ARTICLE IN PRESS

0925-5273/$ - se

doi:10.1016/j.ijp

�CorrespondiE-mail addre

cathal.heavey@

Int. J. Production Economics 103 (2006) 420–437

www.elsevier.com/locate/ijpe

The impact of information sharing and forecasting incapacitated industrial supply chains: A case study

P.J. Byrne�, Cathal Heavey

Department of Manufacturing & Operations Engineering, University of Limerick, Ireland

Received 22 July 2004; accepted 20 October 2005

Available online 20 February 2006

Abstract

This paper models and analyses the effect of information sharing and forecasting on the performance parameters of an

actual industrial supply chain consisting of Small-To-Medium sized enterprises. The paper reports on the industrial supply

chain studied, which was undergoing a Business Process Re-engineering (BPR) exercise. The aim of the BPR exercise was

the streamlining of existing unstructured processes, ultimately culminating in the introduction of an ERP system into the

organisation to improve information sharing between the supply chains echelons.

The paper reviews previous work in this area and expands this work to address the issues posed by a more complex real

industrial example. The model itself has been developed for a complex supply chain structure. This supply chain has

multiple customers, distributors and product families, with customers and distributors face differing demand patterns.

This model and its associated experimentation highlights the significant benefits achievable through the use of improved

information sharing and forecasting techniques on the supply chain performance parameters. Potential total supply chain

cost savings of up to 9.7% have been shown, with increased savings occurring with reduced system capacity. The model

also quantifies the impact of collaboration between all partners in the study and shows that gains are achievable by all

parties in the supply chain.

r 2006 Elsevier B.V. All rights reserved.

Keywords: Supply chain; Business process re-engineering (BPR); Discrete event simulation; Decision support system

1. Introduction

Supply Chain Management (SCM) is made up ofthe control of both material and information flowamong suppliers, manufacturers, distributors andcustomers. SCM involves the management of theseflows both within and between companies andorganisations. However, to coordinate the supply

e front matter r 2006 Elsevier B.V. All rights reserved

e.2005.10.007

ng author. Tel.: +353 61 202948.

sses: pj.byrne@ul.ie (P.J. Byrne),

ul.ie (C. Heavey).

chain, it is necessary for these supply chain partnersto share information. It is widely recognised thatadvances in technologies in the areas of informa-tion, manufacturing, and distribution systems havedriven much change through the supply chain andlogistics management services. This has particularlybeen the case with improving information technol-ogy enabling instantaneous global informationsharing with more powerful information processing.Traditionally, the management of information hasbeen somewhat neglected. The method of informa-tion transfer and forecasts carried out by members

.

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 421

of the supply chain consisted of placing ordersto the member directly above them, termed theirpreceding ‘echelon’. This causes many problems inthe supply chain. These included excessive inventoryholding and shortages, increased lead times andreduced service levels. In addition, increased de-mand variability or the ‘Bullwhip Effect’, com-pounded forecasting problems and led to difficultiesin echelons further up the supply chain.

Thus, as SCM progresses, supply chain managersare realising the need for utilising improvedinformation sharing and forecasting throughoutthe supply chain in order to remain competitive.This is because companies increasingly operate andcompete in an ever-expanding global economy. It iswidely recognised from studies in the area thatimproving information sharing, forecasting andgeneral supply chain collaboration will lead tosupply chain gains. For some recent papers advo-cating these points the interested reader is referredto the following: Yao et al. (2005), Holweg et al.(2005), Lee et al. (2000), Kelle and Akbulut (2005),Chandra and Grabis (2005) and Liberopoulousand Koukoumialos (2005) However, Raghunathan(2001), has suggested that the study by Lee et al.(2000) overestimates the benefit of demand informa-tion sharing by the retailer in the two-level supplychain studied. They suggest that the reason for theoverestimation is the assumption that the manufac-turer uses only the current period’s retailer orderquantity to forecast that of the next period.However, they do accept that information sharingis useful in this model when there is an element ofdemand which cannot be predicted by the manu-facturer using order history. Things such aspromotions, price reductions and advertisementsby the retailer can cause these demand uncertainties.Many of the studies found in the literature haveanalysed simplified supply chains using, in themajority of cases, analytical techniques. Whileanalytical models are computationally efficient, theytend to be highly modified versions of reality inorder for them to be tractable. Such models can beenvisaged as being restrictive in an industrial settingand are therefore only useful to gain simple insights.For such studies to be of real industrial use, theymust have the ability to incorporate the detailsfound in today’s detailed supply networks and havethe flexibility to change within a changing environ-ment. In other words, such models should bereusable and adaptable by operational personnelon an ongoing basis.

The purpose of this study is to explore the effectof utilising information sharing and forecasting in acapacitated supply chain, where minimal forecastingand information had been shared in the past. Themodel developed is that of a real industrial supplychain, with multiple products and multiple echelonswith significant product interactions. In this paper,we consider two information sharing strategies, twoforecasting techniques and three capacity levels thatwere being examined by the case study company inquestion. In order to model the situation in detail, acomputer simulation model of the supply chain hasbeen developed. The aim of this study is to expandexisting theoretical studies to include a real-lifecomplex case study in an attempt to providepractitioners with realistic expected performanceimprovements consequent on initiatives.

The next section reviews the related literature.Section 3 formulates the problem and outlines themodel. Section 4 details a numerical study withexperimentation on information sharing, forecast-ing and capacity constraints and Section 5 discussesthe results and concludes.

2. Literature review

With the advent of improved computing powerand the application of modern information tech-nologies, such as Electronics Data Interchangethrough integrated systems and the Internet, com-panies have the ability to share information seam-lessly (Strader et al., 1999). With information storedin a data warehouse it can be advantageous to usedecision support systems, which will help the usersapply analytical and scientific methods to decisionmaking (Bhargava et al., 1999). However, on theother hand, caution should be exercised as Ramanet al. (2001) state, ‘‘we’ve been promised a world ofsupply chain management, one free of stock-outsand overstocked warehouses. But inaccurate data issabotaging this vision.’’

The value of such information sharing in supplychains is an area which has been studied extensivelyin the literature. For an overview of the completearea, the following two papers are well-roundedreviews on the subject: Lee and Whang (2000) andHuang et al. (2003). The simplest of these systems isthe study of a dyadic supply chain where there areonly two echelons, usually a supplier and a retailer/customer. Because of the simplicity of such systemsit is easy to understand and monitor. Examples ofpapers studying such systems start with Clark and

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437422

Scarf (1960). However, more relevant recent studiesin this area have been carried out by Bourland et al.(1996), Gavirneni and Tayur (1998), Gavirneni et al.(1999), Cachon and Larviere (2001) and Gavirneni(2002). Modelling this type of system allows thepractitioner to look at certain system parameterssuch as capacity, inventory levels, service levels andcosts. However, these models are very restrictive inrelation to their size, flexibility and realism leadingto a growing awareness that a more holisticapproach is needed for studying such structuresand the importance of information sharing andforecasting on them. Huang et al. (2003) identifiedfive different types of supply chain, the dyadicstructure, the serial structure, the divergent struc-ture, the convergent structure and the networkstructure, which represents a complex supply chain.

