Evaluating the cross-efficiency of information sharing in supply chains
Ming-Min Yu
a,1, Shih-Chan Ting
a,*, Mu-Chen Chen
b,2a
Department of Transportation and Navigation Science, National Taiwan Ocean University, No. 2, Pei-Ning Road, Keelung 20224, Taiwan, ROC b
Institute of Traffic and Transportation, National Chiao Tung University, 4F, No. 118, Section 1, Chung Hsiao W. Road, Taipei 10012, Taiwan, ROC
a r t i c l e
i n f o
Keywords:
Supply chain management Data envelopment analysis Simulation
Information sharing Cross-efficiency Bullwhip effect
a b s t r a c t
Supply chain management integrates the intra- and inter-corporate processes as a whole system. Through information technology, companies can efficiently manage the product flow and information related to the issues such as production capacity, customer demand and inventory at lower costs. Infor-mation sharing can significantly improve the performance of the supply chain, how the different combi-nation of information sharing affects the performance is not yet understood. This study designs different information-sharing scenarios to analyze the supply chain performance through a simulation model. Since there are not only desirable measures but also undesirable measures in supply chains, the usual data envelopment analysis (DEA) model allows measuring performance for complete weight flexibility. In this paper, a cross-efficiency DEA approach is applied to solve this problem. We identify the most effi-cient scenario and estimate the each efficiency of information-sharing scenarios. Contrary to the previous findings in the literature suggesting sharing as much as information possible to increase benefits, the results of this study show that the scenario of demand information sharing is the most efficient one. In addition, sharing information on capacity and demand, and full information sharing in general are good practices. Sharing only information on capacity and/or inventory information, without sharing information on demand, interferes with production at manufacturers, and causes misunderstandings, which can magnify the bullwhip effect.
Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction
A supply chain is a logistics network, which consists of all stages (e.g. order processing, purchasing, inventory control, manufactur-ing, and distribution) involved in producing and delivering a final product or service. The entire chain connects customers, retailers, distributors, manufacturers and/or suppliers, beginning with the creation of raw material or component parts by suppliers and end-ing with consumption of the product by customers. Supply chain management (SCM) is related to the coordination of materials, products and information flows among suppliers, manufacturers, distributors, retailers and customers (Simchi-Levi, Kaminsky, & Simchi-Levi, 2000). SCM often needs the integration of inter- and intra-organizational relationships and coordination of different types of flows within the entire chain. With sharing information between trading partners and coordinating their replenishment and production decisions under demand uncertainty, it could be possible to further reduce costs and improve customer service le-vel. The performance of a supply chain could be influenced by
many factors, among which information sharing is the crucial one. Inter-company integration and coordination via information technology play a key role in improving supply chain performance. The application of current information technology, such as elec-tronic data interchange (EDI) and the internet has helped compa-nies to share information and has improved supply chain order fulfillment performance. Sharing both supply and demand infor-mation substantially reduces inventory costs in make-to-stock or assemble-to-order production. Sharing supply information also substantially reduced order cycle time in an assemble-to-order environment (Strader, Lin, & Shaw, 1999).
A supply chain is fully coordinated when all decisions are aligned to approach global system objectives. Lack of coordination occurs when decision makers have incomplete information or incentives that are not compatible with system-wide objectives. Even under conditions of full information availability, the perfor-mance of the supply chain can be sub-optimal when each decision maker optimizes one’s individual objective (Sahin & Robinson, 2002). One line of related research analyzes the benefits of sharing customer demand information with members of the supply chain.
Bourland, Powell, and Pyke (1996)analyze the savings in inventory cost that can be realized when a manufacturer shares point-of-sale (POS) data with suppliers.Ernst and Kamrad (1997)consider a sup-ply chain in which manufacturers and retailers share demand information and analyze the impact of information sharing on 0957-4174/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2009.09.048
* Corresponding author. Tel.: +886 2 24622192x7050; fax: +886 2 24633745. E-mail addresses: yumm@mail.ntou.edu.tw (M.-M. Yu), ericting@mail.ntou. edu.tw(S.-C. Ting),ittchen@mail.nctu.edu.tw(M.-C. Chen).
