An Employee Performance Estimation Model for the Logistics Industry
14
An employee performance estimation model for the logistics industry Yu-Jen Wu 1 , Jiang-Liang Hou ⁎ Department of Industrial Engineering and Engineering Management National Tsing Hua University Hsinchu (300), Taiwan abstract article info Article history: Received 20 March 2009 Received in revised form 18 October 2009 Accepted 6 November 2009 Available online 18 November 2009 Keywords: Performance evaluation Logistics management Logistic information system Human resources management (HRM) In the last decade, the growing economy in Taiwan has brought about rapid growth in the logistics demands of enterprises. An important goal in the field of third party logistics (3PLs) is to improve the performance of logistics activities to enhance operation efficiency and enterprise competency. However, the employee performance must be determined in order to improve the activity performance of 3PLs. Thus, the aim of this research is to develop an employee performance estimation (EPE) model that includes three modules: direct performance determination (DPD), indirect performance determination (IPD), and performance score analysis (PSA). Moreover, a web-based logistics information management (LIM) platform was established via the EPE model in order to assist the managers in collecting and maintaining shop-floor operation data and to identify low-performance logistics tasks as well as inexperienced employees. In addition, a real-world case was used to demonstrate applicability of the proposed model and platform. As a whole, this paper presents an integrated model with the aims to more accurately calculate employee performance and significantly reduce the workload of 3PL decision makers. © 2009 Elsevier B.V. All rights reserved. 1. Introduction As the economy continues to grow in Taiwan, enterprises require more cooperation with professional logistics service providers in order to accomplish logistics activities since the complexity of logistics activities (e.g., distribution or warehousing) has gradually increased. This has resulted in a drastic increase in the number of third party logistics (3PLs) established for the purpose of fulfilling the logistics demands of enterprises. In order to enhance operation competency and efficiency, some 3PLs have utilized a variety of automated techniques and management strategies to improve the performance of logistical tasks. Although conventional 3PLs invest a large amount of money and time in their logistic operations, operation competency and efficiency has not shown significant improvement because managers cannot systematically recognize either low-performance logistics tasks or inexperienced employees. Logistics managers do not take a systematic approach for determining the performance of operators. In addition, logistics-related data (e.g., operation time) from the shop floor cannot be accurately gathered and imported into a logistics database and thus, they cannot be employed for operator performance evaluation. Under such circumstances, 3PL managers have difficulties reusing and analyzing logistics-related data. To overcome these problems, this research proposes a model aimed at determining the performance of different types of employees by utilizing the shop floor data of logistics activities. With regard to employee performance calculation, this research uses quantitative factors to estimate the operational performance of direct workers and indirect managers. Two performance reasoning modules are devel- oped in this study: • Direct Performance Determination: Used to determine the Real Performance (RP), Effective Performance (EP) and Derived Perfor- mance (DP) of direct workers. • Indirect Performance Determination: Used to determine the Verification Performance (VP), Assessment Performance (AP) and Inference Performance (IP) of indirect managers. The two modules can be combined to generate an integrated employee performance estimation model. In the proposed perfor- mance estimation model, the RP may first be calculated via the duration and quantitative outputs of logistical tasks. Subsequently, several quality indices (e.g., the operator trend index and operator idle index) can be formulated to determine the EP and DP. The team- level trend index, team-level quality index, schedule index and budget index can be formulated to estimate the VP, AP and IP. The operator and manager performance indices (i.e., RP, EP, DP, VP, AP and IP) can be given to logistics managers in order to identify both low- performance logistics tasks and inexperienced employees. In sum- mary, the proposed performance estimation approach can be used in the logistics management systems of 3PLs to produce an automatic determination of employee performance in a logistics center. Decision Support Systems 48 (2010) 568–581 ⁎ Corresponding author. Tel.: +886 3 5742658; fax: +886 3 5722685. E-mail addresses: [email protected] (Y.-J. Wu), [email protected] (J.-L. Hou). 1 Tel.: +886 3 5715131x33981; fax: +886 3 5722685. 0167-9236/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.dss.2009.11.007 Contents lists available at ScienceDirect Decision Support Systems journal homepage: www.elsevier.com/locate/dss
Transcript of An Employee Performance Estimation Model for the Logistics Industry
Decision Support Systems 48 (2010) 568–581
Contents lists available at ScienceDirect
Decision Support Systems
j ourna l homepage: www.e lsev ie r.com/ locate /dss
An employee performance estimation model for the logistics industry
Yu-Jen Wu 1, Jiang-Liang Hou ⁎Department of Industrial Engineering and Engineering Management National Tsing Hua University Hsinchu (300), Taiwan
⁎ Corresponding author. Tel.: +886 3 5742658; fax: +E-mail addresses: [email protected] (Y.-J. W
(J.-L. Hou).1 Tel.: +886 3 5715131x33981; fax: +886 3 5722685
0167-9236/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.dss.2009.11.007
a b s t r a c t
a r t i c l e i n f oArticle history:Received 20 March 2009Received in revised form 18 October 2009Accepted 6 November 2009Available online 18 November 2009
Keywords:Performance evaluationLogistics managementLogistic information systemHuman resources management (HRM)
In the last decade, the growing economy in Taiwan has brought about rapid growth in the logistics demandsof enterprises. An important goal in the field of third party logistics (3PLs) is to improve the performance oflogistics activities to enhance operation efficiency and enterprise competency. However, the employeeperformance must be determined in order to improve the activity performance of 3PLs. Thus, the aim of thisresearch is to develop an employee performance estimation (EPE) model that includes three modules: directperformance determination (DPD), indirect performance determination (IPD), and performance scoreanalysis (PSA). Moreover, a web-based logistics information management (LIM) platform was establishedvia the EPE model in order to assist the managers in collecting and maintaining shop-floor operation dataand to identify low-performance logistics tasks as well as inexperienced employees. In addition, a real-worldcase was used to demonstrate applicability of the proposed model and platform. As a whole, this paperpresents an integrated model with the aims to more accurately calculate employee performance andsignificantly reduce the workload of 3PL decision makers.
886 3 5722685.u), [email protected]
.
ll rights reserved.
© 2009 Elsevier B.V. All rights reserved.
1. Introduction
As the economy continues to grow in Taiwan, enterprises requiremore cooperation with professional logistics service providers inorder to accomplish logistics activities since the complexity oflogistics activities (e.g., distribution or warehousing) has graduallyincreased. This has resulted in a drastic increase in the number of thirdparty logistics (3PLs) established for the purpose of fulfilling thelogistics demands of enterprises. In order to enhance operationcompetency and efficiency, some 3PLs have utilized a variety ofautomated techniques and management strategies to improve theperformance of logistical tasks.
Although conventional 3PLs invest a large amount of money andtime in their logistic operations, operation competency and efficiencyhas not shown significant improvement because managers cannotsystematically recognize either low-performance logistics tasks orinexperienced employees. Logisticsmanagers do not take a systematicapproach for determining the performance of operators. In addition,logistics-related data (e.g., operation time) from the shop floor cannotbe accurately gathered and imported into a logistics database andthus, they cannot be employed for operator performance evaluation.Under such circumstances, 3PL managers have difficulties reusing andanalyzing logistics-related data.
To overcome these problems, this research proposes a modelaimed at determining the performance of different types of employeesby utilizing the shop floor data of logistics activities. With regard toemployee performance calculation, this research uses quantitativefactors to estimate the operational performance of direct workers andindirect managers. Two performance reasoning modules are devel-oped in this study:
• Direct Performance Determination: Used to determine the RealPerformance (RP), Effective Performance (EP) and Derived Perfor-mance (DP) of direct workers.
• Indirect Performance Determination: Used to determine theVerification Performance (VP), Assessment Performance (AP) andInference Performance (IP) of indirect managers.
The two modules can be combined to generate an integratedemployee performance estimation model. In the proposed perfor-mance estimation model, the RP may first be calculated via theduration and quantitative outputs of logistical tasks. Subsequently,several quality indices (e.g., the operator trend index and operatoridle index) can be formulated to determine the EP and DP. The team-level trend index, team-level quality index, schedule index andbudget index can be formulated to estimate the VP, AP and IP. Theoperator andmanager performance indices (i.e., RP, EP, DP, VP, AP andIP) can be given to logistics managers in order to identify both low-performance logistics tasks and inexperienced employees. In sum-mary, the proposed performance estimation approach can be used inthe logistics management systems of 3PLs to produce an automaticdetermination of employee performance in a logistics center.
569Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
By estimating the performances of all levels of employee, low-performance logistics tasks as well as inexperienced employees can bedetermined so that the demands of 3PLs for improvement inoperational competency and efficiency can be fulfilled.
2. Literature review
In the field of employee performance evaluation research, relatedstudies focus on evaluation schema construction and measurementitem calculation. The previous groundwork is discussed below.
2.1. Evaluation schema construction
In a performance pyramid model [13], decision makers shoulddetermine performance evaluation factors based on organizationlevels. It is not only education levels and work experience that affectemployee performance, but also job characteristics and workplaceconditions [9]. In the past, many studies utilized literature reviews,expert interviews or questionnaire surveys to identify the appropriateperformance evaluation factors of distinct industries. Sims et al. [15]employed interview and survey methodologies to determine howemployees in the medical and manufacturing industries address taskvariety, task autonomy, task identity, information feedback, dealingwith others, and friendship opportunity. In order to provide acomprehensive structure for performance evaluation, Coleman andBorman [4] first generated twenty-seven citizenship performancebehaviors, based on previous research, and developed a performanceevaluation structure composed of the interpersonal, organizationaljob/task dimensions via questionnaire. In addition, the TQM keycomponents (e.g., problem-solving abilities of employees) wereregarded as important factors for employee evaluation since enter-prise managers could understand the performance of implementingTQM by evaluating employee performance via TQM factors [5].
For room attendants and reception clerks in the hotel industry[3,12], nurse anesthetists in the hospital industry [17], and techniciansin the paper industry [18], expert interviewmethodology was used togenerate factors for candidate evaluation. The importance of thecandidate evaluation factors for managers is analyzed via question-naire. Performance evaluation schemas (i.e., factors and theircorresponding levels) and factor weights may be obtained by theanalytic hierarchy process (AHP) method. In order to ensure theapplicability of candidate performance evaluation factors, Chen [2]and Laio [11] used the Delphi and AHP methods to establishevaluation schemas for advertising executives in the newspaperindustry and technicians in the free-air television industry. A fuzzymultiple criteria algorithm may also be used to analyze theconsistency of performance evaluation factors. The MIJE (MetalIndustry Job Evaluation) system applied to evaluate employeeperformance in the metal industry should improve its evaluationfactor weights since the development of IT technologies and workingconditions have caused managers in the metal industry to stress newfactors. Hence, a revised MIJE system is proposed, using the expertinterview as well as AHP approaches for an optimal evaluationschema in line with the characteristics of the metal industry [8].
2.2. Measurement item calculation
For the scoring of employees using evaluation factors, the PDA(Performance Distribution Assessment) model proposed by Kane andKane [10] requests supervisors to first distribute a subjective score.The performance distributions for all employees can be establishedaccording to the frequencies occurring on different levels of theevaluation factors, while employee performance can be determined interms of specific statistics (e.g., the median or mode). Although thework behavior of R&D engineers in the software industry cannot beeasily measured, measurements of the key competencies for all R&D
engineers could be acquired by their managers via Q&A. Using thedifferences in the measurements and the optimal values of keycompetencies, R&D engineers may be classified into several groupsthrough the use of normal distribution. Furthermore, the performanceof R&D engineers can be rated on a basis of group rank [14].
In order to solve the problem of evaluating employee character-istics, Ahn and Chang [1] regarded the know-how and the humancapabilities of employees as product- and process-related tacitknowledge. In this study, tacit knowledge is transformed intoorganizational and financial performance by means of the DEA (dataenvelopment analysis) approach to investigate employee perfor-mance. It is not only regular work, but also job transfers and influenceactivities that affect employee performance. Eguchi [6] used the timeseries concept to estimate the financial profits that employeesgenerate from regular work. He applied the opportunity cost conceptto calculate the loss due to job transfers and influence activities. Hethen applied the averages of the fuzzy linguistic variables to estimateexpected employee performance using the probabilistic/possibilisticapproach. In order to provide lists to managers for the assignation ofemployees to jobs, employee performance must first be calculatedaccording to the estimated results. In order to generate candidateemployees, ranks of employees in distinct jobs may be determined onthe basis of their job attributes and employee performance [16].
Regarding the benefits generated by employee cooperation on ajob, employee combinations should be emphasized as employeeperformance is calculated for assigning employees to jobs [19]. First,the employee rank of distinct jobs can be determined using thestandard fuzzy arithmetic and then feasible employee combinationscan be generated via the triangular fuzzy number. The optimalemployee assignment plan for designated jobs can be determinedaccording to the job characteristics and may be provided to relevantsupervisors for operational planning. The model proposed by Golecand Kahya [7] quantifies the performance evaluation factors using theheuristic method and calculates the scores of employees using thefactors dictated by the fuzzy rules. The employee assignment programcan be determined by ranking employee scores.
As shown in the above literature review, previous studies forevaluating direct employee performance stressed the analysis of theperformance evaluation factors and calculations of the factor weights.However, it is critical to transform the measurement items of directemployees in the evaluation factors into direct employee perfor-mance. The derived performance can be used to measure the behaviorof the direct employee and the employee assignment plan. Fewstudies have been dedicated to the evaluation factors and the factorweights for indirect employees (e.g., managers). In contrast toprevious studies, this paper focuses on employee performanceevaluation within the logistics industry. Performance evaluationfactors were established according to the characteristics and organi-zational structure of a distribution center (DC). By utilizing the shopfloor operation data, a systematic and quantitative algorithmwas alsodeveloped to automatically calculate the performance of the directworkers and indirect managers of a logistics center.
3. Employee performance estimation model
In order to assist 3PL managers in estimating employee perfor-mance, this research develops an employee performance estimation(EPE) model to determine the performance of direct workers andindirect managers. To enable a determination of personal perfor-mance for each employee in a DC, staff levels and organization unitsmust be defined to serve as input. After defining the DC organization,this study utilized the operation data of logistics activities, the recordsfrom exception reports, work schedules, and budget plans to derivethe performance of each employee (including direct workers andindirect managers). The proposed EPE model can be categorized intothree modules (Fig. 1).
Fig. 1. Process for employee performance determination.
570 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
• Module I — Direct Performance Determination (DPD)
Based on the operation volume and work time of direct workers,the RP may be calculated. To address the error frequency of thelogistics operations, EP and DP can be derived from the idle time oflogistics activities and the growth trend of employee performance foreach direct worker. This module focuses mainly on the performanceestimation of the direct worker.
• Module II — Indirect Performance Determination (IPD)
The DP of direct workers, error frequency, growth trend ofemployee performance and the progress management of a team areconsidered for estimating the VP. The AP can be determined bycombining the VP and checking the budget consumption of thecorresponding office. Then, the IP can be integrated with the derivedAP. In short, this module can be utilized to estimate the performanceof indirect managers.
• Module III — Performance Score Analysis (PSA)
In the PSAmodule, the performance improvement scale of logisticstasks is denoted as an improvement rate. After deriving theimprovement rate, the DP, VP, AP and IP can be transformed intoperformance scores (PSs).
The operations of the above three modules can be further dividedinto seven stages: RP calculation, EP computation, DP determination,
VP estimation, AP estimation, IP estimation and PS analysis. Thedetails of a DC organization and the three performance estimationmodules are described below.
3.1. Organizational structure of a DC
The organizational structure of a typical DC must first be definedbefore applying the employee performance estimation approach. Thedefined organizational architecture can be regarded as the inputs forthe EPE model.
In this section, the relationship WRi,j,k between employees andoperational divisions can be identified. WRi,j,k can be used to denote adirect worker (where i=1 or 2) or an indirect manager (where i=2,3 or 4). That is, in a DC, the direct worker set S(SE)(i.e., {WRi,j,k|i=1,2}) comprises the first-line operators and team-level managerswhile the indirect manager set S(ML) (i.e., {WRi,j,k|i=2,3,4}) includes:team-level managers, office-level managers and division-level man-agers. The organizational structure of a typical DC is illustrated inFig. 2.
3.2. DPD module
The procedures to determine the performance of first-lineoperators and team-level managers consists of three stages (i.e., RP
Fig. 2. The organizational structure of a typical DC.
571Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
Calculation, EP Calculation and DP Calculation) and are discussed asfollows.
3.2.1. RP calculation for direct workersAt this stage, the logistics data must first be acquired and may be
used to calculate the RP of direct workers (including the first-lineoperators and team-level managers). The procedure for calculatingthe RP of direct workers in logistics tasks is discussed below.