Lee et al. (2000) showed that there were manydifferent types of information that can be sharedbetween supply chain partners. Their study listed anumber of general classifications of informationthat are currently being shared across a wide rangeof industries and firms. These include the (1)Inventory Level/Position, (2) Sales Data/DemandInformation, (3) Order Status for Tracking/Tracing,(4) Sales Forecast and (5) Production/DeliverySchedule. The benefit of such information sharinghas been shown in a number of papers. Gaonkarand Viswnadham (2001) conducted experimentson information sharing in supply chains using amathematical model based on a system in acontract-manufacturing environment with colla-boration. The model aimed at quantifying theimpact of collaboration by comparing the costs ofa traditional make-to-stock (MTS) supply chainwith no information sharing between partners and amake-to-order supply chain with ubiquitous infor-mation sharing. They found that informationsharing had a significant impact on supply chainprofitability as a whole when compared to a MTSsupply chain with no information sharing. Thepaper concluded by listing a number of possibleextensions to their work including making morerealistic, less restrictive assumptions. The additionof a forecasting model into the supply chainoptimisation model was proposed. Further, it wassuggested that discrete event simulation studiescould be carried out to investigate lead-time andinventory behaviour with or without informationsharing.

To improve the performance of a supply chain, itis suggested that companies in the chain share

information and coordinate orders (Lee et al.,2000). Zhao et al. (2001), Zhao and Xie (2002)and Zhao et al. (2002a, b) present a series of papersstudying part of a hypothetical supply chainconsisting of one supplier and four retailers. Thesemodels proposed and evaluated information sharingamong retailers and suppliers, using differingdemand forecasting techniques, where the supplieris assumed to be a manufacturer with a fixedproduction capacity to produce a single product forthe retailers. All retailers are assumed to haveidentical demand distributions with the sameaverage demand per period for each retailer. Acomputer simulation model developed in the C++programming language is used to evaluate theassociated cost to the supply chain with this onecapacitated supplier and multiple retailers. Thesepapers analysed the effects of demand forecastingand inventory replenishment decisions by theretailers, and production decisions by the supplierunder different demand patterns and capacitytightness (CT). They conclude that the selection ofa forecasting model, the methods of informationsharing, demand patterns faced by the retailers andCT faced by the supplier significantly affect overallsupply chain performance metrics.

In one such study presented by Zhao et al.(2002a, b), the independent parameters experimen-ted with were the:

�

Demand pattern, � Forecasting methods, � Supplier’s capacity, � Information sharing method between retailers

and suppliers.

A number of performance measures were col-lected, which included: (i) Total cost for retailers, (ii)Total cost for the supplier, (iii) Total cost for thesupply chain, (iv) The service level of the supplierand (v) The customer service level of the retailers. Itwas found as part of this study that the supplierenjoys a greater cost reduction than the retailers as aresult of sharing information regardless of theforecasting model used and the higher the forecast-ing accuracy, the greater the value of informationsharing. Comparison of forecasting model perfor-mance under different levels of information sharingindicates that more accurate forecasts may not helpto improve the performance of the supply chaindramatically when the retailers do not shareinformation with the supplier. Information sharing

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 423

can significantly influence the performance of asupply chain.

It can be seen from the literature studied thatalthough numerous studies have been carried out todetermine the value of information sharing and theuse of forecasting in supply chains, there is littlework published on real systems. Many of themodels have been completed using analyticalmodelling, which, by its nature, is quite restrictive.The results in this paper have been obtainedthrough the study of a real industrial case studyusing discrete event simulation. This is an attemptto study the effect of both information sharing andforecasting on the supply chain performance of thisreal system in an attempt to bridge the gap in theliterature to what is happening in reality. As such,the model developed has been populated with realdata and the experiments carried out in relation toinformation sharing, forecasting, stocking policiesand capacity constraints are conditions/experimentsthat have been put forward by the companythemselves in relation to both their current anddesired operating policies.

3. Problem formulation

The organisation involved in the study may beclassed as a vertically integrated raw materialsupplier, offering a complete range of its relatedindustries products to a global marketplace. Scharyand Skjøtt–Larsen (1995) define vertical integrationas ‘‘the ownership by one organization of otherfirms in its supply or distribution network. Thetotally integrated firm is completely self-sufficient.The non-integrated firm is completely dependant

BufferStock

Bulk Processing Processin

Inventorystorage

point

BulkStock

Fig. 1. Supply cha

on market forces and other organizations for itsoperations.’’ Christopher (1998) states, ‘‘Verticalintegration normally implies ownership of upstreamsuppliers and downstream customers.’’

Company X’s supply chain structure is shown inFig. 1. It can generally be regarded as an extensionof the well-studied serial multi-echelon supplychain, being supplied by a number of raw materialsuppliers and supplying a number of distributors,who in turn supply individual customers. Althoughthere are a number of echelons prior to the bulkstock inventory location, they are not examined inthis study. It is felt that this is a minor restriction, asin the real system material is always available forprocessing at this point due to the use of productsubstitutions. Company X operates a MTS systemthroughout its supply chain. The costs used in themodel are not those directly used by Company X

and are approximations. These approximations takeinto account the typical costs associated with thetransactions in such a supply chain and we highlightthe potential savings associated with using thesecosts. However, as these costs are typical of such asystem, they provide a useful basis for the perfor-mance evaluation of this supply chain case study.

In order to provide general results for the supplychain structure studied, the following independentparameters are experimented on:

Demand pattern: Historical customer demandfrom each of the eight distributors feeding intoCompany X for 16 different product classes has beenstudied using the actual demand as experienced byeach distributor for a period of 1 year. The 16classes consist of eight product families (A, B, C, D,E, F, G and H) with each of the families’ further

g

1

x

3

2

Distributors

FinishedStock

1

n

2

1

n

2

Customers

Set ofoperations

in structure.

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437424

sub-divided into two categories (S and R) dependingon the product routings for the part.

Forecasting models: The impact of two differentforecasting methods will be studied. The twomethods studied are the simple moving average(SMA) method and the double exponential smooth-ing (DES) method. These two methods are beingstudied as they have been found to be the best inrelation to Company X’s demand as investigated byCompany X. Also included will be actual (ACT)demand for baseline comparisons.

Information sharing: Two different types ofinformation sharing will be considered: (1) Fulldemand information sharing and (2) No informa-tion sharing as described in detail in Section 4.

Capacity tightness: It is assumed that the capacityrequired to produce 1 unit of product is 1 resourceunit in both the processing and bulk processingareas. The CT indicates the ratio of the totalavailable capacity to the total capacity needed.Three levels of CT are used: Low (1.33), Medium(1.18) and High (1.05), which correspond toresource utilisation of 75%, 85% and 95%,respectively.

