning period comprises the revenue from selling fresh (Zjdpa * revj) and frozen products (Yfpa *

freezerevf) as well as the shelf life dependent revenue component (the composition of the shelf life component is explained in detail in Chapter 7.2.1). Furthermore, four cost elements are integrated as well. First, manpower costs are included


which consist of cost for regular (cnormcap) and additional (caddcap) capacity. Additional capacity is more expensive as regular capacity due to overtime sup-plements for the workers. Secondly, the costs of packaging (manpower and ma-chines) are considered in the same way. Thirdly, machine costs for fine cutting are respected which are given in form of an hourly rate (cmachinec) multiplied with the produced volume (Vcpa) and divided by the throughput per hour of the cutting-pattern (throughputc). Fourthly, the costs for packaging materials of the final products are deducted as well (cpackj).

Constraints to be respected are the following.

Volume of rough-cuts

rgagrga Urcpoultryipoultry r R; g G; a A: a>1 (9.2)

rgarg Ustockrc r R; g G; a A: a=1 (9.3)

rga

grCc Pp

cpa UV),(

r R; g G; a A (9.4)

The volume of available rough-cuts is determined by multiplying the number of birds per day with a factor (rcpoultrygr) that indicates how many kg of a specific rough-cut (e.g. breast) can be gathered out of an average bird. As the weight of birds depends significantly on its gender, the gender is considered by applying the index g (Equation 9.2). The poultry input and poultry-to-rough-cut factor are given as data (not as decision variables). The poultry-to-rough-cut factor can vary depending on the season and on the grower; nevertheless, for this production planning problem long-term averages are applied as the correct weights can only be known after the processing. The index a refers to the day on which the “ready-to-cook” bird is cut into the different rough cuts. Therefore, this index is decisive for the calculation of the product shelf life. With regard to the data set, it is impor-tant to notice that the real shelf life starts one day earlier as the birds are slaugh-tered the day before. Due to the chilling process that takes one day, rough-cutting cannot take place at the same day as slaughtering. The remaining rough cuts of Friday of the previous week are integrated as stocks (Equation 9.3).

The available rough cuts are then further processed or “fine-cut” into the differ-ent final products. For the fine-cutting of a rough-cut, a cutting-pattern is applied. A cutting-pattern is always specific for a specific rough-cut and a specific gender (c C(r,g); e.g. cutting the breast of a male bird into cutlets). The differentiation of the cutting-patterns by gender is necessary as the composition of the rough-cut parts differs by gender (e.g. different relations between meat and bones). The de-cision variable Vcpa determines the volume of a rough-cut of age a that is fine-cut according to cutting-pattern c on production day p (Equation 9.4). The decision variable Vcpa is only defined for production days p which are higher than or equal to age a of a rough-cut (Equation 9.20). As rough-cutting is a continuous process,


it is possible to rough-cut a ready-to-cook bird and to fine-cut these rough-cuts on the same day.

Volume of products

fpafc

Cc

cpa WrctofcV )( f F; p P; a A (9.5)

fpa

fJj

jpafpa YXW)(

f F; p P; a A (9.6)

Dd

jdpajpa ZX j J; p P; a A: a>1 (9.7)

Dd Pp

jdpaj

Pp

jpa ZstockprodX j J; a A: a=1 (9.8)

The volume produced according to a specific cutting-pattern (Vcpa) has now to be transformed into fine-cuts. The factor rctofcfc indicates how many kg of a fine-cut can be gathered out of one kg of rough-cut if processed according to cutting-pattern c (Equation 9.5). This fine-cut volume can either be packed into different products based on this fine-cut (j J(f )) or it can be frozen (Yfpa). The packed fi-nal products can then be used to cover the demand of fresh meat of different de-mand days (Equation 9.7). The variable Zjdpa indicates the volume of a final prod-uct j, fine-cut and packed on a production day p and based on rough-cuts of a day a, that is used to cover the demand of a demand day d. Equation 9.8 integrates the initial stock of each product at the beginning of the planning week. These products have been produced on Friday of the previous week (a = 1) and can be used in ad-dition to cover the demand of the following week.

Meeting demand

Pp slcradAa

jdpajd

jj

Zde)1()(:

j J; d D(9.9)

The demand of a final product on a specific day is the maximum volume of this product that can be sold on that day. In contrast to the yogurt and sausage models, a full demand satisfaction is not required as the poultry input is not always suffi-cient to cover the entire demand, resulting in unfeasible model instances. To cover the demand of a final product, all products can be considered that fulfil the cus-tomer shelf life requirements (d-a (1-crj)* slj).

Maximum freezing volume

Aa

fpaf Ymfreeze f F; p P (9.10)


The volume of a fine-cut that can be frozen per day is subject to a limitation, as freezing is a sub-optimal possibility due to the fact that fresh products yield sig-nificantly higher revenues. It is mainly an option for a product surplus that cannot be sold as fresh.

Set-up of cutting-pattern

BSV cp

Aa

cpa c C; p P (9.11)

If fine-cutting of rough-cuts according to cutting-pattern c takes place on pro-duction day p, the cutting-pattern must be set-up on that day (Scp = 1). If rough-cuts of different ages are processed according to the same cutting-pattern on the same day, the cutting-pattern must only be set-up once. As in Equation 7.2, the large number B has the function to allow the production of products according to cutting-pattern c if this cutting-pattern is set-up on production day p. Therefore, this number should be as high as the largest possible output of a cutting pattern per day. Higher numbers, however, should be avoided in order not to impair the com-putational performance (see Chapter 7.2.1).

Set-up of packaging

BPX jp

Aa

jpa j J; p P (9.12)

As for the set-up of cutting-patterns, a set-up variable for the packaging lines (Pjp) is introduced that takes the value of 1 if the corresponding product is packed on that day. In analogy to Equation 9.11, the highest possible output of a final product per day is the lower bound for the large number B.

Minimum batch size for cutting-pattern

cpc

Aa

cpa SmbatchV c C; p P (9.13)

A cutting-pattern dependent minimum batch size (mbatchc) is required for a cutting-pattern to be set-up. The set-up of some cutting-patterns such as the set-up of a schnitzel-cutting machine causes relatively high costs. The requirement of a minimum batch size ensures that a sufficiently high volume is processed to justify these costs. For packaging, no extra minimum batch size is required as a final pro-duct is based on one or several fine cuts, which are related to one or several cut-ting-patterns. Therefore, the minimum batch size in fine-cutting leads in almost all cases to the same or even a higher minimum batch size in packaging.

Fine-cutting capacity

normcapNormhoursp p P (9.14)


addcapnormcapAddhoursNormhours pp p P (9.15)

pp

Cc

ccp

Cc Aa

ccpa AddhoursNormhourssetupSthroughputV

p P

(9.16)

The fine-cutting capacities concern only manpower; machine capacities are not considered as they do not constitute a major bottleneck in the production. The de-cision variables Normhoursp and Addhoursp indicate how many man-hours are re-quired on production day p to process the rough-cuts into fine-cuts. The decision variable Normhoursp includes the regular working time; Addhoursp determines the number of overtime hours required. Both Normhoursp and Addhoursp are re-stricted in terms of capacity (Equations 9.14 and 9.15). In Equation 9.16, the ca-pacity usage (Normhoursp and Addhoursp) is related to the volume produced and the time required to set-up the cutting-patterns. In the first term, the produced vol-ume of this cutting-pattern is divided by the throughput of the cutting-pattern per man-hour, which results in the number of man-hours required to produce the vol-ume. In the second term, setupc is the number of man-hours required to set-up a cutting-pattern. This number is multiplied by the binary decision variable Scp,which indicates if cutting-pattern c is set-up on production day p.

Packaging capacity

ppacknormcaursPacknormho p p P (9.17)

packaddcapppacknormcarsPackaddhouursPacknormho pp

p P (9.18)

pp

Jj

jjp

Jj Aa

jjpa

rsPackaddhouursPacknormho

packsetupPhputpackthrougX

p P

(9.19)

The capacity restrictions in packaging are similar to the restrictions in fine-cutting. Again, regular and additional capacities are available. The packed volume of a product (Xjpa) divided by the throughput of the lines per hour (packthrough-

putj) determines the machine-hours required for packaging the products. In addi-tion, a set-up time has to be considered for each product (packsetupj), which con-siders the calibration of the packaging line at the beginning of a batch and the cleaning procedures at the end.


Variable domains

1;0cpS c C; p P (9.20)

1;0jpP j J; p P (9.21)

0rgaU r R; g G; a A (9.22)

0cpaV c C; p P; a A: p a (9.23)

0fpaW f F; p P; a A: p a (9.24)

0jpaX j J; p P; a A: p a (9.25)

0fpaY f F; p P; a A: p a (9.26)

0jdpaZ j J; d D; p P: d p; a A: d a & p a (9.27)

0,,, pppp rsPackaddhouursPacknormhoAddhoursNormhours p P (9.28)

In contrast to the production of sausages, processing and packaging can take place on the same day due to the relatively short shelf life of fresh meat. Only for a > p, the corresponding variables are not defined (Equations 9.23-9.27). Fur-thermore, the packed products can be delivered to the retail outlet on the same day as packaging takes place (Equations 9.27), since the production starts early in the morning and the products are usually delivered directly to the outlet (not via DC).


For the numerical validation of the model, the following data set is applied:

6 rough-cuts, 27 fine-cuts, 40 final products, 28 cutting-patterns, 5 production days (Monday to Friday), 8 demand days (from Monday to Monday of the following week), 8 rough-cut days (from Friday of the previous week to Friday of the planning week).


Table 9.1 Number and type of variables in poultry processing model

Type Symbol Number Sum

Binary Scp 140

Binary Pjp 200 340 Binary Continuous Urga 96

Continuous Vcpa 840

Continuous Wfpa 810

Continuous Xjpa 1,200

Continuous Yjpa 810

Continuous Zjdpa 6,800

Continuous Normhoursp 5

Continuous Addhoursp 5

Continuous Packnormhoursp 5

Continuous Packaddhoursp 5 10,576 Continuous

The resulting model size is given in Table 9.1. Again, it should be noted that not all variables are defined for all elements of an index (e.g. Zjdpa is not defined for a>p). For the interpretability of the objective function that aims at optimizing the contribution margin, the following term is added:

Gg Aa

agg ipoultrycpoultry

The parameter cpoultryg stands for the cost of a living bird of gender g in €. The costs of the living bird multiplied by the number of living birds determine the total costs of living birds per week. As both the costs and the number of the birds are fixed in the model (see Chapter 9.1), they are not a result of the calculation. Nonetheless, these costs must of course be respected in order to determine the contribution margin per week. The computation of the model using ILOG-OPL Studio 3.6.1 (on a AMD XP 2600+ CPU with 1 GB RAM) delivers a near-optimal solution for the objective function in a relatively short time, which is due to the low number of binary variables (340).

The analysis of the model performance reveals that an adequate run-time for the model is ca. eight minutes, when the MIP-gap falls below 2% (see Fig. 9.4). Higher run-times improve the OV only marginally. A run-time of eight minutes is acceptable considering the fact that the model is generally applied only once a week in order to generate the plan for the following week. Most important model results include which cutting-pattern to set-up on which day to fine-cut which vol-ume. For the packaging lines, information is provided on which final product to set-up on which day and which volume of each product to pack.

As the performance of the model in terms of run-times and MIP-gaps depends on the data set used, the demand has been varied as in the sausage production model in order to study the stability of the model in case of changing parameters.


Fig. 9.4 Performance of the poultry processing model

In addition to the base case, two other instances have been analyzed in which the demand, the poultry input and the initial stock of fine-cuts have been multi-plied by 0.8 and 1.2. For all three analyzed model instances, the model shows a similar performance. An MIP-gap below 5% can be obtained at around 125 sec-onds (see Table 9.2). After 30 minutes, all MIP-gaps have fallen below 3%, for the base case even down to 1.47%.

As for the MILP-models for yogurt and sausage production, a numerical analy-sis was performed on the influence of the shelf life depended pricing component. As in the two prior case studies (see Chapters 7.3.3 and 8.3), the factor of one (see Fig. 9.5) represents the pricing component as used in the model (ca. 2.6% of the total revenues); a factor of 0.1 represents a pricing component of only 10% of the initial value and a factor of 10 stands for a model instance in which the initial pric-ing component is multiplied by 10.

Table 9.2 Performance of poultry model

Demand

[% of original demand]

MIP-gap < 10% MIP-gap < 5% MIP-gap

after 1,800 sec

80% 118.09 sec 126.36 sec 2.79%

100% 49.33 sec 122.42 sec 1.47%

120% 124.25 sec 127.01 sec 2.89%

The analysis reveals that starting at a factor of 2, the shelf life depended pricing component has a significant influence on the number of set-up operations per-formed and, hence, on the resulting shelf life. The average shelf life of the consid-ered products increased from 67.5% (factor 2) to 78.0% (factor 5) of the maxi-mum shelf life. A larger factor leads to only minor improvements with regard to

135

136

137

138

139

140

141

142

10 100 1000 10000

Computational time [s]

Ob

jecti

ve v

alu

e [

T€]

0%

1%

2%

3%

4%

5%

6%

7%

MIP

-gap

Objective value MIP-gap

9.4 Conclusion 209

the average shelf life. Therefore, in order to encourage the production of fresher products, the shelf life dependent pricing component must constitute a significant part of the total revenue of the manufacturer. On the other hand, the production of fresher products leads to higher costs in production, mainly due to higher set-up costs because of smaller batch sizes and a more intensive use of overtime. While the production costs per kg of final product are around 34 cent for shelf life factors up to 2, these costs increase with higher shelf life factors, to over 36.5 cent per kg at a shelf life factor of 10.

Fig. 9.5 Variation of the shelf life dependent pricing component (run-time: 5 minutes)

9.4 Conclusion

The integration of shelf life into production planning for poultry processing is less difficult than in the two prior case studies. However, compared to yogurt or sau-sage production, integrating shelf life in fresh meat processing is even more im-portant because in that case shelf life is much shorter. The presented model pro-vides a relatively simple, but well performing algorithm, which takes the shelf life of the products into account for the weekly production planning. Results of the model are volumes to be fine-cut and packed per day.