3.2.1.1. Step (A1): Determine the time interval for RP calculation. A timeinterval on the basis of the logistics operation days should be assignedbefore calculating the real performance of direct employees. The timeinterval T can be derived based on the predefined starting day (t1) andending day (t2) for employee performance calculation.
3.2.1.2. Step (A2): Acquire logistics-related data for RP calculation. Afterdetermining the time interval T for employee performance calcula-tion, the basic predefined data regarding the direct workers WRi,j,k(i=1,2) and the logistics tasks LOl (l=1, ..., p) should be acquiredfrom the logistics information management (LIM) system. In additionto the basic predefined data, the logistics operation data regarding thework date CTt, the operation volume N(WRi,j,k, CTt, LOl) and thecorresponding work time T(WRi,j,k, CTt, LOl) at interval T can also beacquired from the LIM system. In general, the logistics-related datathat should be acquired prior to the employee performancecalculation is summarized in Table 1.
3.2.1.3. Step (A3): Calculate RP of logistics tasks for direct employees. Fora direct employee WRi,j,k, the real performance RC(WRi,j,k, T, LOl) foreach logistics task LOl at T indicates the average operation volume forthe direct employee within a unit time. Based on the above concept,RC(WRi,j,k, T, LOl) can be obtained via one of the following equationsand is summarized in Table 2.
Table 1Summary of the logistics-related data required for employee performance calculation.
LO1 … LOl … LOp
CT1 N(WRi,j,k, LO1, CT1) … N(WRi,j,k, LOl, CT1) … N(WRi,j,k, LOp, CT1)T(WRi,j,k, LO1, CT1) T(WRi,j,k, LOl, CT1) T(WRi,j,k, LOP, CT1)
… … … … … …
CTm N(WRi,j,k, LO1, CTm) … N(WRi,j,k, LOl, CTm) … N(WRi,j,k, LOp, CTm)T(WRi,j,k, LO1, CTm) T(WRi,j,k, LOl, CTm) N(WRi,j,k, LOp, CTm)
• On the basis of interval T:
RCðWRi;j;k; T; LOlÞ = ∑m
t=1NðWRi;j;k;CTt ; LOlÞ=∑
m
t=1TðWRi;j;k;CTt ; LOlÞ:
ð1Þ
• On the basis of the work date CTt:
RCðWRi;j;k; T; LOlÞ = ∑m
t=1
NðWRi;j;k;CTt ; LOlÞTðWRi;j;k;CTt ; LOlÞ=m: ð2Þ
3.2.2. EP calculation for direct workersOwing to the existence of defective items, the operation volume of
logistics tasks is not exactly that of the qualified volume of thelogistics tasks. Thus, the volume of defective items must be deductedfrom the operation volume of the logistics tasks to generate a qualityindex prior to determining the EPs of direct workers. The procedurefor calculating the EPs of direct workers is as follows.
3.2.2.1. Step (B1): Acquire operation error data of logistics activities. Theoperation error data of the logistics activities at T can be acquiredfrom the LIM system in order to calculate the operation error rates oflogistics activities. The operation error data of the logistics activitiesindicate the volume en(WRi,j,k, CTt, LOl) of errors induced by the directworkers WRi,j,k while performing the logistic task LOl in CTt.
3.2.2.2. Step (B2): Derive operation error rates of logistics activities. Fora direct worker, the quality index is defined as the operation errorrates er(WRi,j,k, T, LOl) of the logistics activities and can be determinedusing the ratio of the sum of en(WRi,j,k, CTt, LOl) to the sum of N(WRi,j,k,LOl, CTt) at T. That is, the quality index er(WRi,j,k, T, LOl) can be obtainedby means of the following equation:
erðWRi;j;k; T; LOlÞ = ∑m
t=1enðWRi;j;k;CTt ; LOlÞ=∑
m
t=1NðWRi;j;k;CTt ; LOlÞ
ð3Þ
Table 2Summary of RP for distinct direct employees and the logistics tasks.
LO1 … LOl … LOp
WR1,1,1 RC(WR1,1,1, T, LO1) … RC(WR1,1,1, T, LOl) … RC(WR1,1,1, T, LOp)… … … … … …
WR2,M,n2,MRC(WR2,M,N2,M
,T,LO1) … RC(WR2,M,N2,M,T,LOl) … RC(WR2,M,N2,M
,T,LOp)
572 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
3.2.2.3. Step (B3): Calculate EP of logistics activities for direct workers.The EP of a direct worker in a logistics activity indicates thethroughput of the logistics activity for the direct worker. Therefore,the EP of a direct worker for a specific logistics task can be obtained byconsidering the influence of the quality index on the correspondingRP. This idea is represented by the following equation:
ECðWRi;j;k; T; LOlÞ = RCðWRi;j;k; T; LOlÞ × ½1−erðWRi;j;k; T; LOlÞ�: ð4Þ
3.2.3. DP calculation for direct workersBased on the EP of direct workers derived at the previous stage, the
DPof directworkers can bedetermined by considering the trend indicesand working time indices of the logistics activities. The procedure forderiving the DPs of direct workers is discussed in the following.
3.2.3.1. Step (C1): Calculate trend indices of logistics activities. For adirect worker, the trend index of a specific logistic task can be used toindicate whether its corresponding EP significantly increases withtime. The procedure for calculating the trend index is discussed in thefollowing.
3.2.3.1.1. Step (C1-1): Acquire historical operation data for trendindex calculation. Following Step (A2), the historical operation data inthe working days HCTs, the historical operation volume N(WRi,j,k,HCTs, LOl) and the corresponding historical working time T(WRi,j,k,HCTs, LOl) at time interval HT (from t1 to t2) can also be acquired fromthe LIM system and used to structure the groundwork for comparingthe difference in EP between time intervals T and HT.
3.2.3.1.2. Step (C1-2): Calculate historical EP of logistics activities fordirect workers. In order to determine the variation of EP for a directworkerWRi,j,k at T using the historical EP at HT, the average HEC(WRi,j,k,HT, LOl) and variance HECV(WRi,j,k, HT, LOl) of the historical EP must bederived using the following equations:
HECðWRi;j;k;HT; LOlÞ = ∑n
s=1HECðWRi;j;k;HCTs; LOlÞ=n ð5Þ
HECVðWRi;j;k;HT; LOlÞ =∑n
s=1½HECðWRi;j;k;HCTs; LOlÞ−HECðWRi;j;k;HT; LOlÞ�2
n−1:
ð6Þ
3.2.3.1.3. Step (C1-3): Establish confidence interval of historical EPsfor direct workers. After deriving HEC(WRi,j,k, HT, LOl) and HECV(WRi,j,k,HT, LOl) from the logistics activities of direct workers at HT, theconfidence intervals of historical EPs for a direct worker WRi,j,k can beestablished using the following equations with a significant level α:
C1ðWRi;j;k;HT; LOlÞ = HECðWRi;j;k;HT; LOlÞ + Zα2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiHECVðWRi;j;k;HT; LOlÞ
n
s
ð7Þ
C2ðWRi;j;k;HT; LOlÞ = HECðWRi;j;k;HT ; LOlÞ−Zα2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiHECVðWRi;j;k;HT ; LOlÞ
n
s
ð8Þ
where C1(WRi,j,k, HT, LOl) and C2(WRi,j,k, HT, LOl) represents the upperand lower limits of the confidence intervals for the historical EPs,respectively.
3.2.3.1.4. Step (C1-4): Determine values of trend indices for logisticsactivities. For a direct worker, the trend indices of the logisticsactivities indicate the status (e.g., increase, slack or decrease) of EPsafter comparisonwith the historical EP. The values of the trend indicesTrI(WRi,j,k, T, LOl) at T can be assigned as “1”, “0” or “−1” according tothe following rule. The value of TrI(WRi,j,k, T, LOl) is equal to “1” if EPs
of the logistics activities at T are higher than the upper limits ofconfidence intervals of the historical EPs at HT. That is, the EP of thelogistics activity at T obviously increases as compared with thehistorical EP.
TrIðWRi;j;k; T; LOlÞ=f1 if ECðWRi;j;k; T; LOlÞ ≥ C1ðWRi;j;k; T; LOlÞ0 if C1ðWRi;j;k; T; LOlÞ N ECðWRi;j;k; T; LOlÞN C2ðWRi;j;k; T ; LOlÞ−1 if ECðWRi;j;k; T; LOlÞ ≤ C2ðWRi;j;k; T; LOlÞ
:
ð9Þ
3.2.3.2. Step (C2): Calculate working time index of logistics activities fordirect workers. The working time index of a direct worker in a specificlogistics task can be used to indicate the busyness of the workerrequired to perform the corresponding logistics task. The procedurefor calculating the working time index is discussed below.