Manufacturing lead-time: This is the time it takesto manufacture product, taken as a general lead-time for product families.

The performance measures used to analyse theseindependent parameters are the supply chain costsof the distributors and the company as well asthe overall supply chain costs for each experiment.In addition, the On-Time-In-Full (OTIF) resultsfor four critical products (A-S, A-R, B-S and B-R)and the overall system OTIF are analysed withthe overall average finished stock levels. The OTIFis calculated as the percentage of orders recei-ved into finished stock that can be fully satis-fied from finished stock, as the company areoperating a MTS policy. In order to calculate thesupply chain costs a number of supply chainoperations and inventory holding costs were estab-lished. The costs were broken down into fivedifferent categories:

Transportation: This is the cost of shipmentsbetween finished stock in Company X and thestocking location in the distributor. This cost isborne by the distributor and it is assumed that theproduct is sufficiently small and the mode oftransport is large enough to only require a once-off cost per shipment, which varies depending onthe distributor’s location. The costs used in thismodel are h500, h550, h250, h650, h350, h250, h575

and h350 for the eight distributors, respectively. Ifmore than one product is transported to adistributor in one period it is assumed that theorder is consolidated and shipped as 1 unit, henceonly incurring a single transportation fee per periodwhere shipment occurs.

Order processing: There is an order processingcost associated with an order being received at boththe processing and bulk processing locations, whichhas been set at h50 and h10, respectively, in thisstudy. This cost is borne by Company X.

Production setup: There is a production setup costassociated with an order being setup in both theprocessing and bulk processing areas. This has beenset at h10 and h55 per order, respectively, in thisstudy. This cost is borne by Company X.

Inventory cost: This is the cost of holding stockfor one period. The costs used in this model areh12, h13, h10, h14, h11, h12, h14 and h10 for theeight distributors, respectively, for product typeR. An estimate is used for the S types, as theseproducts are derivatives of the R types and onaverage cost 1% of the cost of the R type. The costsfor holding one unit of finished stock for one periodexperienced by company X are h0.50, h8 and h5 forfinished stock type S, finished stock type R andbuffer stock, respectively. The varying inventoryholding costs in the distributors have to do withdiffering operating conditions in differing countries,which include such things as rates, cost of labour,cost of electricity, etc.

Backorder cost: This is the cost of backordersfor one period. The costs used in this modelare h20, h21, h18, h17, h19, h22, h22 and h20for the eight distributors, respectively, and h4, h21and h8 for finished stock type S, finished stocktype R and buffer stock, respectively, for eachbackordered order or partial order. Again, thevarying backorder costs in the distributors is relatedto the different operating conditions in differingcountries.

4. Model description

The simulation package chosen for the studywas eM-Plant (www.tecnomatix.com). There were anumber of reasons for this choice. Firstly, thepackage is object orientated, which allows for easyreplication of objects (such as work centres, etc.).Secondly, its ability to use an ODBC link (OpenDatabase Connectivity Link) to connect the simula-tion model directly to an information database.

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 425

Thirdly, the flexibility offered by eM-Plant in rela-tion to its programming language (Sim Talk) tocustomise objects to realistically resemble theexisting physical supply chain. This was especiallyuseful when modelling the manufacturing functionand for examining the effect of these functions onthe overall supply chain performance.

The following section will outline the operation ofthe supply chain model as outlined in Fig. 1 for 1year, for each replication of the model. Each objectin the model will be described in detail. The model isto be run on a MTS policy.

4.1. Customers

Customer demand for each distributor is gener-ated using 1 year of historical data obtained fromeach of the eight distributors. An example of thedemand received by one of these distributors forboth categories for two product families (B-S, B-R,C-S and C-R) can be seen in Fig. 2. This customerdemand is fed into the distributors on a period-by-period basis assuming a 5-day working week in boththe distributors and the customers, in relation toprocessing demand orders.

B-S

0

50

100

150

200

250

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

C-S

0

500

1000

1500

2000

2500

3000

3500

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

Fig. 2. Product demand for produ

4.2. Distributors

The distributor in this model holds information oneach distributor in relation to that distributor’sdemand, as it is received from the customer, transittimes for parts and its current stocking situation.Distributors in this model (as they do in reality in thiscase) are controlled by an (R, nQ) inventory policy.When using an (R, nQ) policy an order is triggered assoon as the inventory position ( ¼ inventory on-hand+orders outstanding�backorders) is at orbelow the reorder point R. The order is for nQ

units where n is the minimum integer required toincrease the inventory position to above R. R isthe reorder point and Q the base order quantities(see Table 1). The safety stock held is equal tothe average lead-time demand for that distributor,which is equal to the average amount of stockdemanded over the distributor’s transit lead-time.The reorder point in this study has been set at threetimes the lead-time demand, as recommended byCompany X and as such varies from distributor todistributor.

Table 2 presents an example of the distributorreorder points for Product Y. The transit lead-time

B-R

0

500

1000

1500

2000

2500

3000

3500

4000

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

C-R

0

50

100

150

200

250

300

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

cts B and C in Distributor 8.

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437426

for each distributor is given, as is the yearly demand(260 periods) for Product Y. For example, it takesfour periods to ship products from Company X todistributor 1. The average periodic demand for eachdistributor is equal to their yearly demand forProduct Y divided by 260 (periods in a year). Theaverage lead-time demand is equal to the averagedemand per period multiplied by the transit lead-time for that distributor. This is then rounded to thenearest whole product and is referred to as thesafety stock. The reorder point is similarly calcu-lated by multiplying and rounding the lead-timedemand, in this case by three (as decided byCompany X) to generate the appropriate reorderpoint for each distributor for Product Y. Althoughthis is the policy used by the distributors it is limitedin that it cannot cope well with spikes in demand(i.e. large unexpected orders). However, taking thenature of the business into consideration, this isnot a major problem as customers are generallyprepared to wait on multiple shipments for theselarge orders. Customers and distributors are moreconcerned with receiving their small to mediumorders both on time and in full.

Table 1

Product Q values

Family Product A Product B Product C Product D

Product type A-S A-R B-S B-R C-S C-R D-S D-R

Q 100 1 150 1 100 1 85 1

Family Product E Product F Product G Product H

Product type E-S E-R F-S F-R G-S G-R H-S H-R

Q 75 1 90 1 200 1 250 1

Table 2

Distributor reorder point examples for Product Y

Distributor Transit lead-time

(periods)

Yearly demand Averag

per per

1 4 500 1.92

2 2 0 0.00

3 3 1025 3.94

4 2 2354 9.05

5 1 95 0.37

6 3 63 0.24

7 3 159 0.61

8 1 802 3.08

4.3. Information sharing

In this study, two levels of information sharingare used. The first type is where no direct informa-tion (NO INFO) is shared between the partners inthe supply chain. In this case, the distributors onlysend required orders back to finished stock person-nel when they manually review their own stocklevels. This means that Company X must plan theirproduction levels based on actual orders receivedwith no insight into each distributor’s currentdemand profiles. For the purpose of this study, anorder frequency term was created for each of thedistributors to represent this random demandreview profile. From past experience Company X

determined that, in general, distributors manuallyreview their stock levels between one and fiveperiods but they are not sure how often with anaverage of three periods. Therefore, as Company X

has no more detailed information on this atriangular distribution is used to determine thiscontinuously changing review period.