As for the two prior case studies, run-times and MIP-gaps depend on the data set applied. Therefore, in case of major changes of the problem structure (e.g. cut-ting-patterns or capacities), the appropriate run-time should be re-evaluated with the new data set by the planners. In order to perform this task, intensive training has to be provided to the planner (e.g. guidelines to evaluate the model runtimes). In addition, the planner must be trained on how to handle complex model in-

60,0%

62,0%

64,0%

66,0%

68,0%

70,0%

72,0%

74,0%

76,0%

78,0%

80,0%

0,1 1 10

Factor for the shelf life dependent pricing component

Av

era

ge

sh

elf

lif

e [

in

% o

f m

ax

imu

m s

he

lf lif

e]

33,5

34,0

34,5

35,0

35,5

36,0

36,5

37,0

Pro

du

cti

on

co

st

of

fin

al p

rod

uc

t [i

n c

en

t p

er

kg

]

Shelf life in % of maximum shelf life Production cost per kg of final product


stances (e.g. decomposition of the problem by product type or production step as proposed for the sausage production model).

In order to complete the short-term planning of poultry processing, a further model is required for the subsequent sequencing of cutting-patterns and batches in packaging within a day. This model for daily production planning should take the resulting volumes per day as input factors and should integrate other short-term constraints as well (e.g. machine availability etc.). However, the simultaneous de-termination of batches in fine-cutting / packaging for a week and their sequencing within a day is a too complex planning task to be performed within one model.

10 Conclusions and Recommendations

10.1 Summary of Results

Fresh food industries constitute a major pillar of the German economy with a total turnover of ca. € 50 bn. in 2000. These industries produce primarily products with a short shelf life, the most important single industries being the dairy, the bakery, the processed meat and the fresh meat industries. All these industries are facing strong pressure due to high costs and low margins for various reasons. The biggest threat results from the continued consolidation of the retailers. Further, the rise of the discount channel and the emergence of private labels accelerate this develop-ment. Finally, technological, social/legal and environmental requirements (e.g. RFID, product proliferation or product traceability) will require significant in-vestments in the years to come. In addition, the fresh food production systems are rather complex, due to high variations in the qualities of raw materials, the per-ishability of the products or the extensive product variety. In this context, APS systems offer support in responding to these requirements. APS systems are deci-sion support tools able to compute plans and schedules for multiple variables and constraints simultaneously. They can be optimized with respect to specific criteria. Although the implementation of an APS system generally provides a quick ROI, the number of implementations in process industries and specifically in fresh food industries has remained rather low. Hence, the first research question aims at evaluating the functions an APS system should offer in order to be successfully implemented in fresh food industries:


Which requirements must APS systems cover in order to efficiently and effectively

support production planning in fresh food industries?

In Chapter 2, an overview of the current status of APS systems has been pro-vided in order to understand which kind of support these systems can offer to fresh food industries. The functions of each of the software modules have been evalu-ated in detail, including the planning logic, the most important input and output data and the algorithms applied to solve the planning problems. In addition, the market for APS systems has been characterized, a recommendation for a struc-tured implementation has been proposed and the benefits and shortcomings of cur-rent APS implementations have been assessed.

Following this analysis of the possible level of support of APS systems for fresh food industries, Chapter 3 has looked at the characteristics of these indus-

212 10 Conclusions and Recommendations

tries. After the term “fresh food industries” has been defined and the most impor-tant segments have been presented, the supply chain and production system specif-ics of these industries have been evaluated. Afterwards, the market environment, the supply chain and the production processes of three case study products (yo-gurt, sausages and fresh poultry) have been reviewed in detail. The case studies cover three of the four most important fresh food industries (dairy, processed meat and fresh meat) and have been used throughout this work. These case study prod-ucts have been chosen for two reasons. On the one hand, they cover important sub-segments in terms of revenues and expected growth. On the other hand, these products represent the most challenging production environment within each in-dustry (e.g. high product variety in yogurt production; two-stage production envi-ronment, shrinkage losses and slicing losses in sausage production; the use of rough-cuts and fine-cuts in poultry processing). By first modeling the most com-plicated production environments, most probably the results can later more easily be transferred to other products and industries.

In Chapter 4, a list of 119 specific requirements of fresh food industries regard-ing APS systems has been developed and their respective importance assessed for each of the case study industries. The requirements are based on the characteristics of fresh food industries in general and of the three case study industries in particu-lar. They have been structured around the modules that are usually integrated into APS systems. Most of the requirements concern the DP, the DisP and the PP/PS modules. DP and DisP are of particular importance to fresh food industries due to the MTS production environment, the high perishability of the products and the relative importance of distribution costs. The length of the list of requirements for the PP/PS modules can be attributed to the high diversity of the production sys-tems that differ considerably between industries and even between different com-panies of the same industry. Several requirements are important for almost all fresh food industries and concern several modules. Examples are the support of the collaboration with SC partners or the management of natural products that show high variations in quality, process performance and yields. Probably the most distinguishing factors of fresh food industries is the short shelf life of the products, hence the integration of shelf life becomes a decisive factor for the se-lection of an APS system. Consequently, the second section of the dissertation (Chapters 5 to Chapter 9) has dealt with the integration of shelf life:


How can shelf life be integrated into production planning? How can production

planning contribute to optimizing shelf life output?

In order to answer research question 2, first the role of shelf life in fresh food industries has been assessed in Chapter 5. At the beginning, a definition for shelf life has been provided, and the most important limiting factors for shelf life have been analyzed. Then it has been examined how the shelf life of a product can be determined and which technological options exist to extend shelf life, followed by a detailed analysis of the shelf life of the products considered in the case studies.

10.2 Discussion 213

Finally, an overview has been given on the consideration of shelf life in the OR-related literature.

Chapter 6 has looked at the level of integration of shelf life functions in leading production planning systems (SAP-APO, the Supply Chain Planning suite of Peo-pleSoft’s EnterpriseOne software package and the CSB system). The structure and modules of each system have been portrayed shortly, followed by a description of the offered shelf life support. Several improvement potentials have been high-lighted such as the insufficient support of shelf life propagation, the poor coverage of shelf life in other modules than PP/PS, the lacking optimization of the shelf life output or the missing possibility to prioritize customer orders with regard to shelf life.

In order to close some of these gaps, in Chapters 7 to 9 it has been demon-strated by means of MILP models how shelf life can be integrated into the weekly production planning for the three case study products yogurt, sausages and fresh poultry. For the production of yogurt (Chapter 7), three distinct models have been proposed, all of which focus on the flavoring and packaging step where the actual shelf life of the products is determined. All three yogurt models are based on a combination of a discrete and a continuous representation of time and rely on the principle of block planning. Batch sizing and scheduling of numerous recipes and products on several packaging lines are considered in the models. In order to in-crease the solvability of the models, the sequence of the blocks is fixed in the first two models, whereas the third proposed model applies a position-based approach with a free sequence of recipes. Overnight production and, hence, the necessity for identifying two different shelf life values for the same batch is included in two of the three model formulations. In contrast to the yogurt production model, the model for the production of scalded sausages (Chapter 8) integrates a processing step and the intermediate warehouse in addition to the packaging step, as the shelf life of the products is already determined at the processing stage. While the mod-eling of the processing step relies on a position-based approach, the packaging lines are modeled using the block planning principle. Again, a discrete and a con-tinuous time representation are used simultaneously. In addition, time-dependent shrinkage losses in the intermediate warehouse are considered as well. For the poultry processing model presented in Chapter 9, a two-stage cutting-stock ap-proach has been chosen. Special attention has been paid to the fine-cutting and packaging steps, as the shelf life of the product can be significantly influenced in fine-cutting where the rough-cut parts of the bird are further cut up into fine-cuts by means of specific cutting-patterns.

10.2 Discussion

Both the list of requirements of fresh food industries regarding APS systems and the models developed for the three sample industries offer several advantages. With respect to the list of requirements, its benefit is threefold. First of all, fresh food manufacturers of yogurt, sausages or poultry products can use the list to


evaluate the capabilities of APS software packages and to decide on which soft-ware to implement. Second, other fresh food manufacturers can now derive more

easily their own list of requirements. Many of the requirements can be transferred without difficulty to other fresh food industries and also to other food industries. Finally, APS providers can analyze the suitability of their systems for fresh food

industries and close existing gaps by developing new functions. In particular the support of shelf life functions is rather limited, although it is a decisive buying cri-terion for fresh food producers.

The models developed in Chapters 7 to 9 aim at closing these gaps. All pre-sented models have shown to be suitable to generate near-optimal solutions within reasonable computational time for planning and scheduling problems in realistic dimensions. Compared to the shelf life functions provided by current APS sys-tems, progress has been made in particular with respect to four major issues. First, the models enable the propagation of the shelf life of raw materials and interme-diate products to the shelf life of the final products. For example, in the sausage production model the shelf of the intermediate products determines the shelf life of the final products. In the poultry processing model, the shelf life of the rough-cuts is even propagated over two production steps, from rough-cutting over fine-cutting to packaging. Secondly, the assignment of two different shelf lives to one production batch is possible in two of the three models for yogurt production (in case of overnight production). Thirdly, the prioritization of products with regard to their shelf life is considered in all models by means of the shelf life dependent pricing component. By adjusting the benefit parameter for shelf life, the impor-tance of the shelf life of different products can be weighted. Finally, all model formulations reward the production of fresher products by applying the shelf life dependent pricing component. As the product shelf life has been considered ex-plicitly, the proposed planning tools can be applied to improve the freshness of the products and to assess the additional costs that arise.

In addition to the described benefits on the short-term planning level, the mod-els can also be used to support mid- to long-term planning decisions. In that case, the parameters which are fixed in the short-term model can be varied in order to study the effect of changes in the production environment. A prominent example is the evaluation of the impact of a new production technology. The impact of the new technology in terms of costs and capacities must be reflected by adjusting the corresponding parameters in the models. Then, the models allow to assess the in-fluence of the new technology on the production process, the freshness of the products and the resulting cost structure. In this context, different scenarios can easily be evaluated and compared.

Nonetheless, both the list of requirements and the models for the case study in-dustries are subject to a number of limitations, which must be taken into consid-eration when using the list and the models in an industrial environment. As far as the list of requirements is concerned, it is certainly not possible for a fresh food manufacturer to use the list without prior modifications. The requirements of dif-ferent companies of the same industry and even between sites of the same com-pany can vary considerably. Hence, it might become necessary to change the scores in the list or to add new requirements to it. Nevertheless, the list can be

10.3 Recommendations for Further Research 215

used as starting point for the identification of further requirements. In addition, it can be used to check whether the most important planning aspects have been cov-ered.

For the implementation of the proposed models, several challenges have to be mastered. On the one hand, the APS providers must integrate the presented algo-rithms into their systems. Particularly, the shelf life dependent pricing component

needs to be added as a parameter. Additionally, the production of fresher products needs to become a part of the objective function within the optimization. How-ever, when looking at the current coverage of shelf of life functions in APS sys-tems, the first priority of the providers should be to enlarge the basic shelf life functions such as the provision of shelf life in other than PP/PS modules or the propagation of shelf life. The integration of the shelf life dependent pricing com-ponent and the reward of fresher products within the objective functions will then be implemented based on specific customer requests.

Furthermore, the integration of shelf life integrated planning into existing sup-ply chain management concepts is another important challenge. Developing suit-able incentives for the production of fresher products based on customer satisfac-tion is therefore necessary. In particular, a shelf life dependent pricing component must be added to the existing terms and conditions systems, which may become difficult due to the still rather competitive relationships between manufacturers and retailers. Its value needs to be determined so that the costs of producing fresher products are fairly shared between both partners. A coordination mecha-nism has to be developed in order to achieve a win-win situation between the manufacturer and the retailer. In that case, the increased product freshness will provide an additional quality-oriented feature for markets characterized by intense competition, which can constitute a pivotal competitive advantage for both the manufacturer and the retailer.

10.3 Recommendations for Further Research

Both the list of requirements and the proposed concepts for shelf life integrated planning offer possibilities for further research in this field. Regarding the list of requirements of fresh food industries regarding APS systems, a first recommenda-tion for further research is the enlargement of the list into several directions. At the beginning, other products of the case study industries should be covered (e.g. fresh milk in the dairy industry, fermented sausages in the processed meat industry or pork and beef processing). Then, the remaining fresh food industries should be integrated (e.g. bakery products, fish, or fruits and vegetables). Finally, similar lists should be developed for other food industries as well (e.g. sugar and sweets, oils and fats, beverages etc.). A second direction for further research is the evalua-

tion of APS systems based on the proposed list of requirements. This evaluation will particularly be useful for fresh food manufacturers when assessing the per-formance of different APS systems. In addition, APS providers can use the list to evaluate their systems and to close existing gaps by adding fresh food specific


functions. Finally, the evaluation and assessment of the different systems will most likely reveal further improvement potentials of APS systems. Special atten-tion should be given to the management of variable raw materials, intermediate and final products as well as to variable processing times and yields as these issues are relatively important in fresh food industries. Other functions that should be looked at include the collaboration components, the forecasting algorithms of the DP module, the management of promotions throughout all modules, or all func-tions concerned with product traceability. Similar to the models developed for shelf life integrated planning, additional models should then be developed that remedy the analyzed weaknesses of the systems.

With respect to the shelf life integrated planning concept, the transferability of

the presented models to similar problems concerning the production of other fresh foods has to be examined. Again, as a first step the remaining products of the case study industries should be covered, followed by the remaining fresh food indus-tries such as fish, fruits, vegetables and bakery goods. In particular bakery goods should quickly be covered, by transferring the yogurt and sausage models. An-other direction for further research is the implementation of the proposed models

within selected APS systems. For this purpose, the shelf life dependent pricing component needs to be added to a terms and conditions system between a manu-facturer and a retailer and its value needs to be determined. Therefore, it is neces-sary to assess the benefits of fresher products and the arising costs and to agree on a costs/benefits sharing model. From a technical point of view, some data fields must be added to the APS system, for example the shelf life dependent pricing component that is not included in any system at the moment. Then, the corre-sponding constraints must be modeled in the system and finally, the optimization engines must be adapted in order to integrate the freshness of products within the objective function. Once implemented, the application of the shelf life integrated planning concept will result in optimized product freshness and a higher customer satisfaction.