3.2.3.2.1. Step (C2-1): Determine categories of working date withinthe predefined time interval. In order to accurately evaluate thebusyness of the direct worker in performing the logistics tasks usingthe idle time of the corresponding workers at time interval T, thecategories of the working date at T should first be determined in orderto identify on-job and off-job days. That is, the working date CTt ofthe direct workers at T can be determined by distinguishing whetherCTt belongs to the set of working days NWDSet(WRi,j,k, T). The setNWDSet(WRi,j,k, T) is usually defined by DC managers. Therefore, thenumber of elements in NWDSet(WRi,j,k, T) is equal to the number ofworking days suggested by the DCmanagers for the direct workers at T.
3.2.3.2.2. Step (C2-2): Calculate total working time of thedirect workers within the predefined time interval. After acquiringNWDSet(WRi,j,k, T), the total working time TNWT(WRi,j,k, T), which theDCmanagers suggest for thedirectworkers to perform the logistics tasksat T, can be obtained using the following equation, in order to calculatethe idle time of direct workers while performing the logistics tasks.
TNWTðWRi;j;k; TÞ = DNWT × NNWDðWRi;j;k; TÞ ð10Þ
where DNWT is the daily working time of the direct workers proposedby the DC managers.
3.2.3.2.3. Step (C2-3): Calculate total idle time of the direct workerswithin the predefined time interval. To utilize theworking time index toevaluate the busyness of direct workers, the idle conditions of the directworkers at T should be considered. Using TNWT(WRi,j,k, T) added to thesum of the operation time in which the direct workers perform thelogistics tasks on theworking days at T, the total idle time TIDT(WRi,j,k, T)of a direct worker can be determined via the following equation.
TIDTðWRi;j;k; TÞ = TNWTðWRi;j;k; TÞ
− ∑p
l=1∑n
t=1TðWRi;j;k;CTt ; LOlÞ jCTt∈NWDSetðWRi;j;k; TÞ
n o:
ð11Þ
3.2.3.2.4. Step (C2-4): Calculate idle time of each logistics activity fordirect workers. The idle time IDT(WRi,j,k, T, LOl) of a direct worker in aspecific logistics task at T can be determined by distributing the totalidle time in relation to the proportion of the corresponding logisticstask. The proportion compares the ratio of the time in which the directworker performs this logistics task to the time of the same directworker perform all logistics tasks on the working days at T. Based onthe above idea, the idle time IDT(WRi,j,k, T, LOl) of logistics tasks for adirect worker can be obtained via the following equation.
IDTðWRi;j;k; T ; LOlÞ = TIDTðWRi;j;k; TÞ
×∑m
t=1fTðWRi;j;k;CTt ; LOlÞ jCTt∈NWDSetðWRi;j;k; TÞg
∑p
l=1∑m
t=1fTðWRi;j;k;CTt ; LOlÞ jCTt∈NWDSetðWRi;j;k; TÞg
:
ð12Þ
573Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
3.2.3.2.5. Step (C2-5): Determine values of working time indices forlogistics activities. The working time index of a specific logistics taskindicates the busyness of a direct worker in the correspondinglogistics task. The quantitative values of the working time indices forthe logistics tasks are the proportions of “the difference between thetotal working time and the idle time” to “the total working time”.Therefore, the values of the working time indices for direct workerscan be formulated using the following the equation.
WtIðWRi;j;k; T; LOlÞ
= 1−IDTðWRi;j;k; T; LOlÞ
∑p
l=1∑m
t=1fTðWRi;j;k;CT ; LOlÞ jCTt∈NWDSetðWRi;j;k; TÞg
ð13Þ
3.2.3.3. Step (C3): Calculate DPs of logistics activities for direct workers.Based on EC(WRi,j,k, T, LOl), TrI(WRi,j,k, T, LOl) and WtI(WRi,j,k, T, OLl)of a direct worker in a specific logistics task, the DP of the directworker in the specific logistics task at T can be calculated using thefollowing equation.
FCðWRi;j;k; T; LOlÞ = ECðWRi;j;k; T; LOlÞ × ½1 + TrW × TrIðWRi;j;k; T; LOlÞ�× WtIðWRi;j;k; T; LOlÞ
ð14Þ
where the weight TrW of the trend index for the direct worker in thespecific logistics task at T represents the level of which the EP isgreater or smaller than the historical EP as the confidence interval ofhistorical EPs is utilized. The weight can be determined via thefollowing equation. For example, TrW is the ratio of the differencebetween EC(WRi,j,k, T, OLl) and EC(WRi,j,k, T, OLl) the differencebetween C1(WRi,j,k, HT, OLl) and HEC(WRi,j,k, HT, OLl) if TrI is equal to“1”.
TrW =
ECðWRi;j;k; T ;OLlÞ−C1ðWRi;j;k;HT;OLlÞC1ðWRi;j;k;HT;OLlÞ−HECðWRi;j;k;HT;OLlÞ
; If TrIðWRi;j;k; T ;OLlÞ = 1
0 ; If TrIðWRi;j;k; T ;OLlÞ = 0
C2ðWRi;j;k; T ;OLlÞ−ECðWRi;j;k; T ;OLlÞHECðWRi;j;k;HT ;OLlÞ−C2ðWRi;j;k;HT;OLlÞ
; If TrIðWRi;j;k; T ;OLlÞ = −1
:
8>>>>>>><>>>>>>>:
ð15Þ
3.3. IPD module
In the DPD module, the performance (i.e., RP, EP and DP) of directworkers in distinct logistics tasks at T can be determined by utilizingthe operational data of the logistics tasks. In addition to determiningthe performance of direct workers, the managerial benefits obtainedby indirect managers must be considered in order to evaluate theperformance of indirect managers. The IPD module that determinesthe performance of team-level managers, office-level managers anddivisionmanagers can be classified into four stages: DP Estimation, VPEstimation, AP Estimation and IP Estimation.
3.3.1. VP calculation for team-level managersIn addition to direct participation in the logistics tasks, the team-
level managers must also be responsible for the performance of first-line operators, the performing quality of logistics activities and theachievement of the scheduled operation plans of the logistics tasks.Considering the duties of the team-level managers, the VPs of team-level managers can be determined by utilizing the average perfor-mance index, the average quality index and the schedule index. Thefollowing section describes the calculation procedure for the VPs ofteam-level managers.
3.3.1.1. Step (D1): Calculate integrated DPs of team-level managers. Fora team-level managerWRi=2,j,k, the integrated DP can be determinedby a summation of the derived performance FC(WRi=2,j,k, T)multiplied with the corresponding weighting values βl. The aboveidea can be expressed using the following equation.
IFCðWRi=2;j;k; TÞ = ∑p
l=1βl × FCðWRi=2;j;k; T; LOlÞ ð16Þ
where the weighting value βl denotes the difficulty of direct workersin performing logistics task LOl. That is, βl is the ratio of the efficiencyof LOl to the efficiency of all logistics tasks and can be expressed usingthe following equation:
βl =NðWRi=1;j;k; T;OLlÞTðWRi=1;j;k; T;OLlÞ=∑
p
l=1
NðWRi=1;j;k; T ;OLlÞTðWRi=1;j;k; T;OLlÞ
: ð17Þ
3.3.1.2. Step (D2): Calculate average performance indices of team-levelmanagers. The average performance index of a team-level manager atT is the average of the multi-DPs of all direct workers managed by thecorresponding team-level manager. The procedure for calculating theaverage performance index of a team-level manager is discussedbelow.
3.3.1.2.1. Step (D2-1): Merge DPs of first-line operators in logisticsactivities. Following Step (D1), the blended-DP MFC(WRi=1,j,k, T, LO)of a first-line operator assigned to the logistics tasks at T can beobtained via the following equation:
MFCðWRi=1;j;k; T; LO˙Þ = ∑p
l=1βl × FCðWRi=1;j;k; T ; LOlÞ: ð18Þ
3.3.1.2.2. Step (D2-2): Calculate average of blended-DPs of all first-line operators for team-level managers. The average performance indexof a team-level manager indicates how well the team-level managermanages the first-line operators performing the logistics tasks. Theaverage performance index can be determined using the sum ofblended-DPs for all first-line operators divided by the number of first-line operators in the corresponding team. The above concept can bedescribed using the following equation:
ADCIðWRi=2;j;k; TÞ = ∑n1;j
k=1MFCðWRi=1;j;k; T; LO˙Þ=n1;j: ð19Þ
3.3.1.3. Step (D3): Derive average quality indices for team-levelmanagers. For a team-level manager, the average quality index isdefined as a blended error rate of the direct workers assigned tologistics tasks managed by the team-level manager at T. Theprocedure for deriving the average quality index of a team-levelmanager is discussed below.