1
Tel.: +886 2 24622192x7017; fax: +886 2 24633745. 2 Tel.: +886 2 2349 4967; fax: +886 2 2349 4953.
Contents lists available atScienceDirect
Expert Systems with Applications
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e s w aservice level.Lee, Padmanabhan, and Whang (1997)prove that de-mand variability can be amplified in the supply chain as the orders are passed from retailers to distributors. Therefore, accurate fore-casts can significantly influence the performance of the supply chain in terms of inventory cost, backorders or loss of sales, and good will. Inaccurate forecasts can also cause low utilization of capacity and other problems in production. Cachon and Fisher (2000) and Lee et al. (2000)analyze the benefits of sharing real-time information on demand and/or inventory levels between sup-pliers and customers.Cachon and Fisher (2000)study the value of sharing demand and inventory data. The authors compare a tradi-tional information policy that does not use information-sharing with a full information policy.Lee, So, and Tang (2000)analyze the benefits of demand-side information sharing with a two-eche-lon supply chain. They suggest that this kind of information shar-ing alone could provide significant inventory reduction and cost savings to the manufacturer.Thonemann (2002)derives a better understanding of the benefits of advance demand information (ADI) to identify conditions under which sharing ADI results in sig-nificant cost savings. A typical model is used to capture the basic aspects of a supply chain in which ADI is shared. This enables them to derive analytical results and to gain structural insight into the ADI-sharing problem. The results can be used by decision makers to analyze the cost savings that can be achieved by ADI and help them determine if sharing ADI is beneficial for their supply chain. Another line of related research analyzes the impact of informa-tion sharing on the bullwhip effect and/or the performance of a supply chain.Metters (1997)studies the impact of the bullwhip ef-fect on profitability by establishing an empirical lower bound on the cost excess of the bullwhip effect. Results indicate that the importance of the bullwhip effect to a firm differs greatly between specific business environments, and eliminating the bullwhip ef-fect can increase product profitability by 10–30% under some con-ditions.Chen, Drezner, Ryan, and Simchi-Levi (2000a) Chen, Ryan, and Simchi-Levi (2000b)quantify the bullwhip effect for a simple, two-echelon supply chain consisting of a single retailer and a sin-gle manufacturer. They assume that demand follows an AR(1) pro-cess, and the retailer uses a moving-average model for demand forecast and a simple order-up-to inventory policy for replenish-ment. They conclude that the variance of the orders is always high-er than the variance in demand. Furthhigh-ermore, the magnitude of the variance is significantly influenced by the number of observations used in the moving-average, the lead time between the retailer and the manufacturer, and the correlation coefficient in the demand function. They extend the analytical model to a multiple-echelon supply chain and find that the bullwhip effect could be reduced, but not completely eliminated, by sharing demand among all par-ties in the supply chain.Chen et al. (2000a, 2000b)investigate the impact of forecast methods and demand patterns on the bullwhip effect. They compare an exponential- smoothing forecasting model and a moving-average model, in which the demand is correlated with a linear trend. They find that reduction in ordering lead time and using more demand information in forecasting (a smoother forecast) could decrease the bullwhip effect. Another finding is that negatively correlated demand could lead to a larger increase in or-der variability than positively correlated demand, and that a retai-ler forecasting demand with a linear trend will have more variable orders than a retailer forecasting i.i.d. demand. These two papers evaluate the magnitude of the variance amplifications in the sup-ply chain by considering alternative demand processes and fore-casting models for a simple supply chain structure. However, they do not consider the impact of the variance amplifications on the costs and service levels of the supply chain, nor do they con-sider the costs of either inventory, ordering or setup, or production decisions by the manufacturers.Zhao, Xie, and Leung (2002a) pres-ent the impact of information sharing and ordering coordination
on the performance of a supply chain with one capacity-limited supplier and multiple retailers under demand uncertainty.Zhao, Xie, and Leung (2002b)also present the impact of forecasting mod-el smod-election on the value of information sharing in a supply chain with one capacitated supplier and multiple retailers. Using a com-puter simulation model, this study examines demand forecasting and inventory replenishment decisions by the retailers and pro-duction decisions by the supplier under different demand patterns and capacity tightness. The simulation output indicates that the selection of the forecasting model significantly influences the per-formance of the supply chain and the value of information sharing. Furthermore, demand patterns faced by retailers and capacity tightness faced by the suppliers also significantly influence the va-lue of information sharing. The results also show that substantial cost savings can be achieved through information sharing and then motivating trading partners to share information in the supply chain.