The second type of information sharing is the caseof full information sharing (INFO). In this case, thedistributor shares its requirements directly andcontinuously with the finished stock personnel inCompany X. This is indicative of an entire demandsystem, which is coordinated by an AdvancedPlanning System (APS). In this case, Company X

has full insight into both the customer’s demandprofiles and the requirements from each distributor.Armed with this information Company X can baseits production plans directly on actual orders.

4.4. Forecasting

In this study, Company X produces accordingto its forecasting module. From this forecast, they

e demand

iod

Average lead-time

demand

Safety

stock

Reorder

point

7.69 8 23

0.00 0 0

11.83 12 35

18.11 18 54

0.37 0 1

0.73 1 2

1.83 2 6

3.08 3 9

ARTICLE IN PRESS

B-R

0

1000

2000

3000

4000

5000

6000

7000

0 25 50 75 100 125 150 175 200 225 250Period

0 25 50 75 100 125 150 175 200 225 250Period

Dem

and

C-S

0

500

1000

1500

2000

2500

3000

3500

Dem

and

Fig. 3. Company X product demand (products B-R and C-S).

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 427

predict what the next period’s demand will be foreach individual product family, and attempt toproduce this one period in advance (CT cansometimes restrict production, and will be explainedin detail later). For this study, a benchmarkingtechnique and two forecasting techniques were used.The baseline technique consisted of using the actualnext period’s demand (ACT) to plan production(in reality this is not possible but is used as abenchmarking technique). In other words, thedemand for time period j+1 was used to planproduction in time period j. The two forecastingtechniques used were SMA and DES. These twoforecasting techniques were used as they are ofinterest to Company X. A moving average of fourperiods was used for the SMA. In relation to DES,Monks (1987) stated that low Alpha (a) values areparticularly appropriate when product demand isrelatively stable, which is not the case in this study.Higher values of a are more useful where sub-stantive changes are likely to occur because they aremore responsive to fluctuations in demand. Simi-larly, a low Beta (b) value will give more smoothingof the trend and may be useful if the trend is notwell established. Taking these points into accountan a value of 0.7 and a b value of 0.1 was used inthis study. In addition to this, sensitivity analyseswere carried out using a values of 0.7, 0.8 and 0.9and b values of 0.1, 0.2 and 0.3 (see Section 5). Inthe validation and verification section a number ofdifferent moving average, a and b values werereviewed to determine the system’s sensitivity tochanging these forecast parameters. This forecast isupdated each period as new demand informationis received and is used to plan production inCompany X.

4.5. Demand

The demand object is the main control element inthe model and manages the forecasts for eachproduct family for each different distributor. Thedemand object also reviews current stock levels infinished stock each period and determines theproduction requirements for each product family,which have fallen below their reorder points. Thisobject also manages incoming material to finishedstock and allocates material backordered to relevantdistributors. Taking all these factors into account(and CT) the demand object determines how manyunits of each product family will be produced andhow they will be allocated to different distributors

in the event of capacity issues. If there are capacityissues then parts are issued to the distributors inproportion to their initial order, with the remainderbeing backordered. As a MTS policy is beingoperated, the OTIF is calculated at this point. If ademand order for a distributor is fulfilled comple-tely then that order is said to be OTIF. Otherwise, itis said to be NOT OTIF and this is reviewed foreach product family in Company X. Each experi-ment created a slightly different demand patternexperienced by company X due to differing orderingfrequencies and production rates. Fig. 3 highlightsthe two distinct demand profiles as experienced bycompany X for one such experiment. It can be seenfrom these that product demand for B-R has aregular demand whereas product demand for C-S ismuch less frequent with high spikes. The productdemand for one full experiment received into thefinished stock location for all 16 different productclasses can be seen in the appendix. The informationshown in the graphs has been rescaled for con-fidentiality.

4.6. Stocking locations—bulk, buffer and finished

stock

As stated earlier, for this study it is assumed thatthere is sufficient bulk stock at all times to meetdemand at this point in the model. The buffer

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437428

stock location stores products after they leave bulkprocessing and before they go through finish pro-cessing. The finished stock location stores productsawaiting request from distributors. The finishedstock and buffer stock locations operate in a similarfashion. They are both operating a reorder point,reorder quantity model similar to that used by thedistributors. In these locations, the safety stock heldfor each product family is equal to the historicaverage period demand for that product familymultiplied by a number of safety stock periods. Inthis study, the safety stock periods for the bufferand finished stock locations are four periods. Thereorder points for each product family in these twolocations is equal to the safety stock plus theaverage production lead-time demand. The produc-tion lead-times, as can be seen in Table 3, wereprovided by Company X. These values wereestimates of the average lead-times for each productfamily through each of the processing areas. As withthe distributors, the reorder quantity in eachlocation uses the (R, nQ) policy as describedpreviously.

4.7. Processing areas—finish processing and bulk

processing

It is assumed that it takes 1 unit of a resource toproduce one product. The total historical demand isdetermined for each product family and hence thetotal capacity required producing this demand. For

Table 3

Production lead-times (periods) for each product family

Product Process lead time Bulk lead time

A-S 5

A-R 1 8

B-S 5

B-R 1 6

C-S 6

C-R 2 5

D-S 5

D-R 3 7

E-S 5

E-R 4 10

F-S 4

F-R 2 8

G-S 12

G-R 3 16

H-S 13

H-R 7 9

each processing area there is an associated CT.This is used to generate the available capacity foreach period, which indicates the ratio of the totalavailable capacity to the estimated total capacityneeded. The total capacity available for all theperiods is equal to the estimated total capacityneeded multiplied by the CT factor, where it isassumed that the capacity is evenly distributed overall periods.

When an order is received into the finishprocessing area it checks the availability of bufferstock to produce this order. If sufficient stock isavailable, it will start producing the order immedi-ately, which will move into finished stock after theallocated lead-time for that part. If there is onlystock available to produce part of the order thenthis stock is removed from buffer stock and thatportion is processed, with the remainder placed in abackorder queue. If no stock is available in bufferstock then the entire order is placed in thebackorder queue. The bulk processing area operatesin a similar fashion with the exception of back-ordering products, as it is assumed that there isalways sufficient material to complete any orderreceived.