References

Abad PL (2003) Optimal pricing and lot-sizing under conditions of perishability, finite pro-duction and partial backordering and lost sale. European Journal of Operational Re-search 144: pp 677-685

Allison J (2004) Procurement in the Food and Drink Industry in the early 21st Century. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 49-61

AMR Research (2001) The 2000 Supply Chain Management Market. The AMR Market Analytix Series, Report of AMR Research

AMR Research (2003) The Supply Chain Management Applications Report 2002-2007. Report of AMR Research

Andrae W (1996) Verpacken von Fleisch und Fleischerzeugnissen. In: Sielaff, H. (ed) Fleischtechnologie. B. Behr's Verlag, Hamburg, pp 511-558

ARC Advisory Group (2000) Supply Chain Collaboration Strategies for Food & Beverage Companies. ARC Strategies - E-Business & Automation Strategies for Industry Execu-tives. Report of ARC Advisory Group, Dedham

Arminger G (2002) Absatz- und Bestellprognosen für den Handel. In: Seifert D (ed) Col-laborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 89-102

Arnold U, Warzog F (2001) Supply Chain Management - Konzeptabgrenzung und branchenspezifische Spezifizierung. In: Arnold U, Mayer R, Urban G (eds) Supply Chain Management - Unternehmensübergreifende Prozesse Kollaboration IT-Standards, Lemmens, Bonn, pp 13-47

Auer J (2001) BSE-Krise verschärft Strukturprobleme. Die Auswirkungen für den Markt - Landwirtschaft, Fleischindustrie, Metzger und Lebensmittelhandel unter der Lupe. Le-bensmittel Zeitung 10: 68

Baker RC, Bruce CA (1989) Further Processing of Poultry. In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 251-282

Barberio GG (1986) Problems of Analysis of Products After Ageing - A Review of the In-strumental Analysis of Food Related Products and the Problems Encountered. In: Charalambous G (ed) The Shelf Life of Foods and Beverages - Proceedings of the 4th International Flavor Conference. Rhodes, Greece, 23-26 July 1985, Elsevier Science, New York et al., pp 7-42

Bartsch H, Bickenbach P (2002) Supply Chain Management mit SAP APO, 2nd edn. Gali-leo, Bonn

Bastock B, Baumann F, Smith S (2002) Der Einsatz von CPFR bei Ace Hardware und Manco. In: Seifert, D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 175-186

Baumgarten H, Darkow I-L (2002) Konzepte im Supply Chain Management. In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Management - Theorie und Praxis ef-

218 References

fektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 89-108

Baumgarten H, Thoms J (2002) Trends und Strategien in der Logistik - Supply Chains im Wandel. Bundesvereinigung Logistik und der Bereich Logistik der Technischen Uni-versität Berlin, Berlin

Beamon, BM (1998) Supply chain design and analysis: Models and methods, International Journal of Production Economics 55: 281-294

Beierlein JG, Miller CA (2000) Performance Measures, and Measurement in Supply Chains in the Food System. http://12.237.226.141:81/conferences/2000Congress/2000_forum

_papers.htm [15.01.2003] Bell RG (2001) Meat Packaging: Protection, Preservation, and Presentation. In: Hui YH,

Nip W-K, Rogers RW, Young OA (eds) Meat Science and Applications. Marcel Dek-ker, New York et al., pp 463-490

Biennier F, Favrel L (1998) Robust Production Control. Industrie Journal Européen des Systèmes Automatisées 32 (4): 502-530

Benninger S, Grandjot H-H (2001) Supply Chain Revolution durch E-Commerce. Deutscher Verkehrs-Verlag, Hamburg

Berger I (1999) Optimizing the Supply Chain with APS. http://www.e-optimization.com/

resources/uploads/optimizeAPS.pdf [09.04.2003] pp 24-27 Bergmann R, Rawlings C (1998) Transport Management - Future directions: redefining the

role of transport. In: Gattorna J, Jones T, Danks A, Dhillon Y, Holdforth L (eds) Stra-tegic Supply Chain Alignment - Best practice in supply chain management. Gower, Aldershot et al., pp 369-380

Bernsmann A, Bone T (2003) Die Bahn ist keine echte Alternative. Lebensmittel Zeitung 42: 72

Bhambri A (N.N.) CPFR: Will it fly in the food industry? http://www.internet capi-

tal.com/network/progress/pdf/agribuys-ArvindCPFR.pdf [05.02.2004] Bhushan N, Gummaraju K (N.N.) A petri net based simulation approach for evaluating

benefits of time temperature indicator and wireless technologies in perishable goods retail management. Infosys Technologies, http://www.infy.com [11.03.2003]

Bick M (2002) Einführungskonzept für Supply Chain Management Software am Beispiel von SAP APO. In: Busch, A, Dangelmaier, W (eds) Integriertes Supply Chain Man-agement - Theorie und Praxis effektiver unternehmensübergreifender Geschäfts-prozesse. Gabler, Wiesbaden, pp 235-260

Biehl B (2004) Logistik ist Sache des Handels. Lebensmittel Zeitung 27: 34 Bixby R (2002) Solving real-world linear programs: a decade and more of progress. Opera-

tions Research 50 (1): 3-15 Bjorksten A, Knopf M (1999) Enterprise Software: The Front and Back Office: Overview

and Emerging Trends. http://cci.bus.utexas.edu/research/white/ent_soft.htm

[17.12.2003] Bloemhof-Ruwaard JM, van Beek P, Hordijk L, van Wassenhove LN (1995) Interactions

between operational research and environmental management. European Journal of Operational Research 85: 229-243

Bloemhof-Ruwaard JM, van Nunen JAEE, Vroom J, van der Linden A, Kraal A (2002) One and Two Way Packaging in the Dairy Sector. In: Klose A, Speranza MG, van Wassenhove LN (eds) Quantitative Approaches to Distribution Logistics and Supply Chain Management. Springer, Berlin et al., pp 115-127

Blömer F (1999) Produktionsplanung und -steuerung in der chemischen Industrie – Res-sourceneinsatzplanung von Batchprozessen auf Mehrzweckanlagen. Gabler, Wies-baden

References 219

Blömer F, Günther H-O (1997) OR-Ansätze für die Produktionsplanung in der Prozessin-dustrie. Industrie Management 13 (4): 51-55

Bolton J (1998) Effective demand management - Are you limiting the performance of your own supply chain? In: Gattorna J, Jones T, Danks A, Dhillon Y, Holdforth L (eds) Strategic Supply Chain Alignment - Best practice in supply chain management. Gower, Aldershot et al., pp 138-156

Bothe M (1999) Supply Chain Management mit SAP APO - Erste Projekterfahrungen. HMD Praxis der Wirtschaftsinformatik 207: 70-77

Bothe M (2000) Aus Erfahrung gut - Die Gunst des "Spätentwickelnden": Praxis-erfahrungen bei der Integration des SCM-Moduls "APO" von SAP. it - Industrielle In-formationstechnik 9: 18-21

Bourlakis C, Bourlakis M (2004) The Future of Food Supply Chain Management. In: Bour-lakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Ox-ford et al., pp 221-230

Bourlakis MA, Weightman PWH (2004) Introduction to the UK Food Supply Chain. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 1-10

Bowersox DJ, Closs DJ (1996) Logistical Management - The Integrated Supply Chain Process. McGraw-Hill, New York et al.

Brandes D (1999) Konsequent einfach - Die ALDI-Erfolgsstory. Campus, Frankfurt am Main

Brauckmann-Berger G, Rindle KN (2003) Sich den Transpondern schrittweise nähern. Le-bensmittel Zeitung 42: 66

Brauer H (1999) Brühwursttechnologie - Technologischer Leitfaden für das Kuttern, Um-röten und sensorische Bewerten von Brühwurst. afz-Reihe FLEISCHKAUFMANN & PRAKTIKER, Deutscher Fachverlag, Frankfurt am Main

Braun M (1999) Ausdifferenzierung eines Componentware-PPS-Systems in Richtung auf Branchen und Betriebstypen. Dissertation at Friedrich-Alexander-Universität Erlan-gen-Nürnberg

Brenchley D (2002) CPFR-Implementierung bei Londis in Großbritannien. In: Seifert, D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 299-303

Broekmeulen RACM (1998) Layout and Operations Management of Distribution Centers for Perishables. Ponsen & Looijen, Wageningen

Buffa ES, Sarin RK (1987) Modern Production / Operations Management, 8th eds. John Wiley & Sons, New York et al.

Büker B, Berger A (2004) Radio Frequenz Identifikation wird jetzt erwachsen. Lebensmit-tel Zeitung 36: 58

Bundesverband der Deutschen Fleischwarenindustrie e. V. (2003) Geschäftsbericht 2002/2003. Bonn, September 2003

Burkard RE, Fortuna T, Hurkens CAJ (2002) Makespan minimization for chemical batch processes using non-uniform time grids. Computers and Chemical Engineering 26: 1321-1332

Busch A, Lange H, Langemann T (2002) Marktstudie: Standardsoftware zum Collaborative Supply Chain Management. Fraunhofer-Anwendungszentrum Logistikorientierte Be-triebswirtschaft und Heinz Nixdorf Institut der Universität Paderborn, Paderborn

Buscher R, Jelken O (2000) Von der Plicht zur Kür - Die Integration von ERP- und APS-Systemen wird in den Unternehmen zum entscheidenden Erfolgsfaktor. Logistik Heute 5: 64-66

220 References

Cano-Munoz G (1991) Manual on meat cold storage operation and management. FAO Animal Production and Health Paper 92, Food and Agriculture Organization of the United Nations, Rome

Cap Gemini Ernst & Young (2002a) Advanced Planning Systems (APS) 2003 - An over-view of the vendors, developments and trends in the Dutch APS market. Utrecht

Cap Gemini Ernst & Young (2002b) Inventory Optimization Tools 2003 - An overview of the vendors, developments and trends in the Dutch Inventory Optimization market. Utrecht, http://www.logisticsit.com/inventory_optimisation_ 2003.pdf [11.03.2003]

Cap Gemini Ernst & Young (2002c) Transportation Software 2003 - An overview of the vendors, developments and trends in the Dutch transportation software market. Utrecht, http://www.logisticsit.com/transport-software-2003.pdf [20.03.2004]

Cap Gemini Ernst & Young (2002d) Supply Chain Event Management & Visibility Soft-ware 2003 - An overview of the vendors, developments and trends in the Dutch Supply Chain Event Management & Visibility software market. Utrecht,http://www.worldchain.com/pdf/SCEM-Visibility 2003-CGEY0403.pdf [20.03.2004]

Cap Gemini Ernst & Young (2002e) Forecasting / Demand Planning 2003 - An overview of the vendors, developments and trends in the Dutch Forecasting / Demand Planning market. Utrecht, http://www.logisticsit.com/report-forecasting-2003.pdf [20.03.2004]

Cap Gemini Ernst & Young (2002f) Supply Chain Network Design 2003 - An inventarisa-tion of the vendors, developments and trends in the Dutch Supply Chain Network De-sign market. Utrecht, http://www.logisticsit.com/reports/Supply%20Chain %20Net-

work%20Design%20June%202003.pdf [20.03.2004] Centrale für Coorganisation (1999) Die logistische Kette in Deutschland. Köln Chandan RC, Shahani KM (1993) Yogurt. In: Hui YH (ed) Dairy Science and Technology

Handbook. VCH, New York et al., pp 1-56 Chappell. G, Ginsburg. L, Schmidt P, Smith J, Tobolski J (2002) Auto-ID on Demand: The

Value of Auto-ID Technology in Consumer Packaged Goods Demand Planning. Re-port of Accenture and Auto-IC Center, Cambridge (USA)

Chase RB, Aquilano NJ, Jacobs FR (1998) Production and Operations Management - Manufacturing and Services, 8th edn. McGraw-Hill, Boston et al.

Chowdhury MR, Sarker BR (2001) Manufacturing batch size and ordering policy for prod-ucts with shelf lives. International Journal of Production Research 39 (7): 1405-1426

Cohen MA, Lee HL (1989) Resource Deployment Analysis of Global Manufacturing and Distribution Networks. Journal of Manufacturing and Operations Management 2: 81-104

Connor JM, Wills RL (1988) Marketing and Market Structure of the U.S. Food Processing Industries. In: McCorkle CO Jr (ed) Economics of Food Processing in the United States. Academic Press, San Diego et al., pp 117-166

Cooke DL, Rohleder TR, Silver EA (2004) Finding effective schedules for the economic lot scheduling problem: A simple mixed integer programming approach. International Journal of Production Research 42 (1): 21-36

Corsten H, Gössinger R (2001a) Einführung in das Supply Chain Management. Olden-bourg, München et al.

Corsten H, Gössinger R (2001b) Advanced Planning Systems - Anspruch und Wirklichkeit. PPS Management 6 (2): 32-39

Corsten D, Gabriel C (2002) Supply Chain Management erfolgreich umsetzen - Grundla-gen, Realisierungen und Fallstudien. Springer, Berlin et al.

Crama Y, Pochet Y, Wera Y (2001) A Discussion of Production Planning Approaches in the Process Industry. http://citeseer.nj.nec.com/crama01discussion. html [24.03.2003]

CSB-System AG (2004a) CSB-System Gruppenbericht 2003. Geilenkirchen

References 221

CSB-System AG (2004b) Management & Controlling - BI-Business Intelligence im CSB-System. Geilenkirchen

CSB-System AG (2004c) InterFace - Spezial zur Rückverfolgung. Geilenkirchen Davies R, Diepeveen N, Diks E, Vloemans V (2002) How to get the most out of your sup-

ply chain? - "An overview of APS systems in the consumer products manufacturing and process industry". Report of Deloitte Consulting

Davis G (2002a) Think intelligent. In: Cap Gemini Ernst & Young (eds) Supply Chain Guide 2002 - Fast Track to a Flexible Future. http://www.us.cgey.com/ind_serv/ ser-

vices/supply_chain/default.asp [09.04.2003], pp 5-6 Davis G (2002b) Pallet and chips. In: Cap Gemini Ernst & Young (eds) Supply Chain

Guide 2002 - Fast Track to a Flexible Future. http://www.us.cgey.com/ind_serv/ ser-

vices/supply_chain/default.asp [09.04.2003], p 10 Dawson J (2004) Food Retailing, Wholesaling and Catering. In: Bourlakis MA, Weight-

man, PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 116-135

de Lourdes Pérez-Chabela M, Guerrero Legarreta I (2001) Slaughtering and Processing Equipment. In: Hui, YH, Nip, W-K, Rogers, RW, Young, OA (eds) Meat Science and Applications. Marcel Dekker, New York et al., pp 255-273

de Sitter H (1998) Regulatory aspects of food production systems. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 159-175

Dekker M, Linnemann AR (1998) Product Development in the Food Industry. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 67-86

Dickersbach JT (2003) Supply Chain Management with APO - Structures, Modelling Ap-proaches and Implementation Pecularities. Springer, Berlin et al.