3.3.1.3.1. Step (D3-1): Blend error rates of first-line operators inlogistics activities. Following Step (D1-1), a blended error rate mer(WRi=1,j,k, T, LO) of the first-line operatorWRi=1,j,k can be determinedusing the summation of error rates er(WRi=1,j,k, T, LOl) in the logisticstasks LOl in a team multiplied with the corresponding weightingvalues βl. Here, the weighting value βl denotes the average level ofdifficulty in performing the corresponding logistics task for the first-line operators. The above idea can be expressed using the followingequation:
merðWRi=1;j;k; T; LO˙Þ = ∑p
l=1βl × erðWRi=1;j;k; T; LOlÞ: ð20Þ
3.3.1.3.2. Step (D3-2): Calculate average of blended error rates of allfirst-line operators for team-level managers. The average quality index
574 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
of a team-level manager shows the working quality with which hemanages the corresponding first-line operators performing thelogistics tasks for his team. For a team-level manager, the averagequality index ADQI(WRi=2,j,k, T) can be determined using the averageof the sum of mer(WRi=1,j,k, T, LO) for the first-line operatorsmanaged by the team-level manager. The equation for calculatingADQI(WRi=2,j,k, T) is as follows:
ADQIðWRi=2;j;k; TÞ = ∑n1;j
k=1merðWRi=1;j;k; T ; LO˙Þ= n1;j: ð21Þ
3.3.1.4. Step (D4): Calculate schedule indices for team-level managers.For a team-level manager, the schedule index shows his ability tocontrol the logistics action plans. That is, a schedule index can be usedto evaluate whether the team-level manager accurately implementsthe logistics action plans (e.g., arranging the logistics tasks and theircorresponding operation volumes). The procedure for calculating theschedule index of a team-level manager is discussed below.
3.3.1.4.1. Step (D4-1): Calculate completion levels of logisticsactivities. Before calculating the schedule index of a team-levelmanager, the assigned operation volumes PN(CTt, LOl) of logisticstasks in CTt must be first acquired. Subsequently, the completion levelof a logistics task within CTt can be determined using the ratio ofoperation volume sum to the assigned operation volumes of thelogistics tasks. The completion level CD(CTt, LOl) of a logistics task isrepresented by the following equation:
CDðCTt ; LOlÞ = ∑ni;j
k=1NðWRi;j;k;CTt ; LOlÞ = PNðCTt ; LOlÞ: ð22Þ
After determining CD(CTt, LOl) within CTt, a completion level CD(T,LOl) of the logistics task within T can also be determined using thefollowing equation:
CDðT; LOlÞ = ∑m
t=1CDðCTt ; LOlÞ = m: ð23Þ
3.3.1.4.2. Step (D4-2): Calculate schedule indices of team-levelmanagers. The schedule index of a team-level manager representsthe percentage of completed logistics tasks. Thus, the completion levelCD(T, LOl) of a logistics task within T must be combined to determinethe schedule indices of team-level managers. The schedule index DSCI(WRi=2,j,k, T) of a team-level manager can be obtained using thefollowing equation:
DSCIðWRi=2;j;k; TÞ = ∑p
l=1βl × CDðT; LOlÞ ð24Þ
3.3.1.5. Step (D5): Calculate VP for team-level managers. Beforecalculating VP, the integrated DP and the average performance indicesof each team-level manager should be averaged because a team-levelmanager might participate in the logistics tasks with the first-lineoperators and manage the first-line operators while performing thelogistics activities at the same time. Thus, integrated DPs IFC(WRi=2,j,k,T) and the average performance indices ADCI(WRi=2,j,k, T) of theteam-level managers should first be averaged for VP derivation.Afterwards, considering the importance of the average quality indexADQI(WRi=2,j,k, T) and the schedule index DSCI(WRi=2,j,k, T) forteam-level managers, the VP VC(WRi=2,j,k, T) of a team-level managerWRi=2,j,k can be derived based on the following equation:
VCðWRi=2;j;k; TÞ =IFCðWRi=2;j;k; TÞ + ADCIðWRi=2;j;k; TÞ
2
� �× ½1−ADQIðWRi=2;j;k; TÞ� × DSCIðWRi=2;j;k; TÞ:
ð25Þ
3.3.2. AP calculation for division-level managersAt this stage, the APs of division-level managers can be determined
on the basis of their budget control performance. In the following, thecalculation procedure for the budget index and the AP for a division-level manager are discussed.
3.3.2.1. Step (E1): Calculate the blended VP of division managers.Following Step (D3-2), the blended VP MVC(WRi=3,j,k, T) of thedivision-level manager WRi=3,j,k can be obtained by averaging theVPs VC(WRi=2,j,k, T) for team-level managers WRi=2,j,k in the samedivision at T. The above concept can be described using the followingequation:
MVCðWRi=3;j;k; TÞ = ∑n2;j
k=1VCðWRi=2;j;k; TÞ=n2;j: ð26Þ
3.3.2.2. Step (E2): Calculate budget indices for division-level managers.Before calculating the budget indices, both the budgets andexpenditures of the jth division at T must first be acquired. For adivision-level manager, the budget index shows his/her ability tocontrol the budget. The budget index can be used to indicate whethera division-level manager can effectively reduce expenditure. Thedivision-level manager with a higher budget index is capable ofreducing division expenditure. Accordingly, the ratio of the differencebetween the division budgets PE(j,T) and the division expenditures RE(j,T) and the division budgets can be used to determine the ability of adivision-level manager to control a budget. Thus, the budget index FCI(WRi=3,j,k, T) of a division-level manager can be derived using thefollowing equation:
FCIðWRi=3;j;k; TÞ = 1 + ½PEðj; TÞ−REðj; TÞ= PEðj; TÞ�: ð27Þ
3.3.2.3. Step (E3): Calculate AP for division-level managers. Becausedivision-level managers control division expenditure, the budgetindices FCI(WRi=3,j,k, T) should be consideredwhile using the blendedVP MVC(WRi=3,j,k, T) to calculate the AP of division-level managersWRi=3,j,k at T. Accordingly, the AP IC(WRi=3,j,k, T) of a division-levelmanager at T can be calculated using the following equation:
ICðWRi=3;j;k; TÞ = MVCðWRi=3;j;k; TÞ × FCIðWRi=3;j;k; TÞ: ð28Þ
3.3.3. IP calculation for office-level managersOffice-level managers must propose the operation plans of a DC
and the division-level managers must bring these plans into action.Thus, the performance of division-level managers can be used toevaluate the IPs of office-level managers. The following step revealsthe IP calculation details for office-level managers.
3.3.3.1. Step (F): Calculate IP for office-level managers. Although office-level managers do not directly participate in logistics tasks, theyshould be responsible for leading division-level managers in themanagement of their corresponding divisions. Therefore, the IP foroffice-level managers can be estimated using the APs of division-levelmanagers. Based on the above idea, the IP PC(WRi=4,j,k, T) of an office-level manager at T can be calculated by averaging the APs IC(WRi=3,j,k,T) of division-levelmanagers. The above concept can be expressed usingthe following equation:
PCðWRi=4;j;k; TÞ = ∑ni=4;j
k=1ICðWRi=3;j;k; TÞ = ni=3;j: ð29Þ
3.3.4. Calculation of standard performance score for employeesIn addition to fundamental logistics management tasks (e.g.,
budget control), indirect managers should also be responsible forimproving logistics activities to enhance operational efficiency. In
575Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
order to evaluate the benefits generated by indirect managers, anyimprovement in performance brought about by indirect managers inlogistics activities should also be considered while transforming thedistinct categories of employee performance into the standardperformance scores for all employees. As a result, any bottleneck ofemployees or logistics tasks can be identified using standardperformance scores.
3.3.4.1. Step (G1): Calculate whole growth rate of EPs. The primary goalof an indirect manager is to improve the performance of logistics tasksfor the purpose of enhancing the DC operation competency andefficiency. In other words, the improvement of logistics tasks canenhance operation efficiency. Thus, the entire growth rate of EPsrepresents the benefits of improved logistics tasks in a DC. Thefollowing procedure reveals the process by which to obtain the entiregrowth rate of EPs.