Information sharing can significantly improve the performance of a supply chain. Additionally, companies can redesign their sup-ply chain strategies through information sharing to increase profit. Many studies demonstrate the positive impact of information shar-ing on a supply chain. However, few studies focus on how the dif-ferent combinations of information sharing affect the performance of a supply chain. Provided that the entities of supply chain are aware about how they can benefit from the information sharing, they are more willing to share the necessary information. The pur-pose of this paper is to examine how the different information sharing among the entities influences the performance of the sup-ply chain, and to address the problem of selecting the most appro-priate information sharing for the supply chain partners. This study designs different information-sharing scenarios to analyze the sup-ply chain performance. To measure the performance of each sce-nario, it is necessary to consider not only the desirable indices but also undesirable indices. Thus, the usual data envelopment analysis (DEA) model is applied to measure the performance for complete weight flexibility.
The remainder of the paper is organized as follows. In the next section, the information-sharing scenarios are specified. A brief introduction of cross-efficiency DEA and the analysis for evaluating performance are described in the following section. The analysis and results are demonstrated in Section4. In the final section, some conclusions and recommendations for further research follow.
2. Information-sharing scenarios
We develop a supply chain simulation model (shown inFig. 1) which considers a multi-echelon supply chain (i.e. retailers, dis-tributors, manufacturers and suppliers) and nine information-sharing scenarios. In the first information-information-sharing scenario, denoted by N, no information will be shared between the entities. The second scenario is partial information sharing, which consists of six combinations: (1) C: capacity information sharing; (2) D: de-mand information sharing; (3) I: inventory information sharing; (4) D&C: demand and capacity information sharing; (5) D&I: de-mand and inventory information sharing; (6) C&I: capacity and inventory information sharing. The third scenario, denoted by F is full information sharing with capacity, demand and inventory. The fourth scenario is strategic alliance of supply chain (vendor managed inventory, VMI, is adopted herein).
To compare the performance of each information-sharing sce-nario, a simulation tool, Rockwell Software Arena v5.0, is utilized to analyze performance indices (shown inTable 3). Input parame-ters such as initial inventory level, inventory policy, lead times of production and transportation, customer demand rate, and unit production time are shown inTables 1 and 2.
The above simulation results of performance indices include to-tal costs (consisting of inventory holding cost, shortage cost and or-der cost), fulfillment rate, customer service level and oror-der cycle time. From the performance measures of each scenario illustrated inTable 3, we cannot easily determine the most appropriate sce-nario with respect to the performance data of these eight informa-tion-sharing scenarios, except for the scenario of VMI, since each performance measure is relatively prominent. This can be seen as a problem of discrete alternative multiple criteria evaluation, which is formulated by considering a set of alternatives and a set
of criteria. The aggregation and comparison of various alternatives are based on the values for each criterion. In most approaches, the multi-criteria evaluation for an alternative is represented by a vec-tor of the performance of the alternative on each criterion. This information is then used within the outranking methods to carry out relative rankings and performance evaluations among the val-ues of the alternatives for a given criteria.