Dietz D (2003) Dem Rohstoff auf der Spur. Lebensmittel Zeitung 40: 50-51 Dik RW, von Lewinski H, Whitaker JD, Brooks JD, Lee H, Peleg B, Whang S, van Was-

senhove LN, Guide VDR, Gritsay V (2003) A Global Study of Supply Chain Leader-ship and Its Impact on Business Performance. Industry Study of Accenture, Stanford University and INSEAD

Dudek G, Rohde J, Sürie C (2002) Advanced Planning Systems - Lösungsverfahren und Modellierung. Industrie Management 18 (6): 49-52

Eggert U (1998) Der Handel im 21. Jahrhundert. Metropolitan, Düsseldorf et al. Eilander G (2002) Consumer drivers. In: Grievink J-W, Josten L, Valk C (eds) State of the

Art in Food: The changing face of the worldwide food industry. Elsevier Business In-formation, Arnhem, pp 60-72

Elmaghraby SE (1978) The Economic Lot Scheduling Problem: Review and Extensions. Management Science 24 (6): 587-598

Emblem A (2000) Predicting packaging characteristics to improve shelf-life. In: Kilcast D, Subramaniam P (eds) The stability and shelf-life of food. Woodhead, Cambridge, pp 145-169

Ennen S (2003) Safety tops concerns for coming year. http://www.foodprocessing.com/

Web_First/FP.nsf/ArticleID/SENN-5JPSU9 [05.02.2004] Erdtsieck B (1989) Quality Requirements in the Modern Poultry Industry. In: Mead GC

(ed) Processing of Poultry. Elsevier Science, New York et al., pp 1-30 Fairfield DK (2002) Fünf Fehlannahmen über CPFR. In: Seifert D (ed) Collaborative Plan-

ning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 205-224

Fandel G, François P (1999) Betriebswirtschaftliche Ansprüche an die Planungs-funktionalität von PPS-Systemen und ihre Realisierungen. In: Werners B, Gabriel R

222 References

(eds) Fortgeschrittene Planungsfunktionalität aktueller PPS-Systeme. Arbeitsbericht Nr. 83, pp 1-21

Felger U (2003) Neue Dimensionen der Lebensmittelsicherheit. Lebensmittel Zeitung 40: 54

Felger U (2004) Auch kleine Hersteller nutzen EDI. Lebensmittel Zeitung 36: 67 Ferrar A (2000) Rapid Supply Chain Integration - The software that will help shape the

components industry. Informa Publishing, London Fischer ME (2001) "Available-to-Promise": Aufgaben und Verfahren im Rahmen des Sup-

ply Chain Management. Roderer, Regensburg Fleischmann B (1990) The discrete lot-sizing and scheduling problem. European Journal of

Operational Research 44: 337-348 Fleischmann B (1998) Der Unterschied zwischen MRP und Planung - Warum vernachläs-

sigen traditionelle PPS-Systeme die Ressourcen? it - Industrielle Informationstechnik 2: 52-55

Fleischmann B (2002) Distribution and Transport Planning. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 195-210

Fleischmann B, Meyr H (1997) The general lot sizing and scheduling problem. OR Spek-trum 19: 11-21

Fleischmann B, Meyr H, Wagner M (2002) Advanced Planning. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 71-96

Forrester JW (1958) Industrial dynamics: a major breakthrough for decision makers. Har-vard Business Review July-August: 37-66

Fraede C (1998) Edifact - was heißt das eigentlich? it - Industrielle Informationstechnik 5: 51-53

Francis M (2004) New Product Development and Information Technology in Food Supply Chain Management: The Case of Tesco. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 153-164

Franke S (2002) Einführung von SCM-Systemen bei der BASF AG Ludwigshafen - Erfah-rungen der Geschäftseinheit Styrolpolymere mit SAP APO/DP. In: Busch A, Dangel-maier W (eds) Integriertes Supply Chain Management - Theorie und Praxis effektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 261-276

Fransoo JC, Rutten WGMM (1994) A Typology of Production Control Situations in Proc-ess Industries. International Journal of Operations & Production Management 14 (12): 47-57

Fraser J (2002) CPFR - Status im Konsumgütersektor. In: Seifert, D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 117-138

Fraunhofer Gesellschaft (1999) Marktstudie Supply Chain Management Software. Institut Produktionstechnik und Automatisierung und Institut Materialfluß und Logistik, Fraunhofer IPA und IML, Stuttgart et al.

Freudenreich P, Bach H (1993) Anfall und Verwertung von Schlachtnebenprodukten. In: Institut für Fleischerzeugung und Vermarktung der Bundesanstalt für Fleischforschung Kulmbach (ed) Beiträge zur Erzeugung und Vermarktung von Fleisch. Bundesanstalt für Fleischforschung, Kulmbach, pp 103-130

Friedrich J-M (2002) Supply-Chain-Management-Software für kleine und mittlere Unternehmen - Lösungsmöglichkeiten mit Componentware. Shaker, Aachen

Friend WR (2002) Getting Kinks Out of the Food Supply Chain. The Food CIO Forum,

http://www.processerp.com/article%20master/Food%20Supply%20Chain%20Manage

ment.pdf [05.02.2004]

References 223

Fritsche B (1999) Die Zukunft von PPS und Supply Chain. Logistik Heute 5: 50-56 Frouws J, van der Ploeg, JD (2000) Issues of Quality in Dutch Dairying. In: Schwarzweller

HK, Davidson AP (eds) Dairy Industry Restructuring. Elsevier Science, New York et al., pp 211-228

Galizzi G, Venturini L (1999) Towards a Theory of Successful Vertical Cooperation in the Food System. In: Galizzi G, Venturini L (eds) Vertical Relationships and Coordination in the Food System. Physica, Heidelberg et al., pp 61-92

Ganeshan R, Jack E, Magazine MJ, Stephens P (1999) A Taxonomic Review of Supply Chain Management Research. In: Tayur S, Ganeshan R, Magazine MJ (eds) Quantita-tive Approaches to Supply Chain Management. Kluwer, Boston et al., pp 839-879

Gellynck X, Verbeke W, Viaene J (2002) Consumer Value of Tracebility: Opportunities for Market Orientation in Meat Supply Chains. In: Trienekens JH, Omta SWF (eds) Para-doxes in Food Chains and Networks - Proceedings of the Fifth International Confer-ence on Chain and Network Management in Agribusiness and the Food Industry. Wageningen Academic Publishers, Wageningen, pp 217-228

Geoffrion AM (1977) Twenty A Priori Error Bounds for Procurement Commodity Aggre-gation in Logistics Planning Models. Naval Research Logistics Quarterly 24 (2): 201-212

Geoffrion AM, Powers RF (1995) Twenty Years of Strategic Distribution System Design: An Evolutionary Perspective. Interfaces 25 (5): 105-127

Giles JA (1999) New structures, new customers and new strategies: implications for food safety. Conference Proceedings of "Postharvest Convention 1999", Cranfield Univer-sity, Bedford, 24.03.1999, http://hort.cabweb.org/Postharv/Conferences/conf99.htm

[10.03.2003] Gill CO (2000) Safe Handling of Raw Meat and Poultry Products. In: Farber JM, Todd

ECD (eds) Safe Handling of Foods. Marcel Dekker, New York et al., pp 1-40 Gilpin J, Traill WB (1999) Manufacturer Strategies in the European Food Industry. In: Gal-

izzi G, Venturini L (eds) Vertical Relationships and Coordination in the Food System. Physica, Heidelberg et al., pp 285-310

Goetschalckx M (2002) Strategic Network Planning. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 105-122

Goetschalckx M, Vidal CJ, Dogan K (2002) Modeling and design of global logistics sys-tems: A review of integrated strategic and tactital models and design algorithms. Euro-pean Journal of Operational Research 143: 1-18

Göpfert I (2002) Einführung, Abgrenzung und Weiterentwicklung des Supply Chain Man-agements. In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Management - Theorie und Praxis effektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 25-44

Gould WA (1997) Fundamentals of Food Processing and Technology. CTI, Timonium Goyal SK (1994) A note on "Effect of production cost on shelf life". International Journal

of Production Research 32 (9): 2243-2245 Goyal SK, Gunasekaran A (1995) An integrated production-inventory-marketing model for

deteriorating items. Computers & Industrial Engineering 28 (4): 755-762 Goyal SK, Giri BC (2003) The production-inventory problem of a product with time vary-

ing demand, production and deterioration rates. European Journal of Operational Re-search 147: 549-557

Grandin T (2001) Antemortem Handling and Welfare. In: Hui YH, Nip W-K, Rogers RW, Young OA (eds) Meat Science and Applications. Marcel Dekker, New York et al., pp 221-254

224 References

Gregory NG (1989) Stunning and Slaughter. In: Mead GC (ed) Processing of Poultry. El-sevier Science, New York et al., pp 31-63

Grievink J-W, Josten L, Valk C (2002) State of the Art in Food: The changing face of the worldwide food industry. Elsevier Business Information, Arnhem

Gronau N (2001) Auswahl und Einführung industrieller Standardsoftware. PPS Mana-gement 6 (3): 14-19

Grünauer K-M (2001) Supply Chain Management - Architektur, Werkzeuge und Methode. Difo, Bamberg

Grünewald C (2001) SAP-APO Projekt bei Adventis. In: Lawrenz O, Hildebrand K, Nen-ninger M, Hillek T (eds) Supply Chain Management - Konzepte, Erfah-rungsberichte und Strategien auf dem Weg zu digitalen Wertschöpfungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 303-309

Grunow M, Günther H-O, Viergutz B (2002) APS-System auf dem Prüfstand. PPS Man-agement 7 (1): 14-17

Günther H-O (1987) Planning Lot Sizes and Capacity Requirements in a Single Stage Pro-duction System. European Journal of Operational Research 31: 223-231

Günther H-O, Tempelmeier H (1995) Produktionsmanagement - Einführung mit Übungs-aufgaben, 2nd edn. Springer, Berlin et al.

Günther H-O, Tempelmeier H (2000) Produktion und Logistik, 4th edn. Springer, Berlin et al.

Günther H-O, van Beek P (2003) Advanced planning and scheduling in process industry. In: Günther H-O, van Beek P (eds) Advanced Planning and Scheduling Solutions in Process Industry. Springer, Berlin et al, pp 1-7

Günther H-O, Neuhaus U (2004) Advanced planning and scheduling in the consumer goods industry: Realizing block planning concepts for make-and-pack production using MILP modelling and SAP’s APO software. In: van Wassenhove LN, de Meyer A, Yücesan E, Günes ED, Muyldermans L (eds) Operations management as a change agent. vol 2, INSEAD, pp 757-766

Haase K (1994) Lotsizing and Scheduling for Production Planning. Springer, Berlin et al. Haase K (1996) Capacitated lot-sizing with sequence dependent setup costs. OR Spektrum

18: 51-59Hammer GF (1984) Zusatzstoffe und Zusätze. In: Institut für Technologie der Bundes-

anstalt für Fleischforschung (ed) Technologie der Brühwurst. Fördergesellschaft der Bundesanstalt für Fleischforschung, Kulmbach, pp 80-102

Handfield RB, Nichols EL Jr (1999) Introduction to Supply Chain Management. Prentice Hall, New Jersey et al.

Hartweg E, Bruckner A (2001) Gestaltung von Supply-Chain-Prozessen. Logistik Heute 5: 54-55

Harvey D (2004) The UK Livestock System. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 62-82

Haug H (1999) Der Vertrieb blickt in die Produktion - Wie "Dübel-Fischer" mit "APO" den Kundenversorgungsprozess optimiert. it - Industrielle Informationstechnik 6: 14-17

Hauptmann S, Zeier A (2000) Ein Beitrag zu einer Kern-Schalen-Architektur für Supply-Chain-Management (SCM)-Software - Teil I: Anforderungen an den Kern einer SCM-Software und deren Abdeckung in SAP APO 2.0/3.0. FORWIN-Bericht-Nr.: FWN-2000-007, Bayerischer Forschungsverbund Wirtschaftsinformatik, Bamberg et al.,http://www.forwin.de/download/berichte/Internet_FWN_2000-007.pdf [11.03.2003]

Hauptmann S, Zeier A (2001) Ein Beitrag zu einer Kern-Schalen-Architektur für Supply-Chain-Management (SCM)-Software - Teil II: Anforderung an die Schalen einer SCM-Software und deren Abdeckung in SAP APO 2.0/3.0. FORWIN-Bericht-Nr.:

References 225

FWN-2001-001, Bayerischer Forschungsverbund Wirtschaftsinf., Bamberg et al.,http://www.forwin.de/download/berichte/Internet_FWN_2001-001.pdf [11.03.2003]

Hausruckinger G, Pelousek WF (2003) Lücken im Regal sind mit OSA ein vermeidbares Übel. Lebensmittel Zeitung 35: 59

Heckmann P, Shorten D, Engel H (2003) Supply Chain Management at 21 - The Hard Road to Adulthood. Industry Study of Booz Allen Hamilton Inc.