First, the blended EPMEC(WRi=1,2,j,k, T, LO) of all direct workers atTmust be integrated by using the weighting values βi of logistics taskscombined with the EPs EC(WRi=1.2,j,k, T, LOl) of direct workers. Next,the entire EP AEC(WRi=1,2; ; ,T) of a DC can be obtained by averagingthe blended EP MEC(WRi=1,2,j,k, T, LO) of direct workers at T. Theabove idea can be expressed using the following equations:
MECðWRi=1;2;j;k; T; LO˙Þ = ∑p
l=1βl × ECðWRi=1:2;j;k; T; LOlÞ ð30Þ
AECðWRi=1;2;˙;˙; TÞ
= ∑2
i=1∑b
j=1∑
ni=1;2;j
k=1MECðWRi=1;2;j;k; T; LO˙Þ = 2 × b × ni=1;2;j:
ð31Þ
Following the above formula for calculating AEC(WRi=1,2; ; ,T), theentire historical EP AHEC(WRi=1,2; ; ,HT) of direct workers at HT canbe obtained using the following equation:
AHECðWRi=1;2;˙;˙;HTÞ
= ∑2
i=1∑b
j=1∑
n1=1;2;j
k=1∑p
l=1βl × ECðWRi;1;2;j;k;HT; LOlÞ
� �=2 × b × ni=1;2;j:
ð32ÞFinally, the entire growth rate AECGR(T) of EPs at T can be derived
using the ratio of the difference between “the whole EP for directmanagers at T” and “the whole historical EP for direct managers at HT”and “the whole historical EP for direct managers at HT”. The aboveidea can be represented using the following equation:
AECGRðTÞ = ½AECðWRi=1;2;˙;˙; TÞ−AHECðWRi=1;2;˙;˙; TÞ�= AHECðWRi=1;2;˙;˙; TÞ:ð33Þ
3.3.4.2. Step (G2): Calculate improvement weights. In a DC organization,higher level managers should be assigned a higher improvementweight when transforming the distinct employee performance indicesinto a standard performance score because they are responsible formanaging a larger number of staff. These managers must spend moretime and effort to enhance the EPs of direct operators by improvinglogistics tasks. In order to distinguish the difference betweenimprovement benefits generated by two levels of managers (e.g.,the team-level and the division-level), this model assumes that theimprovement weights for transforming the employee performanceindices into a standard performance score follows a geometric series.Based on the above assumption, the improvement weight IBN(i,T) at Tcan be determined using the whole growth rate AECGR(T) of EPs andthe organization level (i) corresponding to the employee. The aboveidea can be expressed using the following equation:
IBNði; TÞ = ½1 + AECGRðTÞ�i−1: ð34Þ
3.3.4.3. Step (G3): Calculate performance scores of all employees. Inorder to accurately estimate the contributions and analyzethe performance of all employees, employee performance (i.e., DPsFC(WRi=1,j,k, T), VPs VC(WRi=2,j,k, T), APs IC(WRi=3,j,k, T), IPsPC(WRi=4,j,k, T)) derived in previous steps should be transformedinto a standard performance score by utilizing the improvementweights IBN(i, T). Based on the above idea, the standard performancescore CS(WRi,j,k, T) corresponding to each type of employee, includingthe first-line operators, the team-level managers, division-levelmanagers and office-level managers, at time interval T can becalculated based on their role in a DC.
CSðWRi;j;k; TÞ = f FCðWRi;j;k; TÞ × IBNði; TÞVCðWRi;j;k; TÞ × IBNði; TÞICðWRi;j;k; TÞ × IBNði; TÞPCðWRi;j;k; TÞ × IBNði; TÞ
;
If i = 1If i = 2If i = 3If i = 4
: ð35Þ
By analyzing the distribution of the standard performance scoresfor all employees taking part in logistics tasks, the employees whogenerate more benefits to the DC can be identified. That is, theinexperienced employees and the low-performance logistics tasks canalso be determined according to standard performance scores.
4. Logistics information management platform
In order to demonstrate the applicability of the proposed model, aweb-based logistics information management (LIM) platform wasestablished in this research. Under the LIM platform, four mainmodules are provided: Fundamental Logistics Data Maintenance(FLDM), Business Information Management (BIM), Shop Floor StatusReport (SFSR) and Employee Performance Calculation (EPC). In thefollowing, the LIM platform integrated with the employee perfor-mance calculation algorithms is introduced.
Before calculating employee performance, the fundamental logis-tics data (e.g., the logistics activities, the employees and the DCorganization), must first be maintained used the FLDM module(Fig. 3). Using the SFSR and BIM modules, the logistics operation data(e.g., the working days, the operation volume and the defect volume)as well as the business information (e.g., the division budget or thescheduled operation volumes of logistics tasks) can be imported intothe LIM database by the administrators (e.g., team-level or division-level managers) to serve as the inputs for the employee performancecalculation (Fig. 4). Based on the above data, the LIM platform canautomatically generate employee performance (i.e., RP, EP, DP, VP, APand IP) for the decisionmakers of the DC using the EPCmodule (Fig. 5)as the user requests a employee performance calculation (e.g., theemployee name or time interval). The LIM platform also providestabulated and graphical interfaces to display the detailed data (Figs. 6and 7) as well as the statistics (Fig. 8) regarding employeeperformance for the DC decision makers in order to assist them toanalyze the performance of each employee.
In summary, the proposed model for employee performancecalculation and the application modules developed under the LIMplatform can be used in DCs to efficiently and accurately identify low-performance logistics tasks and inexperienced online employees inreal time. Based on the bottleneck analysis of employees and logisticstasks, DC decision makers can improve low-performance logistics andassist inexperienced employees through effective training.
5. Case study
In order to verify the applicability of the employee performanceestimation (EPE) model and logistics information management (LIM)platform, a real-world case, Nung Hsueh distribution center which isthe largest logistics center of the printing industry in Taiwan, wasused in this study. Three evaluation approaches, including the random
Fig. 3. Interface for FLDM.
Fig. 4. Interface for SFSR.
576 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
Fig. 5. Display of employee performance.
577Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
approach, expert (i.e., the DC managers) evaluation and the EPEapproach, were applied to evaluate the performance of employees(i.e., the first-line operators, team-level managers, office-levelmanagers and division-level managers) in Nung Hsueh distribution
Fig. 6. Tabulated logistics data for em
center. After that, the employees can be ranked based on theirperformance generated by the three evaluation approaches and theranks of employee performance determined via the three approachescan be acquired. The related details, including collection of evaluation
ployee performance calculation.
Fig. 7. Visualized display of employee performance.
578 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
data, performance evaluation and evaluation approach analysis, canbe discussed in the following.
1. Collection of evaluation data
The evaluation data of 36 distinct first-line operators (A1, A2, …,A36), 12 distinct team-level managers (B1, B2, …, B12), 4 office-levelmanagers (C1, C2, C3, C4) and 2 division-level managers was acquiredvia the operation records of the shop-floor and the business forms ofthe accounting department from the 1st to 30th September, 2007. Theevaluation data was composed of the following items:
• The operation volumes, operation time and error volumes ofemployees in the logistics tasks
• The scheduled operation volumes of logistics tasks in the team-levelunits
• The budgets and expenditures determined in the office-level units
Fig. 8. Logistics statistics for emplo
After acquiring the evaluation data, the evaluation data can to besummarized via the traditional statistics graphics for employeeperformance appraisal and be imported into the LIM platform toutilize the random approach, expert evaluation and the EPE approachto evaluate the employee performance.
2. Performance evaluation
The procedure to evaluate the employee performance and todeterminate the employee performance ranks via the threeapproaches are described as follows:
• Random approach
The random numbers were used to generate the evaluation score(between 1 and 100) of employees and the employees were rankedbased on their scores. Ten series of employee performance ranks canbe obtained by repeating the above evaluation processes. In addition,the average and standard deviation of employee performance ranks
yee performance calculation.
Table 4The ANOVA table of evaluation scores.
Evaluation approach Source SS DF MS F p-value
Random approach Employee 54962.93 53 1037.04 1.31 0.081Experiment 7290.56 9 810.06 1.02 0.423Error 379161.80 477 794.89Total 441415.30 539
Expert evaluation Employee 16990.01 53 320.56 3.68 b0.001Expert 10115.08 9 1123.89 12.90 b0.001Error 41537.32 477 87.08Total 68642.41 539
579Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
for each employee can be calculated and be regarded as the non-professional evaluation results.