Salminen, Hokkanen, and Lahdelma (1998)compare a number of models and tools based on outranking approaches for multiple criteria decision making (MCDM) and a multi-attribute rating
Suppliers Manufacturers Distributors Retailers
Supply Supply Supply Demand Demand Demand
Supply Chain Simulation Model
1. C: Capacity Information 2. D: Demand Information 3. I: Inventory Information Information Sharing 1. Inventory:
Initial inventory level Inventory policy
Shortage costs (Finished goods) Holding costs (Finished goods) Order costs (Finished goods) 2. Lead Times:
Production Transportation 3. Customer Demand Rate 4. Unit Production Times Parameters N: No Information Sharing Colaboration: Vendor Managed Inventory F: Full Information Sharing Partial Information Sharing:
1. C: capacity 2. D: demand 3. I: inventory
4. D&C: demand and capacity 5. D&I: demand and inventory 6. C&I: capacity and inventory
Scenario 4 Scenario 3 Scenario 2 Scenario 1 1. Shortage costs 2. Holding costs 3. Order costs 4. Total costs 5. Fulfillment rate 6. Customer service level 7. Order cycle time
Performance Index
Cross-efficiency Evaluation by DEA Ranking:
1. D: demand 2. F: full Information 3. D&C: demand and capacity 4. D&I: demand and inventory 5. N: no Information 6. I: inventory 7. C: capacity
8. C&I: capacity and inventory
technique. The outranking techniques were applied with a number of actual decision makers (DMs) providing preference weights for each of the major criteria. The outranking approaches typically re-quire three sets of inputs: preference weights for the criteria, pref-erence and indiffpref-erence thresholds and veto thresholds. The usual approach would be to have the DM provide information relative to the preference weights and some information on veto thresholds levels. The determination of this information for large groups may be cumbersome. The DEA models require little input from the DM.Doyle (1995)has supported the use of the DEA as a lazy DM’s methodology for MCDM. This outlook may alternatively be looked at as a reactive approach to MCDM. That is, the DMs, or their preferences, play an insignificant role in the ranking of alter-natives. Meanwhile those usual MCDM weighting methods to aggregate different criteria into one performance index are more subjective, thus DEA methodology is utilized to measure cross-effi-ciency between different information-sharing scenarios.
3. Methodology
This paper presents an atypical application of data envelopment analysis (DEA) methodology to measure performance of coordina-tion and informacoordina-tion sharing between the supply chain entities at different information sharing scenarios. The DEA method, first pro-posed byCharnes, Cooper, and Rhodes (1978), is known as an eval-uation technique for performance analysis of various entities whose production activities are characterized by multiple inputs
and outputs. A reader can see more details of the DEA method in
Boussofiane, Dyson, and Thanassoulis (1991), Charnes, Cooper, and Lewin (1994), Seiford and Thrall (1990). Nowadays, DEA has become one of the most popular fields in operations research, with applications involving a wide range of contexts. The applicability and practicality of DEA can be easily confirmed inCooper, Huang, and Li (1996), Cooper, Thompson, and Thrall (1996)and numerous previous research efforts. The DEA method is utilized to analyze the performance with multiple inputs and outputs. Thus, we apply this method to evaluate SC information sharing performance. In the supply chain, each unit is permitted to choose the most favor-able weights to be applied to its standings (in our case, the differ-ent information-sharing scenarios are compared by analyzing the resulting performance measures including total cost, order fulfill rate, customer service level and order cycle time) in the usual DEA manner. In the evaluation of this simple efficiency score, the usual DEA model allows for complete weight flexibility. A unit achieves a relative efficiency score of 1 by heavily weighting few favorable inputs and outputs, and completely ignoring the other inputs and outputs. Such units perform well with respect to few in-put/output measures. Thus, considering the scenarios with an effi-ciency score of 1 as the candidates with the best combination of specifications is inappropriate.Cook and Kress (1990)consider a scheme involving an imposed set of weights, which do not provide a fair overall assessment. Nevertheless, the problem of choosing the most favorable weights to be applied to each unit’s standings is still not resolved. The simple efficiency score obtained from Cook and Kress’s model is often misleading. To overcome such problems, a measure more than the simple efficiency score is required in the decision making process. In this section, we provide a review of ba-sic DEA and a cross-efficiency ranking extension to the DEA models and how they may be used to help evaluate discrete alternative MCDM models.
Traditionally, one method for resolving this problem is for the poll organizer to impose a predetermined set of weights on each alternative’s standing. Thus the composite score, Zi, of alternative
i would be given by:
Zi¼ Xk
j¼1
wj
v
ij ð1Þwhere
v
ij represents the value of jth attribute of alternative i(i ¼ 1; 2; . . . ; m; j ¼ 1; 2; . . . ; kÞ, and wjdenotes the weight of the jth
attribute.
The CCR model was initially proposed byCharnes et al. (1978). For each DMU, the CCR model tries to determine the optimal Table 1
Initial inventory level and inventory policy.