Heidrich J (2002) Einsatz von Supply-Chain-Management-Systemen in der Stahlindustrie. Industrie Management 18 (5): 46-49

Heimig D (2003) Joghurt-Drinks treffen den Zeitgeist. Lebensmittel Zeitung 38: 54-56 Hellingrath B, Gehr F, Palm D, Nayabi K (2001) Aufgabenspektrum und Anbieter von

SCM-Softwarelösungen im Überblick. In: Arnold U, Mayer R, Urban G (eds) Supply Chain Management - Unternehmensübergreifende Prozesse Kollaboration IT-Standards. Lemmens, Bonn, pp 211-227

Hellingrath B, Hieber R, Laakmann F, Nayabi K (2002) Die Einführung von SCM-Softwaresystemen. In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Man-agement - Theorie und Praxis effektiver unternehmensübergreifender Geschäfts-prozesse. Gabler, Wiesbaden, pp 187-211

Hemmelmann W (2003) Aldi schiebt die Weiße Linie an, Lebensmittel Zeitung 38: 62-63 Henson S, Northen J (1999) The Role of Quality Assurance in Food Retailer-Manufacturer

Relations: The Case of Food Safety Controls in the Supply of Retailer Own-branded Products in the UK. In: Galizzi G, Venturini L (eds) Vertical Relationships and Coor-dination in the Food System. Physica, Heidelberg et al., pp 347-369

Heymann E (2002) Perspektiven des Ernährungsgewerbes in Deutschland. Presentation of Deutsche Bank Research, Frankfurt am Main

Hieber R (2002) Supply Chain Management - A Collaborative Performance Measurement Appoach. vdf Hochschulverlag, Zürich

Hill CA (2000) Information Technology and Supply Chain Management: A Study of the Food Industry. Hospital Material Management Quarterly 22 (1): 53-58

Hill CA, Scudder GD (2001) A Study of Supply Chain Coordination within the Food In-dustry. http://www.gsu.edu/~mgtcah/research.html [15.01.2003]

Hill CA, Scudder GD (2002) The use of electronic data interchange for supply chain coor-dination in the food industry. Journal of Operations Management 20: 375-387

Hine DJ (1987) Shelf-life prediction. In: Paine FA (ed) Modern Processing, Packaging and Distribution Systems for Food. Blackie, Glasgow et al., pp 62-85

Hitchens D, Birnie E, McGowan A, Triebswetter U, Cottica A (1998) The Firm, Competi-tiveness and Environmental Regulations. Edward Elgar, Cheltenham et al.,http://www.gsu.edu/~mgtcah/research.html [15.01.2003]

Hoffmann K (2003a) Kontrapunkte zum Discountangebot, Lebensmittel Zeitung 35: 38-39 Hoffmann, K (2003b) Doppelstrategie mit Frischevorteil, Lebensmittel Zeitung 35: 41-42 Hoffmann, K (2004a) Die neue Produktions-Taktik, Lebensmittel Zeitung 11: 48-49 Hoffmann, K (2004b) Das lukrative Ost-Engagement, Lebensmittel Zeitung 11: 50 Hoffmann, K (2004c) Hier werden die Preise gemacht, Lebensmittel Zeitung 35: 38-40 Holler D (2003) Aldi riecht den Braten, Lebensmittel Zeitung 17: 39-40 Hoogland JP, Jellema A, Jongen WMF (1998) Quality Assurance System. In: Jongen,

WMF, Meulenberg, MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 139-158

Hughes D (2004) Food Manufacturing. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 99-115

Hurtmanns F, Packowski J (1999) Supply-Chain-Optimierung mit SAP APO in der Che-mieindustrie: Einsatzuntersuchung und Geschäftsprozessszenarien. HMD Praxis der Wirtschaftsinformatik 207: 58-69

226 References

Ierapetritou MG, Floudas CA (1998) Effective Continuous-Time Formulation for Short-Term Scheduling - 1. Multipurpose Batch Processes - 2. Continuous and Semicon-tinuous Processes. Industrial and Engineering Chemistry Research 37: 4341-4374

Jahn G, Kiefer S, Spiller A (2003) Zertifizierung im Umbruch. Lebensmittel Zeitung 40: 55 Jakeman CM (1994) Scheduling needs of the food processing industry. Food Research In-

ternational 27: 117-120 Jakob H (2002) Erfolgreiche Brühwurstrezepturen. Deutscher Fachverlag, Frankfurt am

MainJank K, Wäscher G (1999) Mischungsoptimierung in der Wurstwarenherstellung - Eine

Fallstudie. OR Spektrum 21: 503-523 Jehle E (2000) Steuerung von großen Netzen in der Logistik unter besonderer Berück-

sichtigung von Supply Chains. In: Wildemann H (ed) Supply Chain Management. TCW, München, pp 205-226

Jensen T (1999) Planen nach Bedarf - "Advanced Planning und Scheduling": Ablösung oder Add-on für die traditionelle PPS? it - Industrielle Informationstechnik 1: 33-34

Jimenez S, Brown T, Jordan J (1998) Network-modelling tools - Enhancing supply chain decision making. In: Gattorna J, Jones T, Danks A, Dhillon Y, Holdforth L (eds) Stra-tegic Supply Chain Alignment - Best practice in supply chain management. Gower, Aldershot et al., pp 302-324

Jones JM, Grey TC (1989) Influence of Processing on Product Quality and Yield. In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 127-181

Jongen WMF (1996) Added value of agri-food products by an integrated approach. In: Fenwick GR, Hedley C, Richards RL, Khokhar S (eds) Agri-Food Quality - An Inter-disciplinary Approach. The Royal Society of Chemistry, Cambridge, pp 263-271

Kaftan K, Kaftan H-J (2002) ERP-Software zur Simulation von Geschäftsprozessen branchenspezifischer Modellunternehmen. Industrie Management 18 (1): 48-51

Kallrath J (2002) Planning and scheduling in the process industry. OR Spectrum 24: 219-250

Kampker R, Wienecke K (2001) Produktionsprogrammplanung mit PPS/ERP-Systemen. Zeitschrift für Unternehmensentwicklung und industrial engineering 50 (2): 52-64

Kansky D (1999) Supply Chain Management - Der direkte Weg zur Steigerung von Ertrag und Wettbewerbsfähigkeit. Industrie Management 15 (5): 14-17

Kapell E (2003) Mehr Qualität in der Supply Chain. Lebensmittel Zeitung 44: 25 Karmarkar US, Schrage L (1985) The Deterministic Dynamic Product Cycling Problem.

Operations Research 33: 326-345 Kilcast D (2000) Sensory evaluation methods for shelf-life assessment. In: Kilcast D,

Subramaniam P (eds) The stability and shelf-life of food. Woodhead, Cambridge, pp 79-105

Kilcast D, Subramaniam P (2000) The stability and shelf-life of food. Woodhead, Cam-bridge

Kilger C (2002a) The Definition of a Supply Chain Project. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 241-259

Kilger C (2002b) The Selection Process. In: Stadtler H, Kilger C (eds) Supply Chain Man-agement and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 261-273

Kilger C, Müller A (2002) Integration von Advanced Planning Systemen in die innerbetrie-bliche DV-Landschaft. In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Management - Theorie und Praxis effektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 213-234

References 227

Kilger C, Reuter B (2002) Collaborative Planning. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 223-237

Kilger C, Schneeweiss L (2002a) Demand Fulfilment and ATP. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 161-175

Kilger C, Schneeweiss L (2002b) Computer Assembly. In: Stadtler H, Kilger C (eds) Sup-ply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 335-352

Kim Y-S (2001) Meat Production. In: Hui YH, Nip W-K, Rogers RW, Young OA (eds) Meat Science and Applications. Marcel Dekker, New York et al., pp 563-579

Kinsey J (2000) "Electronic Systems in the Food Industry: Entropy, Speed and Sales". Pro-ceedings of "E-Business Transformation: Sector Developments and Policy Implica-tions" Conference, Washington D.C., 26.-27.09.2000, http://www.e-conomy.berkeley.

edu/conferences/9-2000/EC-conference2000_papers/JeanKinsey.pdf [09.04.2003] Kirch R, Manger M (2001) Supply Chain Management in der Pharmaindustrie. PPS Man-

agement 6 (2): 52-54 Kirschmeier T (2003) Lichtblicke zwischen zwölf Metern Joghurt, Lebensmittel Zeitung

38: 59Klein R (2002a) Genetic Algorithms. In: Stadtler H, Kilger C (eds) Supply Chain Manage-

ment and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 403-410 Klein R (2002b) Constraint Programming. In: Stadtler H, Kilger C (eds) Supply Chain

Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 411-418 Kleineicken A (2002) Electronic Procurement. In: Wannenwetsch HH, Nicolai S (eds) E-

Supply Chain Management - Grundlagen-Strategien-Praxisanwendungen. Gabler, Wiesbaden, pp 100-126

Klettner P-G (1984) Zerkleinerungstechnik. In: Institut für Technologie der Bundesanstalt für Fleischforschung (ed) Technologie der Brühwurst. Fördergesellschaft der Bunde-sanstalt für Fleischforschung, Kulmbach, pp 103-122

Klostermeyer H (1996) Konsummilch und Milcherzeugnisse. In: Heiss R (ed) Lebens-mitteltechnologie - Biotechnologische, chemische, mechanische und thermische Ver-fahren der Lebensmittelverarbeitung, 5th edn. Springer, Berlin et al., pp 1-11

Knolmayer G (2001a) Advanced Planning and Scheduling Systems: Optimierungs-methoden als Entscheidungskriterium für die Beschaffung von Software-Paketen? In: Wagner U (ed) Zum Erkenntnisstand der Betriebswirtschaftslehre am Beginn des 21. Jahrhunderts. Duncker & Humblot, Berlin, pp 135-155

Knolmayer G (2001b) Architekturen von Supply Chain Management Systemen. Pro-ceedings of speech at Gesellschaft für Operations Research (GOR) e.V., Rot, 12.01.2001, http://130.149.112.124:81/pm/inhalt/de/forschung/scm-ag/2001_01_11/

Knolmayer.pdf [05.04.2004] Knolmayer G, Mertens P, Zeier A (2002) Supply Chain Management Based on SAP Sys-

tems - Order Management in Manufacturing Companies. Springer, Berlin et al. Koçlar A, Süral H (2005) A note on “The general lot sizing and scheduling problem”. OR

Spectrum 27 (to appear) Kodweiss J, Nadjmabadi K (2001) Supply Chain Management als strategische Herausfor-

derung. In: Lawrenz O, Hildebrand K, Nenninger M, Hillek T (eds) Supply Chain Management - Konzepte, Erfahrungsberichte und Strategien auf dem Weg zu digitalen Wertschöpfungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 71-87

Kortmann J, Lessing H (2000) Marktstudie: Standardsoftware für Supply Chain Manage-ment. Fraunhofer-Anwendungszentrum Logistikorientierte Betriebswirtschaft und Heinz Nixdorf Institut der Universität Paderborn, Paderborn

228 References

Kotzab H (2000) Zum Wesen von Supply Chain Management vor dem Hintergrund der be-triebswirtschaftlichen Logistikkonzeption - erweiterte Überlegungen. In: Wildemann H (ed) Supply Chain Management. TCW, München, pp 21-47

KPMG Corporate Finance UK (2000) Europe's recipe for success...innovate and consoli-date. Industry report of KPMG Corporate Finance UK

Krampe G (2004) Discounter und kein Ende. Lebensmittel Zeitung 1: 20-21 Kuhn A, Hellingrath H (2002) Supply Chain Management - Optimierte Zusammenarbeit in

der Wertschöpfungskette. Springer, Berlin et al. Kuznesof S, Brennan M (2004) Perceived Risk and Product Safety in the Food Supply

Chain. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 32-48

Labuza TP (2000) The Search for Shelf Life - An update on continued efforts in under-standing practical strategies for determining and testing the shelf life of food products. Food testing Analysis, May 2000, http://fscn.che.umn.edu/Ted_Labuza/PDF_files/ pa-

pers/Search%20For%20Shelf%20Life-FTA.pdf [24.03.2003] Labuza TP, Taoukis PS (1990) The Relationship Between Processing and Shelf-Life. In:

Birch GG, Campbell-Platt G, Lindley MG (eds) Food for the '90s. Elsevier Science, New York et al., pp 73-106

Lang M (2002a) eDistribution - Distributionsstrategien im eZeitalter. In: Wannenwetsch HH, Nicolai S (eds) E-Supply Chain Management - Grundlagen-Strategien-Praxisanwendungen. Gabler, Wiesbaden, pp 182-206

Lang M (2002b) eProduction - Von der Push- zur Pull-Produktion. In: Wannenwetsch HH, Nicolai S (eds) E-Supply Chain Management - Grundlagen-Strategien-Praxisanwen-dungen. Gabler, Wiesbaden, pp 127-155

Langemann T (2002) Collaborative Supply Chain Management (CSCM). In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Management - Theorie und Praxis ef-fektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 421-438

Lebensmittel Zeitung (2001) LZ Report 2001/2002. Verlagsgruppe Deutscher Fachverlag, Frankfurt am Main

Lebensmittel Zeitung (2003) LZ Report 2003/2004. Verlagsgruppe Deutscher Fachverlag, Frankfurt am Main

Lebensmittel Zeitung, PwC Consulting (2002) SCM in der Konsumgüterindustrie - Eine Studie über die Strukturen der Logistik und des Supply Chain Managements. Industry Study, October 2002

Lee HL, Padmanabhan V, Whang S (1997) The Bullwhip Effect in Supply Chains. Sloan Management Review, Spring 1997, pp 93-102

Lehtonen J-M (1999) Supply Chain Development in Process Industry. Acta Polytechnica Scandinavia, Helsinki

Leung SCH, Wu Y (2004) A robust optimization model for stochastic aggregate production planning, Production Planning & Control 15 (5): 502-514

Li Y, Man F, Tang S, Kwong S, Ip WH (2000) Genetic algorithm to production planning and scheduling problems for manufacturing systems. Production Planning & Control 11 (5): 443-458

Loderhose B (2003) RFID bewirkt Quantensprung. Lebensmittel Zeitung 25: 95 Loderhose B (2004) Paletten plagen Konsumgüterbranche. Lebensmittel Zeitung 23: 26 Loderhose B, Kapell E (2003) Neue Technik soll alte Probleme lösen. Lebensmittel Zei-

tung 42: 61MacCarthy BL, Fernandes FCF (2000) A multi-dimensional classification of production

systems for the design and selection of production planning and control systems. Pro-duction Planning & Control 11 (5): 481-496

References 229

Manna SK, Chaudhuri KS (2001) An economic order quantity model for deteriorating items with time-dependent deterioration rate, demand rate, unit production cost and shortages. International Journal of System Science 32 (8): 1003-1009

Marbacher A (2001) Demand & Supply Chain Management. Paul Haupt, Berlin et al. Marshall D (2004) The Food Consumer and the Supply Chain. In: Bourlakis MA, Weight-

man PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 11-31 Mayr M (1996) Hierarchische Produktionsplanung mit zyklischen Auflagemustern.