• EPE approach
The EPE approach was used to estimate the employee perfor-mance, including the derived performance of first-line operations,verification performance of team-level managers, inference perfor-mance of office-level managers and assessment performance ofdivision-level managers. The employees were also ranked based onthe employee performance and the employee performance ranks canbe regarded as the systematical evaluation results.
• Expert evaluation
Ten experienced DC managers were selected and requested toassign the appropriate scores between 1 and 100 to the employeesbased on the traditional statistics graphics of the employee perfor-mance in order to denote the grade of employee performance. Theevaluation scores of employees in the 10 evaluation experiments canbe acquired and the employees were ranked based on their scores.Furthermore, the average and standard deviation of employeeperformance ranks for each employee can be calculated and regardedas the professional but non-systematical evaluation results.
In the random and expert evaluations, the employees can be re-ranked based on their average ranks. The average and standarddeviation of performance ranks and the final performance rank foreach employee can be summarized in Table 3.
3. Evaluation approach analysis
In order to evaluate the applicability of the three evaluationapproaches, the variance of the evaluation scores, reasonableness andsimilarity of employee performance ranks can be analyzed based onthe evaluation scores and employee performance ranks. The proce-dure to analyze the variance of the evaluation scores, reasonableness
Table 3Employee performance ranks of three evaluation approaches.
Employee Random EPE Expert
Avg. Std. Rk. Rk. Avg. Std. Rk.
A1 26.7 8.9 36 5 9.3 6.59 2A2 15.7 7.6 11 6 15.6 9.18 14A3 20.6 11.7 28 4 15.1 11.37 13A4 15.5 9.5 8 10 17.1 9.46 18A5 20.0 8.9 24 11 9.7 8.32 3A6 18.3 10.2 16 20 14.4 10.21 11A7 14.9 9.7 5 19 12.8 8.11 6A8 23.3 11.1 33 33 21.1 8.65 27A9 25.7 4.8 35 30 14.3 8.61 10A10 21.3 11.0 29 14 14.0 8.69 8A11 11.4 6.1 3 18 20.2 8.87 25A12 7.7 6.4 1 15 15.7 10.01 15A13 20.3 12.6 27 2 8.8 5.78 1A14 15.1 10.5 6 3 12.9 9.20 7A15 14.5 5.9 4 7 19.5 10.05 23A16 15.5 11.2 8 9 15.0 8.32 12A17 26.7 8.9 23 13 21.0 5.87 26A18 15.7 7.6 8 25 28.5 8.11 35A19 19.9 10.4 22 31 27.6 7.12 33A20 15.5 10.7 19 29 28.4 6.68 34A21 19.2 10.3 30 26 29.0 6.43 36A22 19.0 9.8 2 1 14.1 11.09 9A23 21.4 12.0 34 16 25.2 8.95 31A24 10.3 6.0 30 21 21.3 9.83 28A25 24.7 8.9 32 8 9.8 7.59 4A26 21.4 5.7 21 12 16.4 8.44 16A27 22.5 9.1 14 22 18.9 9.35 21
• Avg.: Average of employee performance ranks.• Std.: Standard deviation of employee performance ranks.• Rk.: Final employee performance rank.
and similarity of the employee performance ranks are described asfollows:
• Analysis of the variance of evaluation scores
In order to analyze the difference of appraisal results in eachevaluation experiment, the distinct evaluation experiments andemployees were regarded as factors and the two-way analysis ofvariation (i.e., ANOVA) was used to analyze the evaluation scoresdetermined via the random approach and expert evaluation (Table 4).
As shown in Table 4, the p-values of the null hypotheses (i.e., theevaluation scores are identical and the appraisal criteria applied toevaluate the employee performance are consistent in each evaluationexperiment) in the random approach are less than 0.01 and the twonull hypotheses of the random evaluation cannot be rejected.Therefore, the employee performance can be consistently evaluatedvia the identical appraisal criteria. However, employees with distinctperformance cannot be effectively identified. In addition, the two nullhypotheses of the expert evaluation approach should be rejected (i.e.,p-valueb0.01). Although the employees with distinct performancecan be effectively distinguished, the appraisal criteria of distinctexperts are significantly different. In other words, the evaluationresults are inconsistent for different experts. As a result, the
Employee Random EPE Expert
Avg. Std. Rk. Rk. Avg. Std. Rk.
A28 19.1 8.1 26 34 19.0 8.45 22A29 17.5 8.7 12 32 24.4 8.45 30A30 20.2 12.3 15 28 10.2 8.20 5A31 16.8 9.7 7 24 16.8 5.60 17A32 17.6 9.6 19 17 19.5 11.02 23A33 15.3 9.3 24 23 17.1 8.25 18A34 19.0 10.1 16 27 22.2 8.68 29A35 20.0 10.1 13 35 17.6 10.13 20A36 18.3 12.1 18 36 25.5 9.66 32B1 6.4 3.8 6 2 4.5 2.94 3B2 6.1 3.6 4 5 3.3 3.13 2B3 4.7 3.5 2 11 7.4 2.73 8B4 5.0 3.3 3 8 4.6 1.56 4B5 7.0 3.2 8 1 3.1 1.70 1B6 6.3 3.4 5 9 6.1 3.08 7B7 6.4 3.2 6 4 9.7 1.90 11B8 8.0 3.7 11 3 10.7 1.10 12B9 8.6 1.8 12 7 8.7 2.57 1B10 4.5 3.0 1 10 7.4 2.69 8B11 7.0 2.4 8 6 5.9 2.47 6B12 7.1 3.0 10 12 4.7 3.10 5C1 2.3 1.1 2 1 1.8 1.08 1C2 2.7 1.2 4 3 2.6 0.92 3C3 2.6 1.1 3 2 2.9 1.04 4C4 2.2 1.0 1 4 2.5 1.12 2D1 1.4 0.5 1 1 1.4 0.49 1D2 1.6 0.5 2 2 1.6 0.49 2
Table 6Similarity analysis of the three evaluation approaches.
Index Random vs. Expert EPE vs. Expert
Similarity 51.06% 69.27%Improvement ratio 35.78%
580 Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
evaluation results determined via the experts (i.e., the DC managers)have to be rechecked in order to obtain the reasonable performanceevaluation. However, it might take the DC a lot of time and cost torecheck the evaluation results. Thus, a systematical approach (e.g., theEPE model) should be developed in order to acquire reasonableperformance evaluation results.
• Analysis of reasonableness of final employee performance ranks
The final employee performance ranks determined by expertswere regarded as the nominal employee performance ranks. Reason-ableness of the final employee performance ranks can be determinedby analyzing whether the final employee performance ranks fall intothe acceptable rank intervals accepted by the DC managers. Thereasonableness of final employee performance ranks determined viathe EPE approach can be derived via the following procedure.
(1) Determine the confidence intervals of employee performanceranks based on the average and standard deviation of finalemployee performance ranks. The confidence intervals can beregarded as the reasonable rank ranges of employee perfor-mance ranks accepted by the DC managers.
(2) Analyze whether the final employee performance ranks fallinto the reasonable rank ranges in order to obtain areasonableness index for the final employee performanceranks. If a final employee performance rank falls into thereasonable rank range, the corresponding reasonablenessindex is “1”; otherwise, “0” is assigned.
(3) Accumulate the reasonableness indices of the EPE approach toobtain the total number of acceptable employee performanceranks. The number of acceptable employee performance ranksdetermined via the EPE approach is “49”.
(4) Determine the acceptance rate of final employee performanceranks via the ratio of “the number of accepted employeeperformance ranks” to “the number of employees”. Therefore,the acceptance rate of the EPE approach is “90.74%” (i.e., 49/54).
As a result, the final employee performance ranks determined viathe EPE approach are acceptable by the DC managers. Therefore, theEPE model can effectively generate reasonable and acceptableperformance evaluation results.
• Analysis of similarity of final employee performance ranks
The final employee performance ranks of the expert evaluationwere regarded as the nominal employee performance ranks. Theconsistency of the final employee performance ranks derived via therandom and EPE approaches between the ones derived via expert
Table 5Rank sequence patterns and pairwise sequences of three evaluation approaches.