Inventory Retailer 1 Retailer 2 Retailer 3 Distributor Manufacturer
Initial inventory level 40 40 40 130 200
Inventory policy (s, S): s 21 24 21 61 83
Inventory policy (s, S): S 39 43 39 129 153
Table 2
Parameters for simulation.
Parameters Inputs
Iteration 30
Simulation time 120 days
Interval distribution of customer order
Exponential distribution (mean = 0.15 day) Quantity distribution of
customer order
Discrete distribution (Q = 1 or 4, Prob. = 0.167; Q = 2 or 3, Prob. = 0.333)
Frequency of
replenishment review
Once daily Transportation lead times 1 (day)
Production lead times Normal distribution (mean = 0.1 h, standard deviation = 0.02 h)
Unit holding costs 1
Unit shortage costs 5
Order costs 10 (retailers), 50 (distributors and manufacturers)
Table 3
Simulation results of performance indices.
Scenario N C D I D&C D&I C&I F VMI
Shortage costs 219.78 66.15 29.44 109.84 26.088 54.05 95.80 24.79 –
Holding costs 103.06 293.63 233.36 189.94 401.25 262.12 371.13 479.88 –
Order cost 130 180 130 130 180 130 180 180 –
Total costs 452.84 539.78 392.8 429.77 607.34 446.17 646.94 684.68 241.08
Fulfillment rate (%) 65.22 75.25 79.26 72.16 79.13 77.04 72.36 79.35 98.50
Customer service level (%) 61.83 72.62 77.44 68.84 77.26 74.66 76.42 81.30 98.09
weight of jth attribute of alternative i, wij, using linear
program-ming so as to maximize the composite score Zii which are used
in the objection function (2) to emphasize that this is alternative i’s own evaluation of its own desirability.
Maximize Zii¼ Xk j¼1 wij
v
ij ð2Þ Subject to Zip¼ Xk j¼1wij
m
pj6 1for all DMUs p; including i ð3ÞwijP0 ð4Þ
where Zip denotes cross-efficiency of alternative i’s evaluation of
alternative p’s desirability i.e. DMU p is evaluated by the weights of DMU i. Constraints (3) represent that no alternative p should have a desirability greater than 1 under i’s weights.
The optimal weights may vary from one DMU to another. Thus, the weights in DEA are derived from the data instead of being fixed in advance, such as given by decision makers. Each DMU is as-signed a best set of weights with values that may vary from one DMU to another.Cook and Kress (1990)suggest that each alterna-tive be allowed to propose its own weights in order to maximize its own desirability subject to certain reasonable constraints on the desirability of all the alternatives. Sexton, Silkman, and Hogan (1986)argue that decision makers do not always have a reasonable mechanism from which to choose assurance regions. Thus they recommend the cross-evaluation matrix (CEM) for ranking alterna-tives. Cross-efficiencies in DEA can effectively be used to surmount the problems associated with simple efficiency scores. Cross-efficiencies of a DMU provide information on how well it is per-forming with the optimal DEA weights of other m 1 DMUs. The cross-efficiencies of all the DMUs can be arranged in a CEM as shown inTable 4. The pth row and ith column of the CEM repre-sents the cross-efficiency of DMU p with the optimal weights of DMU i. The usual simple efficiency measurements for each DMU are found in the leading diagonal of this matrix. The cross-efficiency method simply calculates the cross-efficiency score of each DMU m times using the optimal weights evaluated by m LPs. The cross-efficiency ranking method in the DEA context utilizes the results of cross-efficiency matrix Zipin order to rank scale the
DMUs. It could be argued that Zi¼Pmp¼1Zpi=m is more
representa-tive than Zii, the standard DEA efficiency score, since all the
ele-ments of the cross-efficiency matrix are considered, including the diagonal. While the standard DEA score, Zii, is non-comparable,
since each uses different weights, the Ziis comparable because it
uses the weights of all units equally (i.e. all the units’ standing). A limitation with the CEM evaluated fromSexton et al. (1986)
model weights is that the optimal weights obtained from their model may not be unique. This condition occurs if multiple opti-mum solutions exist. This ambiguity can be solved by using formu-lations proposed byDoyle and Green (1994). These formulations can be categorized into aggressive and benevolent approaches, in whichDoyle and Green (1994)not only maximize the efficiency
of target DMU, but also take a second goal into account. The second goal, in the case of aggressive formation, is to minimize the effi-ciency of the composite DMU constructed from other m 1 DMUs. The aggressive formulation is shown below:
Minimize X k j¼1 wij Xm p¼1;p – i
m
pj ! ð5Þ Subject to Zip61 for all DMUs i – p ð6ÞXk j¼1
wij
m
ij Zii¼ 0 ð7ÞwijP0 ð8Þ
where DMU i is the target DMU, Pkj¼1 wijPmp¼1;p – i
m
pj
is the weighted attributes of composite DMU, and Zii is the simple
effi-ciency of DMU i obtained from usual DEA.