Roderer, Regensburg McCorkle CO Jr (1988) U.S. Food Manufacturing Industries: An Overview. In: McCorkle

CO Jr (ed) Economics of Food Processing in the United States. Academic Press, San Diego et al., pp 1-46

McKinnon A (1999) Vehicle Utilisation and Energy Efficiency in the Food Supply Chain - Full Report of the Key Performance Indicator Study. http://www.som.hw.ac.uk/ logis-

tics/download.html [15.01.2003] McKinnon A (2004) Third Party Logistics in the Food Supply Chain. In: Bourlakis MA,

Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 165-178

McKinnon A, Campbell J (1998) Quick-response in the Frozen Food Supply Chain: The Manufacturer's Perspective. http://www.som.hw.ac.uk/logistics/download.html

[15.01.2003] McLaughlin E (2002) The food industry in 2005, a forecast. In: Grievink J-W, Josten L,

Valk C (eds) State of the Art in Food: The changing face of the worldwide food indus-try. Elsevier Business Information, Arnhem, pp 598-608

Mead GC (1989) Hygiene Problems and Control of Process Contamination. In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 183-220

Mebus J (2004) KPI als Ansporn. Lebensmittel Zeitung 36: 64 Mekschrat M (2002) Umsetzung des SCM-Systems SAP APO bei der Degussa AG in den

Business Units der Röhm GmbH & Co KG. In: Busch A, Dangelmaier W (eds) Inte-griertes Supply Chain Management - Theorie und Praxis effektiver unternehmensüber-greifender Geschäftsprozesse. Gabler, Wiesbaden, pp 277-294

Méndez CA, Cerdá J (2002) An MILP-based approach to the short-term scheduling of make-and-pack continuous production plants. OR Spectrum 24: 403-429

Menz WW (1986) Shelf-Life of Milk. In: Charalambous G (ed) Handbook of Food and Beverage Stability - Chemical, Biochemical, Microbiological, and Nutritional Aspects. Academic Press, San Diego et al., pp 622-664

Meulenberg MTG, Viaene J (1998) Changing food marketing systems in western countries. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 5-36

Meyr H (1999) Simultane Losgrößen- und Reihenfolgeplanung für kontinuierliche Produk-tionslinien. Gabler, Wiesbaden

Meyr H (2002) Forecast Methods. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 379-390

Meyr H, Rohde J, Schneeweiss L, Wagner M (2002a) Architecture of Selected APS. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 293-304

Meyr H, Wagner M, Rohde J (2002b) Structure of Advanced Planning Systems. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 99-104

Michael J, Funder J, Huppertz S (2002) KonsumGüter 2010: Handel-Macht-Marken - Eine Expertenbefragung und Trendstudie über die Entwicklung des deutschsprachigen Le-

230 References

bensmitteleinzelhandels und der Konsumgüterindustrie bis zum Jahr 2010. Industry Report of Accenture, Frankfurt am Main

Mizrahi S (2000) Accelerated shelf-life tests. In: Kilcast D, Subramaniam P (eds) The sta-bility and shelf-life of food. Woodhead, Cambridge, pp 107-128

Morris CE (2000) Scheduling Software: A Key B2B Link. http://www.food engineering-

mag.com/articles/2000/1100/0011tech.htm [11.03.2003] Muir DD, Banks JM (2000) Milk and milk products. In: Kilcast D, Subramaniam P (eds)

The stability and shelf-life of food. Woodhead, Cambridge, pp 197-219 Müller W-D (1984) Füllen und Räuchern. In: Institut für Technologie der Bundesanstalt für

Fleischforschung (ed) Technologie der Brühwurst. Fördergesellschaft der Bundesan-stalt für Fleischforschung, Kulmbach, pp 144-164

Murmann C (2004a) Milchbranche vor Paradigmenwechsel. Lebensmittel Zeitung 1: 18 Murmann C (2004b) Grenzverkehr nach Osten. Lebensmittel Zeitung 9: 51-52 Murmann C (2004c) Molkereien machen Front gegen Rewe. Lebensmittel Zeitung 42: 20 Murmann C, Wolfskeil J (2004) Magere Zeiten für Molkereien. Lebensmittel Zeitung 9: 48 N.N. (2001) Will APS replace MRP? APS Insight, Issue 5, http://www.apsinsight.com/

Contents.html [10.03.2003] N.N. (2002a) MRP vs. APS the difference is in the detail. APS Insight, Issue 9,

http://www.apsinsight.com/Contents.html [10.03.2003] N.N. (2002b) APS and Process Industries. APS Insight, Issue 11, http://www.aps in-

sight.com/Contents.html [10.03.2003] N.N. (2002c) Implementing APS. APS Insight, Issue 15, http://www.apsinsight.com/ Con-

tents.html [10.03.2003] N.N. (2003a) Globus plant Disposition im Stundentakt. Lebensmittel Zeitung: 48: 28 N.N. (2003b) Olaf setzt auf Chargen-Transparenz. Lebensmittel Zeitung 38: 30 N.N. (2003c) Discount Anteil höher. Lebensmittel Zeitung 41: 24 N.N. (2003d) Händler öffnen Datenlager zögerlich. Lebensmittel Zeitung 47: 26 N.N. (2004a) Hoffnungsträger RFID. Lebensmittel Zeitung, 11: 30 N.N. (2004b) Funketiketten zum Aufdrucken. Lebensmittel Zeitung 16: 32 N.N. (2004c) Europas Milchindustrie im Umbruch. Lebensmittel Zeitung, 36: 25 N.N. (2004d) Qualitätsüberwachung erfolgreich ausgebaut. Lebensmittel Zeitung 32: 50 N.N. (2004e) Chilled Food schärft das Profil. Lebensmittel Zeitung 35: 40-42 Nahmias S (1982) Perishable Inventory Theory: A Review. Operations Research 30 (4):

680-708Nakhla M (1995) Production control in the food processing industry - The need for flexibil-

ity in operations scheduling. International Journal of Operations & Production Man-agement 15 (8): 73-88

Neuhäuser S (1996) Fleischwaren. In: Heiss, R (ed) Lebensmitteltechnologie – Biotechno-logische, chemische, mechanische und thermische Verfahren der Lebensmittelverar-beitung, 5th edn. Springer, Berlin et al., pp 51-61

Neumann K, Schwindt C, Trautmann N (2002) Advanced production scheduling for batch plants in process industries. OR Spectrum 24: 251-279

Nicolai S (2002) Praxisinstrumente für eine erfolgreiche eSCM-Realisierung. In: Wannen-wetsch HH, Nicolai S (eds) E-Supply Chain Management - Grundlagen-Strategien-Praxisanwendungen. Gabler, Wiesbaden, pp 69-99

Nielsen VC (1989) Treatment and Disposal of Processing Waters. In: Mead GC (ed) Proc-essing of Poultry. Elsevier Science, New York et al., pp 361-411

Nienhaus J, Schnetzler M, Sennheiser A, Weidemann M, Glaubitt K, Pierpaoli F, Heinzel H (2003) Trends im Supply Chain Management. Zentrum für Unternehmens-wissenschaften (BWI) der ETH Zürich, Forschungsinstitut für Rationalisierungs-

References 231

technik (FIR) der RWTH Aachen, Institut für Industrielle Produktionstechnologie und Automation der CNR Mailand, SCOR Training Center Europe

Nosbers F, Plewnia M (2001) Supply Chain Management und Logistikdienstleister. In: Lawrenz O, Hildebrand K, Nenninger M, Hillek T (eds) Supply Chain Management - Konzepte, Erfahrungsberichte und Strategien auf dem Weg zu digitalen Wertschöp-fungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 151-168

Nossent S, de Groot B, Verschuren R (1995) Working Conditions in the European Meat Processing Industry. European Foundation for the Improvement of Living and Work-ing Conditions, Dublin

Ockerman HW, Hansen CL (1988) Animal By-Product Processing. Ellis Horwood, Chich-ester

Ody P (2002) Demand reality. In: Cap Gemini Ernst & Young (eds) Supply Chain Guide 2002 - Fast Track to a Flexible Future. http://www.us.cgey.com/ind_serv/services/ sup-

ply_chain/default.asp [09.04.2003], pp 8-9 Oelker P (1996a) Brühwurstherstellung. In: Sielaff H (ed) Fleischtechnologie, B. Behr's

Verlag, Hamburg, pp 397-425 Oelker P (1996b) Kochwurstherstellung. In: Sielaff H (ed) Fleischtechnologie, B. Behr's

Verlag, Hamburg, pp 383-396 O'Leary DE (2000) Enterprise Resource Planning Systems - Systems, Life Cycle, Elec-

tronic Commerce, and Risk. Cambridge University Press, Cambridge Olhager J, Wikner J (2000) Production planning and control tools, Production Planning &

Control 11 (3): 210-222 Otto F (2004) Ideales Klima für Eigenmarken. Lebensmittel Zeitung 16: 66 Parry RT (1989) Technological Developments in Pre-slaughter Handling and Processing.

In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 65-101 Pearson AM, Gillett TA (1996) Processed Meats, 3rd edn. Chapman & Hall, New York et

al. PeopleSoft Inc. (2004a) About PeopleSoft. http://www.peoplesoft.com/corp/en/about/ in-

dex.jsp?site=corp [13.09.2004] PeopleSoft Inc. (2004b) Corporate Acquisitions. http://www.peoplesoft.com/corp/en/about/

overview/corp_acq.jsp [13.09.2004] PeopleSoft Inc. (2004c) EnterpriseOne Product Lines. http://www.peoplesoft.com/corp/en/

products/ent_one/index.jsp [13.09.2004] PeopleSoft Inc. (2004d) PeopleSoft EnterpriseOne Pure Internet Architecture.

http://www.peoplesoft.com/corp/en/products/ent_one/tools_tech/solution_pure_interne

t_arch.jsp [13.09.2004] PeopleSoft Inc. (2004e) PeopleSoft EnterpriseOne: Supply Chain Management.

http://www.peoplesoft.com/corp/en/products/ent_one/scm/index.jsp [13.09.2004] PeopleSoft Inc. (2004f) PeopleSoft eXtended Business Processes. http://www peoplesoft.

com/corp/en/products/dev_integration_portals/module/extended_business_process.jsp

[13.09.2004] PeopleSoft Inc. (2004g) PeopleSoft Demand Consensus. http://www.peoplesoft.com/corp/

en/products/ent_one/scm/module/demand_consen.jsp [13.09.2004] PeopleSoft Inc. (2004h) PeopleSoft Demand Forecasting and Advanced Forecast Modeling.

http://www.peoplesoft.com/corp/en/products/ent_one/scm/module/demand_forecast.

jsp [13.09.2004] PeopleSoft Inc. (2004i) PeopleSoft Order Promising. http://www.peoplesoft.com/corp/en/

products/ent_one/scm/module/order_promising.jsp [13.09.2004] PeopleSoft Inc. (2004j) PeopleSoft Production Scheduling - Discrete - Data Sheet. http://

www.peoplesoft.com/media/en/pdf/datasheet/ent1_scm_prod_discrete_ds_1203.pdf

[13.09.2004]

232 References

PeopleSoft Inc. (2004k) PeopleSoft Production Scheduling – Process. http://www. people-

soft.com/corp/en/products/ent_one/scm/module/prod_sched_process.jsp [13.09.2004] Perosio DJ, McLaughlin EW, Cuellar S, Park K (2001) Fresh Track 2001 - Supply Chain

Management in the Produce Industry. Department of Agricultural, Resource, and Managerial Economics, Cornell University, Ithaca, New York, http://aem.cornell.edu/

research/rb.htm [10.03.2003] Peterson K, Jimenez M, White A (2002) Gartner Research Note 2Q02: Supply Chain Plan-

ning Magic Quadrant: Out of the Trough. 25.06.2002 Pfohl H-C (2000) Supply Chain Management: Konzept, Trends, Strategien. In: Pfohl H-C

(ed) Supply Chain Management: Logistik plus? 15. Fachtagung der Deutschen Gesell-schaft für Logistik e.V. in Darmstadt, 23.05.2000, Erich Schmidt, Berlin, pp 1-42

Pillep R, von Wrede P (1999) Anspruch und Wirklichkeit - Nutzenpotentiale und Mark-tübersicht von SCM-Systemen. Industrie Management 15 (5): 18-22

Poirier CC, Reiter SE (1997) Die optimale Wertschöpfungskette - Wie Lieferanten, Pro-duzenten und Handel bestens zusammenarbeiten. Campus, Frankfurt am Main et al.