5. Verification and validation

The simulation model was developed in eM-Plant, an object-orientated simulation package andrun on a standard Dell workstation. As the modelwas being developed, it was continually validatedand the codes verified. The model itself wasdeveloped in a modular fashion starting with thecustomer module and working back through thesupply chain in a logical way. As each module wasdeveloped the inputs and outputs for that selectionwere analysed and verified and where possible thesevalues were confirmed through manual calculations.After each of the modules were individually verifiedthe entire model was run and orders for products,product movement and stocking levels were mon-itored and traced for one complete run throughoutthe entire supply chain. Animation was also used inthe process to follow the flow of these productseasily.

The simulation model was run for 1 year (260periods) on the principle of a 5-day working week,with no product operations or order processingbeing carried out on Saturday or Sundays. Perfor-mance measurements are truncated for the first 20periods (equal to 4 weeks) in the model to allow the

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 429

system to reach steady state. In addition to thistruncation, periods of 10, 15, 25 and 30 periods werealso tested, which resulted in similar findings tothose presented here.

In addition, sensitivity analysis was performed onthe forecasting parameters using varying factorsaround those suggested by Company X for both theSMA and the DES forecasting techniques. Whenusing the SMA forecasting technique the model wasrun using a moving average of four periods and an avalue of 0.7 and a b value of 0.1 were used whenusing the DES forecasting technique. In order totest the sensitivity of these two forecasting techni-ques, the model was run with varying forecastinginputs as shown in Tables 4 and 5 for the systemusing information sharing set to NO INFO, and aCT of 1.05 and 1.33. The output analysed is thedistributors’ supply chain cost (Dist Totals), thecompany’s supply chain cost (Comp Totals) and thetotal supply chain cost (Totals). The outputs showthat, given the current operating conditions of thesystem, the forecasting factors (moving average, aand b values within the given ranges) themselves donot have a significant effect on the performance ofthe model with the type of demand sets being used.Therefore, the forecasting parameters as determinedby Company X were used to produce the resultsdescribed in this paper.

6. Supply chain results

Eighteen experiments, as shown in Table 6, werecarried out using the above described model todetermine the effect of information sharing, fore-casting and CT on the overall supply chainperformances in the described industrial case study.The performance measures were the supply chain

Table 4

Forecasting—simple moving average supply chain costs with varying in

Simple moving average Moving average

SMA CT ¼ 1.05 4 5

NO INFO Totals 121,233,333 124,169,858

Dist Totals 44,470,287 46,041,643

Comp Totals 76,763,046 78,128,215

SMA CT ¼ 1.33

NO INFO Totals 67,839,461 67,607,876

Dist Totals 25,411,606 24,822,549

Comp Totals 42,427,855 42,785,327

costs of the distributors and the company as well asthe overall supply chain costs for each experiment.In addition, the OTIF results for the four criticalproducts (A-S, A-R, B-S and B-R) and the overallsystems OTIF are analysed with the overall averagefinished stock levels.

Although the model presents the data in relationto the individual transportation costs, the inventorycosts and the backorder costs for each distributor,this paper will present the overall distributor supplychain costs. In the same way, the company costscould be sub-divided into order processing, setup,inventory and backorder costs for each of thedifferent processing areas, but in this study, it isthe overall company supply chain costs that arepresented. These results can be seen in Tables 7and 8.

6.1. Supply chain costs

The supply chain costs for each set of experimentshas been analysed separately and the results can beseen in Table 9 and Fig. 4. It can be seen from theseresults that the capacity in the supply chain has thebiggest impact on the supply chain costs for boththe distributors collectively and for the companyand hence for the entire supply chain. Although it isimpossible to know what actual demand is going tobe prior to receiving orders, the actual demand is auseful benchmark to show us how well the system isoperating as compared with a perfect, but unrealis-tic, situation. It can be seen that the SMA forecastoperated best under the given demand conditions,with the exception of the case where capacity is at itsmaximum (CT ¼ 1.33) and information sharing isshared between one and five periods. In this case(experiments 12 and 18), the DES forecasting

put parameters

7 9 11 13

121,749,942 125,871,177 123,365,204 121,393,184

44,334,831 45,999,322 45,268,756 44,575,690

77,415,111 79,871,855 78,096,448 76,817,494

71,259,087 66,811,429 69,072,155 66,322,614

27,034,267 25,023,159 26,233,784 24,323,012

44,224,820 41,788,270 42,838,371 41,999,602

ARTICLE IN PRESS

Table

5

Forecasting—

double

exponentialsm

oothingsupply

chain

costswithvaryinginputparameters

Double

expsm

ooth

Alpha

0.7

0.8

0.9

DES

CT¼

1.05

Beta

0.1

0.2

0.3

0.1

0.2

0.3

0.1

0.2

0.3

NO

INFO

Totals

123,620,430

125,928,611

121,264,532

131,063,042

131,398,885

127,969,572

125,195,648

128,013,378

127,963,110

DistTotals

45,054,254

45,932,388

43,689,628

48,821,416

48,697,478

47,049,329

46,072,986

47,433,080

47,371,381

CompTotals

78,566,176

79,996,223

77,574,904

82,241,626

82,701,407

80,920,243

79,122,662

80,580,298

80,591,729

Double

expsm

ooth

Alpha

0.7

0.8

0.9

DES

CT¼

1.33

Beta

0.1

0.2

0.3

0.1

0.2

0.3

0.1

0.2

0.3

NO

INFO

Totals

66,428,797

68,739,172

69,816,363

70,451,716

70,145,943

67,709,713

69,728,026

69,910,145

70,474,313

DistTotals

24,867,521

26,346,650

26,463,430

26,736,045

27,136,159

25,668,904

26,798,563

26,643,089

26,670,879

CompTotals

41,561,276

42,392,522

43,352,933

43,715,671

43,009,784

42,040,809

42,929,463

43,267,056

43,803,434

Table 6

Model experiments

Exp Model conditions

Forecast Bulk CT Proc CT Information sharing

1 ACT 1.05 1.05 INFO

2 ACT 1.05 1.05 NO INFO

3 ACT 1.18 1.18 INFO

4 ACT 1.18 1.18 NO INFO

5 ACT 1.33 1.33 INFO

6 ACT 1.33 1.33 NO INFO

7 SMA 1.05 1.05 INFO

8 SMA 1.05 1.05 NO INFO

9 SMA 1.18 1.18 INFO

10 SMA 1.18 1.18 NO INFO

11 SMA 1.33 1.33 INFO

12 SMA 1.33 1.33 NO INFO

13 DES 1.05 1.05 INFO

14 DES 1.05 1.05 NO INFO

15 DES 1.18 1.18 INFO

16 DES 1.18 1.18 NO INFO

17 DES 1.33 1.33 INFO

18 DES 1.33 1.33 NO INFO

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437430

technique was superior. From these results, it can beseen that choosing the correct forecasting techniquefor the situation experienced in this case leads to areduction in supply chain costs in the ranges1.2–4%, 0.3–2.3% and 0.3–2.9% for the distribu-tors, company and total supply chain costs,respectively, with an average saving of h500,000for both the distributors and the company alike.Although selecting the correct forecasting techniqueprovides distinct benefits the use of improvedinformation sharing techniques has a much moresignificant impact on the supply chain costs for boththe distributors and the company. It can be seenfrom the results of the experiments carried out onthis particular supply chain that it is possible toachieve reductions in the distributors’ supply chaincosts in the range of 5.5–9.7%, with the companyexperiencing reductions in the range 3.3–6.3%, withmaximum savings of h3.5 million and h5 million forthe distributors and company, respectively, fromoperating a seamless demand information sharingprocess. As one would expect, these savings areexperienced when using a CT of 1.05 wheninformation sharing is extremely important. Whenusing a CT of 1.33 savings of h1.3 million and h1.6million for the distributors and company, respec-tively, were achieved using the same policies asabove.