Maximizing the other DMUs’ cross-efficiencies in the same way is known as a benevolent formulation:
Maximize X k j¼1 wij X p¼1;p–immpj 0 @ 1 A ð9Þ
Subject to Zip61for all DMUs i–p ð10Þ Xk
j¼1
wij
m
ij Zii¼ 0 ð11ÞwijP0 ð12Þ
When aggressive models (5)–(8) are solved for alternative i, as well as obtaining Zii, we are also provided with values Zipwhich
can be thought of as evaluations of p’s desirability from i’s point of view within this modeling framework. The values obtained in a complete run of the model can be organized in a matrix Z in which the values down a column p ðZpÞ represent how alternative
p is appraised by all alternatives, and values across a row i ðZiÞ
rep-resent how alternative i appraises all alternatives. Thus, this matrix can be regarded as the summary of a self- and peer-appraisal pro-cess in which on-diagonal elements represent self-appraisals, and off-diagonal elements represent peer-appraisals.
Sexton et al. (1986)propose the column averages of Z as suit-able overall ratings of the alternatives. In essence, each alternative is being accorded a weight of 1=m in determining any alternative’s overall rating. In order to mitigate the rank reversal effect,Green, Doyle, and Cook (1996)relax the assumption that each alternative be accorded a weight of 1=m in the establishment of overall rat-ings. They suggest that each alternative apply a weight in propor-tion to its original overall rating rather than uniformly 1=m. 4. Results and discussions
The data for this study are shown inTable 3. A total of 8 scenar-ios and six criteria (performance measures) are introduced. The six performance measures include three minimizing criteria (holding
Table 4
Matrix of cross-efficiencies for m DMUs.
Rating DMU Rated DMU
1 2 3 . . . m Averaged appraisal of peer
1 Z11 Z12 Z13 . . . Z1m B1 2 Z21 Z22 Z23 . . . Z2m B2 3 Z31 Z32 Z33 . . . Z3m B3 .. . .. . .. . .. . . . . ... .. . m Zm1 Zm2 Zm3 . . . Zmm Bm Averaged peer-appraisal Z1 Z2 Z3 . . . Zm
cost, shortage cost, order cost and order cycle time). The remaining criteria (fulfillment rate, customer service level) are defined as the maximizing criteria. There are many studies which consider treat-ing minimiztreat-ing criteria. We begin with a brief summary of these recent works related to the treatment of minimizing criteria into four categories. First are some studies which regard them as fol-lowing weak disposability, such asFäre, Grosskopf, Lovell, and Pas-urka (1989), Boyd and McClelland (1999), Zofío and Prieto (2001). Weak disposability indicates that the undesirable outputs can be reduced only at the expense of a reduction in the other outputs or an increase in the use of inputs. The second possibility is to envi-sion them as inputs, such asHaynes, Ratick, and Cummings-Saxton (1994) and Korhonen and Luptacik (2004). This method considers both inputs and undesirable outputs (minimizing criteria) to have the same improvable direction when inefficient DMUs wish to im-prove their performance, and then, undesirable outputs are treated as inputs. Thirdly, some studies treat them as desirable outputs by taking their reciprocal, such asLovell, Pastor, and Turner (1995), and the fourth group treats them by subtracting them from some sufficiently large numbers, such asSeiford and Zhu (2002), Jahan-shahloo, Hadi Vencheh, Foroughi, and Kazemi Matin (2004). The translated data in the third and fourth categories have the same improvable direction with desirable output, and then the efficiency scores will be obtained by employing a traditional DEA model. In our study, we treat those minimizing criteria as desirable outputs by taking their reciprocal, since there is no production relationship between minimizing criteria and maximizing criteria.