Poluha RG (2001) SCM in der Praxis - Projektmanagement komplexer SCM Projekte. In: Lawrenz O, Hildebrand K, Nenninger M, Hillek T (eds) Supply Chain Management - Konzepte, Erfahrungsberichte und Strategien auf dem Weg zu digitalen Wertschöp-fungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 311-327

Raafat F (1991) Survey of Literature on Continnuously Deteriorating Inventory Models. Journal of the Operational Research Society 42 (1): 27-37

Radespiel E (1996) Kältetechnologische Verfahren und Ausrüstungen. In: Sielaff H (ed) Fleischtechnologie. B. Behr's Verlag, Hamburg, pp 621-662

Rau H, Wu M-Y, Wee H-M (2003) Integrated inventory model for deteriorating items un-der a multi-echelon supply chain environment. International Journal of Production Economics 86: 155-168

Richardson RI (1989) Utilization of Turkey Meat in Further-processed Products. In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 283-324

Richmond B, Burns A, Mabe J, Nuthall L, Toole R (1998) Supply chain management tools - Minimizing the risks: maximizing the benefits. In: Gattorna J, Jones T, Danks A, Dhillon Y, Holdforth L (eds) Strategic Supply Chain Alignment - Best practice in sup-ply chain management. Gower, Aldershot et al., pp 509-520

Richter M, Stockrahm V (2002) Scheduling of Synthetic Granulate. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 305-320

Rindfleisch A, Heide JB (1997) Transaction Cost Analysis: Past, Present, and Future Ap-plications. Journal of Marketing 61: 30-54

Ristic M (1993) Schlachtkörperwert und Fleischqualität von Geflügel. In: Institut für Fleischerzeugung und Vermarktung der Bundesanstalt für Fleischforschung Kulmbach (ed) Beiträge zur Erzeugung und Vermarktung von Fleisch. Bundesanstalt für Fleisch-forschung, Kulmbach, pp 82-102

Rode J (2003a) Die Zukunft funkt mit kleinen Chips. Lebensmittel Zeitung 35: 54 Rode J (2003b) Kostensenkung und Regalverfügbarkeit im Fokus der Branche. Lebens-

mittel Zeitung 35: 56 Rode J (2003c) Wal-Mart gibt bei RFID den Takt an. Lebensmittel Zeitung 46: 26 Rode J (2004a) E-Contor beschleunigt Plus-Einkauf. Lebensmittel Zeitung 13: 24 Rode J (2004b) Wal-Mart gibt bei EPC weiter Gas. Lebensmittel Zeitung 26: 25 Rode J (2004c) Informationen für bessere Geschäfte. Lebensmittel Zeitung 30: 25 Rode J (2004d) Rückverfolgung beschäftigt Branche. Lebensmittel Zeitung 35: 26 Rode J (2004e) Eine Technologie in Entwicklung. Lebensmittel Zeitung 36: 57 Rode J (2004f) Prozesseffizienz steht im Mittelpunkt. Lebensmittel Zeitung 36: 62

References 233

Rohde J (2002) Coordination and Integration. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 211-222

Rohde J, Wagner M (2002) Master Planning. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 143-160

Rojek D (2000) Treffsichere Prognosen - Logistikpotentiale ausschöpfen durch Bündelung der verfügbaren Informationen. In: Pfohl H-C (ed) Supply Chain Management: Lo-gistik plus? 15. Fachtagung der Deutschen Gesellschaft für Logistik e.V. in Darmstadt, 23.05.2000, Erich Schmidt, Berlin, pp 185-204

Rose H (2001) Innovationsbedarf in der Lebensmittelindustrie. In: Rose H, Düll K, Klöden W (eds) Kooperative Produktionssteuerung - Innovationsstrategien für Unternehmen der Lebensmittelindustrie. Forschungsberichte aus dem Institut für Sozialwissen-schaftliche Forschung e.V., München, pp 9-12

Rosenthal I (1991) Milk and Dairy Products - Properties and Processing. VCH, New York et al.

Russell SM, (2001) Spoilage bacteria associated with poultry. In: Sams AR (ed) Poultry Meat Processing. CRC Press, Boca Raton et al., pp 159-180

Rutten WGMM (1995) The use of recipe flexibility in production planning and inventory control. Ponsen & Looijen, Wageningen

Sahinidis NV, Grossmann IE (1991) MINLP Model for cyclic multiproduct scheduling on continuous parallel lines. Computers and Chemical Engineering15 (2): 85-103

Salehi F (2003) Supply Chain Management Solutions. Proceedings of Lecture at the Chair of Production Management at TU Berlin, 19.06.2003, http://130.149.112.124:81/ in-

halt/AT-Kearney-Vortrag-SCM-TU-Berlin-19JUN2003.pdf [17.12.2003] SAP AG (2002) SAP Advanced Planner and Optimizer (SAP APO) - Release 3.1 - Release

Notes. Walldorf SAP AG (2003a) Taking shelf life into account with SNP optimizer. SAP Note No.

579556, Walldorf SAP AG (2003b) Planning with shelf life in APO 3.1. SAP Note No. 563863, Walldorf SAP AG (2003c) Planen mit Haltbarkeitsdaten. Help-function of SAP APO 3.1, Walldorf SAP AG (2003d) SNP does not support any shelf life. SAP Note No. 574321, Walldorf Sarker BR, Babu PS (1993) Effect of production cost on shelf life. International Journal of

Production Research 31 (8): 1865-1872 Sarker BR, Chowdhury MR (2002) Response to technial note on "Manufacturing batch size

and ordering policy for products with shelf lives". International Journal of Production Research 40 (8): 1971-1972

Savic IV (1985) Small-scale sausage production. FAO Animal Production and Health Paper 52, Food and Agriculture Organization of the United Nations, Rome

Schaub A, Zeier A (2000) Eignung von Supply-Chain-Management-Software für unter-schiedliche Betriebstypen und Branchen - untersucht am Beispiel des Produktionspro-zessmodells zum System SAP APO. FORWIN-Bericht-Nr.: FWN-2000-004, Bay-erischer Forschungsverbund Wirtschaftsinformatik, Bamberg et al.,http://www.forwin.de/download/berichte/Internet_FWN_2000-004.pdf [11.03.2003]

Schiefer G (2000) Environmental Management and Life-Cycle Assessment in Supply Chain Management - The Case of the Meat Chain. Proceedings of IAMA Agribusiness Forum in Chicago, 25.06.2000, http://www.ifama.org/conferences/2000Congress/

2000_forum_papers.htm [09.04.2003] Schimitzek P (2004) The Efficient Enterprise - Increased Corporate Success with Industry-

Specific Information Technology and Knowledge Management. CRC Press, Boca Raton et al.

Schmelmer M, Seiling K (2002) Erfahrungsbericht über die Einführung des Supply Chain Management Systems von i2 bei Infineon. In: Busch A, Dangelmaier W (eds) Integri-

234 References

ertes Supply Chain Management - Theorie und Praxis effektiver unternehmens-übergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 295-309

Schmidt W (1997) Effiziente und sichere Auftragserfassung per EDI, Industrie Mana-gement 13 (4): 78-80

Schneeweiss L, Wetterauer U (2002) Semiconductor Manufacturing. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 321-334

Schneider R, Grünewald C (2001) Supply Chain Management-Lösung mit mySAP.com. In: Lawrenz O, Hildebrand K, Nenninger M, Hillek T (eds) Supply Chain Management - Konzepte, Erfahrungsberichte und Strategien auf dem Weg zu digitalen Wertschöp-fungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 169-201

Schonberger RJ, Knod EM (1994) Operations Management - Continuous Improvement, 5th edn. Irwin, Burr Ridge et al.

Schraft RD, Kuhn C, Huen E, Huen J (2001) Situationsanalyse in der deutschen Lebensmit-telindustrie - Ergebnisse der Umfrage 2001. Fraunhofer-Institut für Produktionstechnik und Automatisierung (IPA) und LT Food Medien-Verlag, Stuttgart et al.

Schreiber C (2002) Prozessnahe Informationsverarbeitung in kleinen und mittelständischen Unternehmen der Lebensmittelindustrie, Dissertation at TU Dresden

Schwetz W (2003) Der Konsument rückt jetzt in den Fokus. Lebensmittel Zeitung 38: 73-74

Seidl K (2000) Supply Chain Management Software: Einsatzmöglichkeiten und Nutzener-wartungen. In: Pfohl H-C (ed) Supply Chain Management: Logistik plus? 15. Fachta-gung der Deutschen Gesellschaft für Logistik e.V. in Darmstadt, 23.05.2000, Erich Schmidt, Berlin, pp 161-183

Seifert D (2001) Efficient Consumer Response - Strategische Erfolgsfaktoren für die Wert-schöpfungspartnerschaft von Industrie und Handel. Rainer Hampp, München et al.

Seifert D (2002a) Efficient Consumer Response als Ausgangspunkt von CPFR. In: Seifert D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 27-53

Seifert D (2002b) CPFR als neuer Strategieansatz. In: Seifert D (ed) Collaborative Plan-ning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 55-88

Seifert D (2002c) CPFR: State-of-the-Art-SCM für Europa. In: Seifert D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 227-230

Seiler T (2004) Anlagenbelegungsplanung für die Produktion frischer Lebensmittel - Entwicklung eines Optimierungsmodells am Beispiel der Produktion von Joghurt. Master Thesis at the Chair of Production Management, Technische Universität Berlin

Shepherd J, Lapide L (1999) Supply Chain Planning Optimization: Just the Facts. In: Montgomery Research (ed) Achieving supply chain excellence trough technology. Montgomery Research, San Francisco, pp 166-176

Siebel L (2000) Food Logistics: Lebensmittel via Internet - Trends, Konzepte und lo-gistische Probleme. Symposion, Düsseldorf

Sielaff H (1996) Fleischtechnologie. B. Behr's Verlag, Hamburg Sikora R, Chhajed D, Shaw MJ (1996) Integrating the lot-sizing and sequencing decisions

for scheduling a capacitated flow line. Computers and industrial Engineering 30 (4): 659-679

Silver EA (1989) Shelf life considerations in a family production context. International Journal of Production Research 27 (12): 2021-2026

References 235

Silver EA (1995) Dealing with a shelf life constraint in cyclic scheduling by adjusting both cycle time and production rate. International Journal of Production Research 33 (3): 623-629

Sloan AE, McNutt KW, Powers M (1986) Consumer Attitudes About Shelflife and Tech-nology. In: Charalambous G (ed) The Shelf Life of Foods and Beverages - Proceedings of the 4th International Flavor Conference. Rhodes, Greece, 23-26 July 1985, Elsevier Science, New York et al., pp 63-72

Smith D, Sparks L (2004) Temperature Controlled Supply Chains. In: Bourlakis MA, Weightman PWH (eds) Food Supply Chain Management. Blackwell, Oxford et al., pp 179-198

Smith T, Mabe J, Beech J (1998) Components of demand planning - Putting together the details for success. In: Gattorna J, Jones T, Danks A, Dhillon Y, Holdforth L (eds) Strategic Supply Chain Alignment - Best practice in supply chain management. Gower, Aldershot et al., pp 123-137

Soman CA, van Donk DP, Gaalman GJC (2002) Combined make-to-order and make-to-stock in a food production system. http://www.ub.rug.nl/eldoc/som/a/02A49/

02A49.pdf [24.03.2003] Soman CA, van Donk DP, Gaalman GJC (2004) A basic period approach to the economic

lot scheduling problem with shelf life considerations. International Journal of Produc-tion Research 42 (8): 1677-1689

Spiller A; Engelken J, Gerlach S, Lüth M, Schramm M, Staack T, Zimmermann M (2002) Qualitätskonzepte für Nahrungsmittelverarbeitung und –handel. Gutachten für das Büro für Technikfolgenabschätzung beim Deutschen Bundestag, Georg-August-Universität, Göttingen, http://univis.uni-goettingen.de/formbot/dsc_3Danew_2Fresrep

_indepub_26dir_3Dfakagrarw_2Finstit_6_2Farbeit_3_26ref_3Dresrep [11.12.2003] Spreer E (1995) Technologie der Milchverarbeitung. B. Behr's Verlag, Hamburg Stache U (1997) Einsatz von Genetischen Algorithmen in der Produktionsplanung. Indus-

trie Management 13 (4): 38-40 Stadtler H (2002) Production Planning and Scheduling. In: Stadtler H, Kilger C (eds) Sup-

ply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 177-193

Steven M, Krüger R (2002) Advanced Planning Systems - Grundlagen, Funktionalitäten, Anwendungen. In: Busch A, Dangelmaier W (eds) Integriertes Supply Chain Man-agement - Theorie und Praxis effektiver unternehmensübergreifender Geschäfts-prozesse. Gabler, Wiesbaden, pp 169-186

Stiebing A (1984) Erhitzen - Haltbarkeit. In: Institut für Technologie der Bundesanstalt für Fleischforschung (ed) Technologie der Brühwurst. Fördergesellschaft der Bundesan-stalt für Fleischforschung, Kulmbach, pp 165-186

Stieglitz A (1998) ZP-Stichwort: Efficient Consumer Response. Zeitschrift für Planung 9: 333-339

Stölzle W, Placzek TS (2004) Gähnende Leere statt der Ware. Lebensmittel Zeitung 36: 68 Sürie C, Wagner M (2002) Supply Chain Analysis. In: Stadtler H, Kilger C (eds) Supply

Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 29-44 Tamime AY, Robinson RK (1999) Yogurt - Science and Technology, 2nd edn. Woodhead,

CambridgeTamminen M, Battezzati L, Piccirillo E (2004) What are the infrastructural constraints for

Radio Frequency Identification (RFID) implementation in Unilever's supply chain? In: van Wassenhove LN, de Meyer A, Yücesan E, Günes ED, Muyldermans L (eds) Op-erations management as a change agent. vol 2, INSEAD, pp 579-588

236 References

Tändler K (1984a) Frischware und Vorverpackung. In: Institut für Technologie der Bunde-sanstalt für Fleischforschung (ed) Technologie der Brühwurst. Förder-gesellschaft der Bundesanstalt für Fleischforschung, Kulmbach, pp 187-205

Tändler K (1984b) Rohstoffauswahl und Zusammensetzung von Brühwürsten. In: Institut für Technologie der Bundesanstalt für Fleischforschung (ed) Technologie der Brüh-wurst. Fördergesellschaft der Bundesanstalt für Fleischforschung, Kulmbach, pp 40-59

Taoukis PS (2001) Modelling the use of time-temperature indicators in distribution and stock rotation. In: Tijskens LMM, Hertog MLATM, Nicolai BM (eds) Food process modelling. Woodhead, Cambridge, pp 402-431

Tappe D, Mussäus K (1999) Efficient Consumer Response als Baustein im Supply Chain Management. HMD Praxis der Wirtschaftsinformatik 207: 47-57

TEC Group (N.N.) Software Selection RFP Templates. http://www.technologyevaluation.

com/rfptemplates/Template.asp [05.02.2004] Teixeira AA, Shoemaker CF (1989) Computerized Food Processing Operations. AVI, New

York et al. Tempelmeier H (1999a) Advanced Planning Systems. Industrie Management 15 (5): 69-72 Tempelmeier H (1999b) Materiallogistik - Modelle und Algorithmen für die Produktions-

planung und -steuerung und das Supply Chain Management, 4th edn. Springer, Berlin et al.

Tempelmeier H (2001) Supply Chain Planning with Advanced Planning Systems.http://pom-consult.de/DOC/TempelmeierTinos2001.pdf [11.03.2003]

ten Kate HA (1995) Order acceptance and production control. Labyrint, Capelle a/d IJssel Teng J-T, Chern M-S, Yang H-L, Wang YJ (1999) Deterministic lot-size inventory models

with shortages and deterioration for fluctuating demand. Operations Research Letters 24: 65-72

Thaler K (2001) Supply Chain Management - Prozessoptimierung in der logistischen Kette, 3rd edn. Fortis, Köln et al.