ARTICLE IN PRESS

Table 7

Overall supply chain costs (h)

Exp Distributor supply chain costs Company supply chain costs Overall supply

chain costTransport Inventory Backorder Total Order proc Prod setup Inventory Backorder Total

1 755,025 6,549,077 31,568,026 38,872,128 180,010 430,755 12,702,463 54,324,731 67,637,959 106,510,087

2 722,275 5,696,184 31,266,980 37,685,439 175,780 422,015 12,898,772 52,940,259 66,436,826 104,122,265

3 692,275 8,167,420 19,801,652 28,661,347 173,160 417,055 15,272,353 32,398,676 48,261,244 76,922,591

4 683,575 6,926,958 24,236,641 31,847,174 165,320 401,060 15,152,524 35,338,958 51,057,862 82,905,036

5 632,650 9,120,278 12,553,051 22,305,979 163,080 389,765 16,982,105 19,925,263 37,460,213 59,766,192

6 613,275 7,696,408 15,336,311 23,645,994 153,810 371,480 16,785,369 21,469,112 38,779,771 62,425,765

7 765,700 6,046,410 34,038,656 40,850,766 180,520 431,460 11,621,311 60,449,198 72,682,489 113,533,255

8 757,525 4,846,181 38,866,581 44,470,287 176,550 423,000 11,639,228 64,524,268 76,763,046 121,233,333

9 701,125 7,765,647 21,635,814 30,102,586 174,120 418,660 13,815,838 37,543,767 51,952,385 82,054,971

10 690,475 6,511,450 25,178,268 32,380,193 168,670 407,410 13,753,141 39,421,033 53,750,254 86,130,447

11 649,875 8,703,704 13,870,734 23,224,313 165,480 394,315 15,393,458 24,114,614 40,067,867 63,292,180

12 643,775 7,432,047 17,335,784 25,411,606 157,080 377,165 15,336,981 26,556,629 42,427,855 67,839,461

13 769,525 5,954,170 34,778,560 41,502,255 177,730 425,915 11,435,940 61,542,214 73,581,799 115,084,054

14 755,675 4,726,931 39,571,648 45,054,254 175,120 420,135 10,961,466 67,009,455 78,566,176 123,620,430

15 706,925 7,701,604 22,055,161 30,463,690 168,540 407,570 13,409,557 38,082,113 52,067,780 82,531,470

16 710,975 6,157,121 26,862,612 33,730,708 164,490 398,945 12,800,722 41,623,363 54,987,520 88,718,228

17 653,450 8,643,953 14,202,066 23,499,469 157,450 378,525 14,821,110 24,611,979 39,969,064 63,468,533

18 646,100 7,256,100 16,965,321 24,867,521 151,970 367,910 14,180,684 26,860,712 41,561,276 66,428,797

Table 8

Overall supply chain performance measures

EXP A-S OTIF A-R OTIF B-S OTIF B-R OTIF Overall OTIF Average finished stock

1 68.4 51.9 78.6 40.5 42.7 1093

2 63.4 40.0 84.3 41.6 40.6 977

3 76.4 77.4 78.6 62.6 55.0 1418

4 77.1 57.0 84.9 62.3 50.8 1459

5 87.3 89.4 78.6 76.6 63.9 1578

6 76.9 72.7 80.0 70.7 57.0 1666

7 65.0 42.1 75.0 38.7 38.8 892

8 54.5 24.6 71.7 37.2 33.1 772

9 72.2 68.1 75.0 59.9 49.8 1198

10 70.5 50.8 70.6 56.1 46.3 1107

11 77.2 79.6 75.0 70.3 57.2 1333

12 74.2 65.8 72.0 64.1 52.2 1232

13 65.4 41.7 75.0 37.8 38.1 889

14 55.8 25.0 70.6 35.2 32.2 648

15 73.4 67.2 75.0 59.5 49.7 1181

16 72.5 41.8 70.0 50.4 43.3 1017

17 78.9 78.3 75.0 68.9 56.8 1325

18 77.0 55.2 73.5 66.4 49.7 1186

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 431

6.2. Supply chain performance measures

In addition to monitoring the supply chain costs,the supply chain performance is also monitored.This includes the percentage of finished stockorders that are OTIF and the quantity of finished

stock that is held. As the supply chain underinvestigation is that of a vertically integratedsystem, these are key performance metrics of theentire supply chain as this is the central hublinking the production system to the distributionsystem. For the purpose of this study, the OTIF is

ARTICLE IN PRESS

Table 9

Supply chain cost (h) breakdown

Information

sharing CT

INFO NO INFO INFO NO INFO INFO NO INFO

1.05 1.05 1.18 1.18 1.33 1.33

Distributor supply chain costs

ACT 38,872,128 37,685,439 28,661,347 31,847,174 22,305,979 23,645,994

SMA 40,850,766 44,470,287 30,102,586 32,380,193 23,224,313 25,411,606

DES 41,502,255 45,054,254 30,463,690 33,730,708 23,499,469 24,867,521

Company supply chain costs

ACT 67,637,959 66,436,826 48,261,244 51,057,862 37,460,213 38,779,771

SMA 72,682,489 76,763,046 51,952,385 53,750,254 40,067,867 42,427,855

DES 73,581,799 78,566,176 52,067,780 54,987,520 39,969,064 41,561,276

Overall supply chain costs

ACT 106,510,087 104,122,265 76,922,591 82,905,036 59,766,192 62,425,765

SMA 113,533,255 121,233,333 82,054,971 86,130,447 63,292,180 67,839,461

DES 115,084,054 123,620,430 82,531,470 88,718,228 63,468,533 66,428,797

1.05INFO

1.05NO INFO

1.18INFO

1.18NO INFO

1.33INFO

1.33NO INFO

ACT

SMA

DES50,000,000

60,000,000

70,000,000

80,000,000

90,000,000

100,000,000

110,000,000

120,000,000

130,000,000

Capacity Tightness & Info Sharing

Overall Supply Chain Costs

ACT

SMA

DES

1.05

INF

O

1.05

NO

INF

O

1.18

INF

O

1.18

NO

INF

O

1.33

INF

O

1.33

NO

INF

O

ACTSMA

DES20,000,000

25,000,000

30,000,000

35,000,000

40,000,000

45,000,000

50,000,000

Capacity Tightness & Info Sharing

Fore

cast

Distributor Supply Chain Costs

ACT

SMA

DES

1.05INFO

1.05NO INFO

1.18INFO

1.18NO INFO

1.33INFO

1.33NO INFO

ACT

SMA

DES30,000,000

35,000,000

40,000,000

45,000,000

50,000,000

55,000,000

60,000,000

65,000,000

70,000,000

75,000,000

80,000,000

Capacity Tightness & Info Sharing

Fore

cast

Fore

cast

Company Supply Chain Costs

ACT

SMA

DESC

C

C

Fig. 4. Supply chain cost breakdown (h).