Approach Sequence First-line
Random (Ra) Rank sequence pattern A12→A22→A11→…→A9→A1Paired sequence (A12,A22), (A12,A11), … (A9,A1)Number 630
EPE (Mp) Rank sequence pattern A22→A13→A14→…→A35→APaired sequence (A22,A13), (A22,A14), …, (A35,A3Number 630
Expert (Ex) Rank sequence pattern A13→A1→A5→…→A18→A21Paired sequence (A13,A1), (A13,A5), …, (A18,A21)Number 630
Identical paired sequencesbetween (Ra) and (Ex)
(A13,A1), (A13,A25), …,(A18,A21)
Number 319Identical paired sequencesbetween (Mp) and (Ex)
(A22,A3), (A22,A2), …,(A35,A36)
Number 446
evaluation was analyzed. Before analyzing the consistency, the finalemployee performance ranks have to be processed according to thefollowing procedure.
(1) Determine the rank sequence pattern (e.g., A12→A22→A11→…→A9→A1) with respect to each employee levelbased on the final ranks (as shown in Table 3).
(2) Acquire the pairwise sequences of employees (e.g., (A12,A22),(A12,A11),…, (A9,A1)). The numbers of pairwise sequences forthe first-line, team-level, office-level and division-level em-ployees are C2
36=630, C212=66, C24==6 and C22=1. Therefore,
the total number of pairwise sequences corresponding to eachapproach is 703 (i.e., 630+66+6+1).
(3) Identify the identical pairwise sequences in the “random andexpert evaluations” and in the “EPE approach and expertevaluations”. In the random and expert evaluations, thenumbers of identical pairwise sequences of the first-line,team level, office level and division level employees are 319,36, 3 and 1 respectively. In the EPE approach and expertevaluations, the numbers of identical pairwise sequences forthe four distinct levels of employees are 446, 37, 3 and 1respectively.
As shown in Table 5, among the 703 pairwise sequences withrespect to the random evaluation, a total of 359 (i.e., 319+36+3+1)pairwise sequences (i.e., (A13,A1), (A13,A25), …, (C4,C2), (D1,D2))are identical to the pairwise sequences with respect to the expertevaluation. That is, the similarity between the random and expertevaluations is 51.06% (i.e., 359/703). Similarly, the similarity betweenEPE approach and expert evaluations is 69.27% (i.e., 487/703). As aresult, the improvement ratio of the similarity is 35.78% (Table 6).
According to the analysis results of the three approaches, the EPEmodel can effectively estimate the employee performance (similar tothe results determined by several DC managers), and the employeeperformance generated by the EPE model can be accepted by the DCmanagers. Therefore, the inexperienced DC managers can used theEPE model to estimate the employee performance and regard theemployee performance generated by the EPE model as the initialevaluation results in order to reduce the time and cost required forevaluating the employee performance.
Team-level Office-level Division-level
B10→B3→B1→…→B8→B9 C3→C1→C4→C2 D1→D2(B10,B3), (B10,B1), … (B8,B9) (C3,C1), (C3,C4), … (C4,C2) (D1,D2)66 6 1
36 B5→B1→B8→…→B3→B12 C1→C3→C2→C4 D1→D26) (B5,B1), (B5,B8), … (B3,B12) (C1,C3), (C1,C2), … (C2,C4) (D1,D2)
66 6 1B5→B2→B1→…→B7→B8 C1→C4→C2→C3 D1→D2(B5,B2), (B5,B1), … (B7,B8) (C1,C4), (C1,C2), … (C2,C3) (D1,D2)66 6 1(B5,B12), (B5,B11), …, (B7, B7) (C1,C4), (C1,C2), (C4,C2) (D1,D2)
36 3 1(B5,B1), (B5,B8), …, (B6,B3) (C1,C3), (C1,C2), (C1,C4) (D1,D2)
37 3 1
581Y.-J. Wu, J.-L. Hou / Decision Support Systems 48 (2010) 568–581
6. Conclusion
To accurately estimate the employee performance in a DC, thispaper developed an integrated model to calculate the employeeperformance based on the logistics operation data and businessinformation. In the proposed model, three critical modules includingDPD, IPD and PSA were developed to estimate the logisticsperformance of employees at distinct levels in a DC organizationhierarchy. According to the proposed model, this paper alsoestablished a web-based LIM platform to reduce the workload of DCdecision makers, to assist in the management of logistics operationdata and to support the bottleneck analysis of employees and logisticstasks. In addition, the Nung Hsueh DC in Taiwan was used to analyzeapplicability of the EPE model and the LIM platform. According toanalysis results of the random approach, EPE approach and expertevaluation, the proposedmodel and platform can effectively assist theinexperienced DC managers to acquire the employee performanceaccepted by the experienced DC managers. As a whole, this paperpresents a feasible approach for LSPs to accurately determineemployee performance.
However, the applications (e.g., identification of employees withsignificant performance increase or decrease) of employee perfor-mance are not investigated in this paper. In order to assist the DCmanagers to manage the employees and logistics tasks, the futureresearch can focus on applying the employee performance generatedvia the EPE model to identify the employees with the significantperformance increase or decrease and that might be beneficial toenhance the operation efficiency.
References
[1] J.H. Ahn, S.G. Chang, Assessing the contribution of knowledge to businessperformance: the KP3 methodology, Decision Support Systems 36 (2004)403–416.
[2] C.C. Chen, Performance appraisal of advertising executive in newspaper, GraduateSchool of Communication Management, Ming Chuan University, Masters Thesis,(2001).
[3] S.C., Chen, The performance evaluation model for reception clerks in thehospitality industry, Institute of Human Resource Management, Sun Yat-SenUniversity, Masters Thesis, (2005).
[4] V.I. Coleman, W.C. Borman, Investigating the underlying structure of thecitizenship performance domain, Human Resource Management Review 10 (1)(2000) 25–44.
[5] D.L. Deadrick, D.G. Gardner, Performance distributions: measuring employeeperformance using total quality management principles, Journal of QualityManagement 4 (2) (1999) 225–241.
[6] K. Eguchi, Job transfer and influence activities, Journal of Economic Behavior andOrganization 56 (2005) 187–197.
[7] A. Golec, E. Kahya, A fuzzy model for competency-based employee evaluation andselection, Computers & Industrial Engineering 52 (2007) 143–161.
[8] E. Kahya, Revising the metal industry job evaluation system for blue-collar jobs,Compensation Benefits Review 38 (6) (2006) 49–63.
[9] E. Kahya, The effects of job characteristics and working conditions on jobperformance, International Journal of Industrial Ergonomics 37 (2007) 515–523.
[10] J.S. Kane, K.F. Kane, The analytic framework: the most promising approach for theadvancement of performance appraisal, Human Resource Management Review2 (1) (1992) 37–70.
[11] C.Y. Laio, Free-air television stations technicians performance appraisal research,Graduate School of Communication Management, Ming Chuan University,Masters Thesis, (2000).
[12] K.H. Liu, Application of analytical hierarchy process on the research of discussionof the performance appraisal of room attendant in hotel industry, Graduate Schoolof Communication Management, Ming Chuan University, Masters Thesis, (2004).
[13] C.J. Mcnair, R.L. Lynch, K.F. Cross, Do financial and non-financial measures ofperformance have to agree? Management Accounting (November, 1990) 28–36.
[14] S.H. Ow, H.W. Chen, An overview of the development of a computerized employeeperformance measurement tool-ECAS, Chiang Mai University Journal 5 (2) (2006)229–241.
[15] H.P. Sims, A.D. Szilagyi Jr., R.T. Keller, The measurement of job characteristics,Academy of Management Journal 19 (2) (1976) 195–212.
[16] P.L. Sonja, Personnel selection fuzzy model, International Transactions inOperational Research 8 (1) (2001) 89–105.
[17] H.M. Tsai, Constructing performance evaluation criteria with AHP for nurseanesthetist, Institute of Health CareManagement, Sun Yat-Sen University, MastersThesis, (2005).
[18] J.T. Tzeng, A study on applying analytic hierarchy process to technicianperformance evaluation for the paper industry, Institute of Human ResourceManagement, Sun Yat-Sen University, Masters Thesis, (2004).
[19] S.B. Yaakob, S. Kawata, Worker's placement in an industrial environment, FuzzySets and Systems 106 (3) (1999) 289–297.
Yu-JenWu is a currently Ph.D. student in the Department of Industrial Engineering andEngineering Management at NTHU. His research interests are knowledge managementand logistics management.
Jiang-Liang Hou is a professor in the Department of Industrial Engineering andEngineering Management at National Tsing-Hua University (NTHU). Dr. Hou receivedhis Ph.D. in Industrial Engineering at NTHU and his research interests are knowledgemanagement and logistics management. He has participated in several industrialprojects with high-tech companies and non-profit R&D centers in Taiwan.