Thiemig F (1996a) Geflügelschlachtung. In: Sielaff H (ed) Fleischtechnologie. B. Behr's Verlag, Hamburg, pp 317-327

Thiemig F (1996b) Schlachtbetriebe und Vermarktungsformen. In: Sielaff H (ed) Fleis-chtechnologie. B. Behr's Verlag, Hamburg, pp 257-261

Thonemann U, Behrenbeck K, Diederichs R, Großpietsch J, Küpper J, Leopoldseder M (2003) Supply Chain Champions - Was Sie tun und wie Sie einer werden. Gabler, Wiesbaden

Tiemeyer E (1999) Supply Chain Management - ein neues Managementinstrument zur Unterstützung von Planungs- und Entscheidungsaufgaben. Zeitschrift für Unterneh-mensentwicklung und industrial engineering 48 (3): 100-107

Treeck S, Seishoff M (2002) CPFR in Deutschland, Österreich und der Schweiz. In: Seifert D (ed) Collaborative Planning, Forecasting and Replenishment - Ein neues Konzept für state-of-the-art Supply Chain Management. Galileo, Bonn, pp 265-277

Treutlein P, Kipp R (2002) Online-Plattform für die effiziente ERP/PPS-Auswahl. PPS Management 7 (1): 38-41

Trienekens JH, Omta SWF (2002) Paradoxes in Food Chains and Networks - Proceedings of the Fifth International Conference on Chain and Network Management in Agribusi-ness and the Food Industry. Wageningen Academic Publishers, Wageningen

Trier M (2002) IT-Unterstützung im Supply Chain Management. Industrie Management 18 (5): 58-66

Troeger K (1984) Behandlung und Bevorratung der Rohstoffe. In: Institut für Technologie der Bundesanstalt für Fleischforschung (ed) Technologie der Brühwurst. Fördergesell-schaft der Bundesanstalt für Fleischforschung, Kulmbach, pp 21-39

References 237

Tseng FT, Stafford EF Jr, Gupta JND (2004) An empirical analysis of integer programming formulations for the permutation flowshop. Omega - The International Journal of Management Science 32: 285-293

Turkki P (1994) Production Control of Integrated Meat Plants. Acta Polytechnica Scandi-navia, Helsinki

Tuttle H (2001) Manufacturing Systems Integration http://www.emgt.ku.edu/emgt867/ ses-

sions/EMGT86706A.ppt [09.04.2003] Twardawa W (2004) Verschnaufpause nach dem Aufstieg zum Hochplateau. Lebensmittel

Zeitung 16: 68 USDA, FSIS, AFDO (1999) Safe Practices for Sausage Production. U.S. Department of

Agriculture (USDA), Food Safety and Inspection Service (FSIS) and The Association of Food and Drug Officials (AFDO), Washington D.C.

van Beek P (2002) Operationele Research in actie: 26 jaar OR in Wageningen. Speech at Wageningen University, 21.03.2002

van Beek P, Beulens AJM, Meffert HFT (1998) Logistics and ICT in Food Supply Chains. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems – Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 117-138

van Boekel MAJS (1998) Developments in technologies for food production. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 87-116

van Dam P (1995) Scheduling Packaging Lines in the Process Industry. Labyrint, Capelle a/d IJssel

van Dam P, Gaalman G, Sierksma G (1993) Scheduling of packaging lines in the process industry: An empirical investigation. International Journal of Production Economics 30-31: 579-589

van der Gaag MA (2004) Epidemiological and economic simulation of Salmonella control in the pork supply chain. Dissertation at Wageningen University

van der Vorst JGAJ (2000) Effective Food Supply Chains: Generating, modelling and evaluating supply chain scenarios. Ponsen & Looijen, Wageningen

van Donk DP (2001) Make to stock or make to order: The decoupling point in the food processing industries. International Journal of Production Economics 69: 297-306

van Donk DP, van Dam P (1996) Structuring complexity in scheduling: a study in a food processing industry. International Journal of Operations & Production Management 16 (5): 54-63

van Hoek RI (1998) Reconfiguring the Supply Chain to Implement Postponed Manufactur-ing. International Journal of Logistics Management 9 (1): 95-110

van Ossel G (2002) The new marketing paradigm: manufacturer or retail brand? In: Grievink J-W, Josten L, Valk C (eds) State of the Art in Food: The changing face of the worldwide food industry. Elsevier Business Information, Arnhem, pp 609-618

van Sonsbeek JTM, van Beek P, Urlings HAP, Bijker PGH, Hagelaar JFL (1997) Mixed in-teger programming for strategic decision support in the slaughter by-product chain. OR Spektrum 19: 159-168

van Trijp JCM, Steenkamp JEBM (1998) Consumer-oriented new product development: principles and practice. In: Jongen WMF, Meulenberg MTG (eds) Innovation of Food Production Systems - Product Quality and Consumer Acceptance. Wageningen Pers, Wageningen, pp 37-66

van Wezel WMC (2001) Tasks, hierarchies, and flexibility - Planning in Food Processing Industries. Labyrint, Capelle a/d IJssel

238 References

Vandenheede J-J (2002) An outlook on an industry that's in perpetual reinvention. In: Grievink J-W, Josten L, Valk C (eds) State of the Art in Food: The changing face of the worldwide food industry. Elsevier Business Information, Arnhem, pp 45-59

Varnam AH, Sutherland JP (1994) Milk and Milk Products - Technology, chemistry and microbiology. Chapman & Hall, London et al.

Varnam AH, Sutherland JP (1995) Meat and Meat Products - Technology, chemistry and microbiology. Chapman & Hall, London et al.

Veerkamp CH (1989) Chilling, Freezing and Thawing. In: Mead GC (ed) Processing of Poultry. Elsevier Science, New York et al., pp 103-125

Vidal CJ, Goetschalckx M (1997) Strategic production-distribution models: A critical re-view with emphasis on global supply chain models. European Journal of Operational Research 98: 1-18

Vis IFA, Roodbergen KJ (2002) Examining Supply Chains from Practice. In: Klose A, Speranza MG, van Wassenhove LN (eds) Quantitative Approaches to Distribution Lo-gistics and Supply Chain Management. Springer, Berlin et al., pp 3-18

Viswanathan S (1995) A note on "Effect of production cost on shelf life". International Journal of Production Research 33 (12): 3485-3486

Viswanathan S, Goyal SK (1997) Optimal cycle time and production rate in a family pro-duction context with shelf life considerations. International Journal of Production Re-search 35 (6): 1703-1711

Viswanathan S, Goyal SK (2000) Incorporating planned backorders in a family production context with shelf life considerations. International Journal of Production Research 38 (4): 829-836

Viswanathan S, Goyal SK (2002) Technical note on "Manufacturing batch size and order-ing policy for products with shelf lives". International Journal of Production Research 40 (8): 1965-1970

von Steinaecker J, Kühner M (2001) Supply Chain Management - Revolution oder Mode-wort? In: Lawrenz O, Hildebrand K, Nenninger M, Hillek T (eds) Supply Chain Man-agement - Konzepte, Erfahrungsberichte und Strategien auf dem Weg zu digitalen Wertschöpfungsnetzen, 2nd edn. Vieweg, Braunschweig et al., pp 39-69

Voß S, Woodruff D (2000) Supply Chain Planning: Is mrp a Good Starting Point? In: Wildemann H (ed) Supply Chain Management. TCW, München, pp 177-203

Wagner M (2002) Demand Planning. In: Stadtler H, Kilger C (eds) Supply Chain Manage-ment and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 123-141

Wagner M, Meyr H (2002) Food and Beverages. In: Stadtler H, Kilger C (eds) Supply Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 353-370

Walle DL (1999) Operations Management - A Supply Chain Approach. Thomson, London et al.

Walstra P, Geurts TJ, Noomen A, Jellema A, van Boekel MAJS (1999) Dairy Technology - Principles of Milk Properties and Processes. Marcel Dekker, New York et al.

Weber B (2003) Streit um Pflicht- und Kürprogramm. Lebensmittel Zeitung 42: 26 Weber B (2004) Warenverkehr online im Blick. Lebensmittel Zeitung 30: 37 Wee H-M (1995) A deterministic lot-size inventory model for deteriorating items with

shortages and a declining market. Computers & Operations Research 22 (3): 345-356 Wee H-M, Shum Y-S (1999) Model development for deteriorating inventory in material

requirement planning systems. Computers & Industrial Engineering 36: 219-225 Werner H (2002a) Supply Chain Management - Grundlagen, Strategien, Instrumente und

Controlling, 2nd edn. Gabler, Wiesbaden

References 239

Werner H (2002b) Elektronische Supply Chains (E-Supply Chains). In: Busch A, Dangel-maier W (eds) Integriertes Supply Chain Management - Theorie und Praxis effektiver unternehmensübergreifender Geschäftsprozesse. Gabler, Wiesbaden, pp 403-419

Wessel A (2004a) Geflügel bleibt Trendprodukt. Lebensmittel Zeitung 37: 41-42 Wessel A (2004b) Aufregung auf Europas Hühnerhof. Lebensmittel Zeitung 37: 43 Wetterauer U (2002) The Implementation Process. In: Stadtler H, Kilger C (eds) Supply

Chain Management and Advanced Planning, 2nd edn. Springer, Berlin et al., pp 275-289

Whitaker D, Cammell S (1990) A Partitioned Cutting-stock Problem Applied in the Meat Industry. Journal of the Operational Research Society 41 (9): 801-807

Wietschel V (2004) Software objektiviert Listungsentscheidungen der Metro. Lebensmittel Zeitung 36: 66

Wirth F, Leistner L, Rödel W (1990) Richtwerte der Fleischtechnologie, 2nd edn. Deutscher Fachverlag, Frankfurt am Main

Wolfram G, Spalink H (2004) Tags lohnen sich für beide Seiten. Lebensmittel Zeitung 36: 58

Wouda FHE, van Beek P, van der Vorst JGAJ, Tacke H (2002) An application of mixed-integer linear programming models on the redesign of the supply network of Nutricia Dairy & Drinks Group in Hungary. OR Spectrum 24: 95-112

Wünsche K (2002) Die Nahrungsmittelindustrie - eine stabile Branche mit hohem Wachstum. In: Ernst & Young, Deutsche Investitions- und Entwicklungsgesellschaft, F.A.Z.-Institut, manager magazin (eds) Nahrungsmittelindustrie - Trends und Perspek-tiven im Food Processing. Köln, pp 9-15

Xiong YL, Mikel WB (2001) Meat and Meat Products. In: Hui YH, Nip W-K, Rogers RW, Young OA (eds) Meat Science and Applications. Marcel Dekker, New York et al., pp 351-369

Yang PC, Wee H-M (2001) An excess inventory model of deteriorating items taking ac-count of present value. The Engineering Economist 46 (2): 139-152

Yang P-C, Wee H-M (2003) An integrated multi-lot-size production inventory model for deteriorating item. Computers & Operations Research 30: 671-682

Zeier A (2002a) Ausdifferenzierung von Supply-Chain-Management-Standardsoftware in Richtung auf Betriebstypen und Branchen unter besonderer Berücksichtigung des SAP APO. Röll, Dettelbach

Zeier A (2002b) Identifikation und Analyse branchenspezifischer Faktoren für den Einsatz von Suppy-Chain-Management-Software - Teil I: Grundlagen, Methodik und Kern-anforderungen. FORWIN-Bericht-Nr.: FWN-2002-002, Bayerischer Forschungsver-bund Wirtschaftsinformatik, Bamberg et al., http://www.forwin.de/download/berichte/

Internet_FWN_2002-002.pdf [11.03.2003] Zeier A (2002c) Identifikation und Analyse branchenspezifischer Faktoren für den Einsatz

von Suppy-Chain-Management-Software - Teil II: Betriebstypologische Branchen-segmentierung. FORWIN-Bericht-Nr.: FWN-2002-003, Bayerischer Forschungs-verbund Wirtschaftsinformatik, Bamberg et al., http://www.forwin.de/download/

berichte/Internet_FWN_2002-003.pdf [11.03.2003] Zeier A (2002d) Identifikation und Analyse branchenspezifischer Faktoren für den Einsatz

von Suppy-Chain-Management-Software - Teil III: Evaluation der betriebs-typologischen Anforderungsprofile auf Basis des SCM-Kern-Schalen-Modells in der Praxis für die Branchen Elektronik, Automobil, Konsumgüter und Chemie/Pharma. FORWIN-Bericht-Nr.: FWN-2002-004, Bayerischer Forschungsverbund Wirtschafts-informatik, Bamberg et al., http://www.forwin.de/download/berichte/Internet_FWN_

2002-004.pdf [11.03.2003]

240 References

Zeier A (2002e) Identifikation und Analyse branchenspezifischer Faktoren für den Einsatz von Suppy-Chain-Management-Software - Teil IV: Anwendungsbeispiel. FORWIN-Bericht-Nr.: FWN-2002-005, Bayerischer Forschungsverbund Wirtschaftsinformatik, Bamberg et al., http://www.forwin.de/download/berichte/Internet_FWN_ 2002-005.pdf

[11.03.2003] Zhang S (2004) Zweistufige Anlagenbelegungsplanung für die Produktion frischer Le-

bensmittel - Entwicklung eines Optimierungsmodells am Beispiel der Produktion von Brühwurst. Master Thesis at the Chair of Production Management, Technische Univer-sität Berlin

Zijm WHM (2000) Towards Intelligent Manufacturing Planning and Control Systems. OR Spektrum 22: 313-345

Zikakis JP (1986) Factors affecting the Shelf Life of Dairy Products. In: Charalambous G (ed) The Shelf Life of Foods and Beverages - Proceedings of the 4th International Fla-vor Conference. Rhodes, Greece, 23-26 July 1985, Elsevier Science, New York et al., pp 313-334

Zuckerman A (2002) Supply Chain Management. Capstone, Oxford Zweig K, Sielaff H (1996) Ausrüstungen für die Fleischverarbeitung. In: Sielaff H (ed)

Fleischtechnologie. B. Behr's Verlag, Hamburg, pp 217-256

Advanced planning...

Food

Transcript of Advanced planning...