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437432

analysed for the four critical products in thesupply chain (A-S, A-R, B-S and B-R) as well asthe overall OTIF for all products as can be seen in

Table 10 and Fig. 5. It can clearly be seen fromthese results that, in general, improved informationsharing has a significant positive impact on the

ARTICLE IN PRESS

Table 10

OTIF summary

Information sharing INFO NO INFO INFO NO INFO INFO NO INFO

Capacity tightness 1.05 1.05 1.18 1.18 1.33 1.33

Forecast ¼ SMA

A-S OTIF 65.0 54.5 72.2 70.5 77.2 74.2

A-R OTIF 42.1 24.6 68.1 50.8 79.6 65.8

B-S OTIF 75.0 71.7 75.0 70.6 75.0 72.0

B-R OTIF 38.7 37.2 59.9 56.1 70.3 64.1

Overall OTIF 38.8 33.1 49.8 46.3 57.2 52.2

Forecast ¼ DES

A-S OTIF 65.4 55.8 73.4 72.5 78.9 77.0

A-R OTIF 41.7 25.0 67.2 41.8 78.3 55.2

B-S OTIF 75.0 70.6 75.0 70.0 75.0 73.5

B-R OTIF 37.8 35.2 59.5 50.4 68.9 66.4

Overall OTIF 38.1 32.2 49.7 43.3 56.8 49.7

OTIF using SMA

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

1.05 1.05 1.18 1.18 1.33 1.33

INFO NO INFO INFO NO INFO INFO NO INFO

Capacity Tightness and Info Sharing

%

A-S OTIF

A-R OTIF

B-S OTIF

B-R OTIF

Overall OTIF

OTIF using DES

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

1.05 1.05 1.18 1.18 1.33 1.33

INFO NO INFO INFO NO INFO INFO NO INFO

Capacity Tightness and Info Sharing

%

A-S OTIF

A-R OTIF

B-S OTIF

B-R OTIF

Overall OTIF

Fig. 5. OTIF results.

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 433

OTIF. In a similar manner, an increase in capacityin the system also has a significant positive im-pact on the OTIF. For example, when using theSMA forecast, the OTIF increased from 24.6%(with a CT of 1.05 and no information sharing) to42.1% using improved information sharing, to65.8% using increased capacity (CT ¼ 1.33) and79.6% using both improved information sharingand increased capacity combined (INFO andCT ¼ 1.33).

7. Conclusions

This paper presents a simulation model of asupply chain, which has been used to highlight thepotential savings from utilising improved forecast-ing and information sharing techniques in a realindustrial study. The work extends earlier analyticalstudies such as those presented by Gaonkar andViswnadham (2001) and work carried out on smallhypothetical supply chains such as those presented

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437434

by Zhao et al. (2002a, b). This study extends theseidealised, but useful studies to incorporate realworld complexities into the analysis. In this case,the supply chain model was developed using adedicated simulation package. Such specific model-ling packages allow for a more flexible environmentwhere objects representing system parts can beeasily manipulated to the industrial supply chaincondition. Thus, the building blocks from such amodel can be manipulated to address numerousextensions to the current system in relation todiffering and diverse supply chain questions andstructures. With this in mind, this work extractsuseful techniques from the study, which can beapplied to a complex industrial systems andquantifies the benefits from such strategies.

It is felt that this study is not only useful in thesingle case as presented in this paper, but also in themore general sense as the results here correlate withthose results published in relation to these morerestrictive models. The model presented here in-corporates multiple products flowing through multi-ple echelons with capacity constraints in a complexnetwork structure. Each of these products hasdifferent processing times and has detailed reorder-ing policies and bills of material. This studyprovides practitioners with further proof that thereis benefit in using improved forecasting andinformation sharing techniques, but does so withreal industrial data in a complex supply chainstructure.

It can be seen from the information given in thisparticular study that both the distributors andCompany X have benefited significantly in all casesexamined and in most cases approximately equally,thus providing motivation for all participants in thesupply chain to adopt and cooperate with improvedinformation sharing and forecasting techniques andprocesses. From analysis carried out on the resultsfrom this study it was found that the informationsharing technique has a greater impact on thesupply chain performance particularly in relation tothe potential supply chain cost savings. Thedistributors experienced up to 9.7% savings (equat-ing to h5 million) and Company X experienced up to6.3% savings (equating to h3.5 million) using animproved periodic information sharing technique. Itshould be noted that these potential savings wereachieved on the system operating with a CT of 1.05.When additional capacity was added to the model(using a CT of 1.33), savings of h1.3 million andh1.6 million for the distributors and company,

respectively, were achieved using the same policiesas above. This point highlights the fact that morecapacitated systems benefit to a larger extent fromimproved information sharing. However, it shouldalso be noted that in this case it was assumed thatthere was no cost associated with this extra capacity.In other words, the resources were already in placeand could be used at no additional cost, which ingeneral will not be the case. Although suchconclusions have been drawn in the past they haveonly been drawn on more idealised supply chainstructures. This study provides results for a real casestudy thus providing encouragement and potentialincentives for companies to pursue such strategies inthe real world.

This study has taken an industrial example andhas quantified the impact of collaboration betweenpartners in relation to information sharing and theuse of intelligent forecasting using this improvedinformation. It can be seen from this study that allparties have benefited with the distributors comingout slightly better when information is sharedinstantaneously, thus encouraging distributors toshare such information. This is primarily due to thetimely receiving of orders and hence the reduction inbackorders due to the production being triggered byactual customer demand.

8. Future model extensions

Although this work provides useful insights intothe quantifiable benefits of improving informationsharing and forecasting, much more work is neededto study similar problems in all aspects of differingsupply chains as has been highlighted in theprevious section. The first and most obvious one isto generate similar models for multiple supply chainstructures as currently present in the industrialworld. In this study, the CT was set to the samevalue across all production elements in the supplychain depending on the experiment being studied.Further extensions to this work could be carried outto review the overall effect of differing capacityconstraints throughout the supply chain as pre-liminary studies into this have shown that capacitiesfurther down the supply chain (closer to thecustomers) have a bigger impact on the supplychain performance in relation to cost and servicelevels than capacities higher up the chain. Inaddition, further research should be carried out toinvestigate the varying cost of capacity in thesesystems, whether it is overtime costs or additional

ARTICLE IN PRESS

A-S

0

5000

10000

15000

20000

25000

30000

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

A-R

0

500

1000

1500

2000

2500

3000

3500

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

B-S

0

200

400

600

800

1000

1200

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

B-R

0

1000

2000

3000

4000

5000

6000

7000

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

C-S

0

500

1000

1500

2000

2500

3000

3500

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

C-R

0

200

400

600

800

1000

1200

1400

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

D-S

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

0 25 50 75 100 125 150 175 200 225 250

Dem

and

D-R

0

50

100

150

200

250

300

350

400

0 25 50 75 100 125 150 175 200 225 250

Dem

and

Fig. 6. Company X product demand (products A–D).

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 435

resource costs. Further extensions to this papershould entail a review on the effect of differingdemand patterns on the overall supply chainperformances. This is in contrast to the fixeddemand patterns used in this study. Similarly, as

this study is trying to review industrial supplychains, much more work is needed to research theuse of alternate cost structures, reorder parametersand methods and the use of additional forecastingtechniques.

ARTICLE IN PRESS

E-S

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

E-R

0

100

200

300

400

500

600

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

F-S

0

500

1000

1500

2000

2500

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

F-R

0

50

100

150

200

250

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

G-S

0

5000

10000

15000

20000

25000

0 25 50 75 100 125 150 175 200 225 250

Period

Dem

and

G-R

0

200

400

600

800

1000

1200

1400

0 25 50 75 100 125 150 175 200 225 250

Period

Period Period

Dem

and

H-S

0

50

100

150

200

250

300

350

400

0 25 50 75 100 125 150 175 200 225 250

Dem

and

H-R

0

500

1000

1500

2000

2500

0 25 50 75 100 125 150 175 200 225 250

Dem

and

Fig. 7. Company X product demand (products E–H).

P.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437436

ARTICLE IN PRESSP.J. Byrne, C. Heavey / Int. J. Production Economics 103 (2006) 420–437 437

Appendix

See Figs. 6 and 7.

References

Bhargava, H.K., Sridhar, S., Herrick, C., 1999. Beyond

spreadsheets: Tools for building decision support systems.

Computer 32 (3), 31.

Bourland, K.E., Powell, S.G., Pyke, D.F., 1996. Exploiting

timely demand information to reduce inventories. European

Journal of Operational Research 92, 239–253.

Cachon, G.P., Larviere, M.A., 2001. Contracting to assure

supply: How to share demand forecasts in a supply chain.

Management Science 47 (5), 629–646.

Chandra, C., Grabis, J., 2005. Application of multi-steps

forecasting for restraining the bullwhip effect and improving

inventory performance under autoregressive demand. Eur-

opean Journal of Operational Research 166, 337–350.

Christopher, M., 1998. Logistics and Supply Chain Management,

Strategies for Reducing Cost and Improving Service. Finan-

cial Times Professional Limited, England.

Clark, A.J., Scarf, H., 1960. Optimal policies for a multi-echelon

inventory problem. Management Science 6, 475–490.

Gaonkar, R., Viswnadham, N., 2001. Collaboration and

information sharing in global contract manufacturing net-

works. IEEE/ASME Transactions on Mechtronics 6 (4),

366–376.

Gavirneni, S., 2002. Information flows in capacitated supply

chains with fixed ordering costs. Management Science 48 (5),

644–651.

Gavirneni, S., Tayur, S., 1998. Value of information sharing and

comparison with delayed differentiation. In: Tayur, S.,

Ganeshan, R., Magazine, M. (Eds.), Quantitative Models

for Supply Chain Management. Kluwer Academic Publishers,

Dordecht.

Gavirneni, S., Kapuscinski, R., Tayur, S., 1999. Value of

information in capacitated supply chains. Management

Science 45 (1), 16–24.

Holweg, M., Disney, S., Holmstrom, J., Smoros, J., 2005. Supply

chain collaboration: Making sense of the strategy consortium.

European Management Journal 23 (2), 170–181.

Huang, G.Q., Lau, J.S.K., Mak, K.L., 2003. The impacts of

sharing production information on supply chain dynamics: A

review of the literature. International Journal of Production

Research 41 (7), 1483–1517.

Kelle, P., Akbulut, A., 2005. The role of ERP tools in supply

chain information sharing, cooperation, and cost optimiza-

tion. International Journal of Production Ecconomics 93-94,

41–52.

Lee, H.L., Whang, S., 2000. Information sharing in a supply

chain. International Journal of Technology Management 20

(3/4).

Lee, H.L., So, K.C., Tang, C.S., 2000. The value of information

sharing in a two-level supply chain. Management Science 46

(5), 626–643.

Liberopoulous, G., Koukoumialos, S., 2005. Tradeoffs between

base stock levels, numbers of kanbans, and planned supply

lead times in production/inventory systems with advance

demand information. International Journal of Production

Ecconomics 96, 213–232.

Monks, J.G., 1987. Operations Management Theory and

Practice. McGraw-Hill, New York, London.

Raghunathan, S., 2001. Information sharing in a supply chain: A

note on its value when demand is nonstationary. Management

Science 47 (4), 605–610.

Raman, A., DeHoratius, N., Ton, Z., 2001. The Achilles’ heel of

supply chain management. Harvard Business Review 79 (5),

25.

Schary, P.B., Skjøtt-Larsen, T., 1995. Managing the global

supply chain. Copenhagen, Handelshøjskolens forlag; Dis-

tribution, Munksgaard International.

Strader, T.J., Lin, F.R., Shaw, M.J., 1999. Business-to-business

electronic commerce and convergent assembly supply chain

management. Journal of Information Technology 14 (4),

361–373.

Yao, D.D., Yue, X., Wang, X., Liu, J.J., 2005. The impact of

information sharing on a returns policy with the addition of a

direct channel. International Journal of Production Ecco-

nomics 97, 169–209.

Zhao, X.D., Xie, J.X., 2002. Forecasting errors and the value of

information sharing in a supply chain. International Journal

of Production Research 40 (2), 311–335.

Zhao, X.D., Xie, J.X., Lau, R.S.M., 2001. Improving the supply

chain performance: Use of forecasting models versus early

order commitments. International Journal of Production

Research 39 (17), 3923–3939.

Zhao, X.D., Xie, J.X., Leung, J., 2002a. The impact of

forecasting model selection on the value of information

sharing in a supply chain. European Journal of Operational

Research 142 (2), 321–344.

Zhao, X.D., Xie, J.X., Zhang, W.J., 2002b. The impact of

information sharing and co-ordination on supply chain

performance. Supply Chain Management 7 (1), 22–40.