Models (2)–(4) are initially used to obtain the simple efficiency of all SCM information-sharing scenarios. The standard DEA identi-fied scenarios N, D, I, D&I, and F to be efficient with a relative effi-ciency score of 1. The remaining 3 scenarios (C, D&C and C&I) obtained an efficiency score of less than 1. No specific argument is advanced for preferring an aggressive over a benevolent ap-proach. However, since the major interest is in finding the best SCM information sharing rather than a group of projects to make up a program, an aggressive approach, in the eye of some neutral evaluator, may be seen as appropriate in this context. Thus, simple efficiency scores are then used in aggressive models (5)–(8) to ob-tain the optimal attribute weights for each scenario. These weights also minimize the relative efficiency of the composite scenarios that is constructed from the remaining m 1 scenarios for each case. Such a matrix and overall rating is shown inTable 5. It is evi-dent from this table that scenarios D and F have several high cross-efficiency values. Some of the simple efficient scenarios such as N and I have several low cross-efficiency values. The adjusted weighted column means of the Z matrix can be used to effectively differentiate among the overall efficient scenarios.
Scenario D&C, which was inefficient with a relative efficiency score of 0.999 and mean score of 0.819, is rated as a better overall performer than efficient scenarios N, D&I and I, and as almost equal
to scenario F. Based on these results, the optimal choice is scenario D – a good overall alternative performing well in many dimen-sions. This methodology allows the decision maker to rank the SCM information-sharing scenarios based on their overall performance.
Demand information has a tendency to amplify, delay and oscil-late from downstream to upstream along the supply chain ( For-rester, 1998; Lee et al., 2000). This information is fundamental and important to supply chain partnership. Furthermore, demand information has a major impact on supply chain performance since it has a direct impact on production scheduling, inventory control and delivery plans (Thonemann, 2002). Therefore, sharing demand information is usually taken as the first step for supply chain part-nership. For example, more than 50% of manufacturers in the per-sonal computer industry share their demand information with suppliers (Austin, Lee, & Kopczak, 1997). From our results shown inTable 5, the scenarios with sharing demand information outper-form the other scenarios.
The results also show that the no information-sharing scenario (N) is better than some partial information-sharing scenarios (C, I, C&I). This seems most unreasonable, but is an interesting and meaningful result. According to the simulation, sharing only capac-ity and/or inventory information, without any demand information sharing, causes interference with production at manufacturers and misunderstandings, and magnifies the bullwhip effect. The busi-ness activities are triggered by demand. The activities, such as pro-duction in the upstream of the supply chain, try to meet the actual demand of end customers. Better meeting of actual demand results in better consequent decisions in the supply chain. Therefore, shar-ing only capacity and/or inventory information, without any de-mand information sharing, may mislead the sales forecast, inventory control and production plan.
5. Concluding remarks and further research
After proceeding with international management, enterprises have to face the challenge of SCM mainly because of the rapid change in the business environment and severe competition in market and customers’ diverse demand. Therefore, how to operate information technology to upgrade the efficiency of a supply chain has currently become one of the most important issues for enter-prises. Information sharing is usually taken as a basic treatment for supply chain collaboration. In a supply chain, more direct and immediate information results in higher accuracy of forecasts. The effective SCM is not achievable by any single enterprise, but in-stead requires a virtual entity by faithfully integrating all involved partners, who should come up with the insightful commitment of real-time information sharing and collaborative management. Thus assessing the effects of different degrees of information
Table 5
Cross-efficiency and overall rating for 8 SCM information sharing scenarios.
N C D I D&C D&I C&I F
N 1.000 0.351 0.442 0.543 0.257 0.393 0.278 0.215 C 0.826 0.949 1.000 0.911 0.996 0.971 0.912 1.000 D 0.483 0.557 1.000 0.563 0.965 0.728 0.474 1.000 I 1.000 0.722 1.000 1.000 0.722 1.000 0.722 0.722 D&C 0.810 0.940 1.000 0.891 0.999 0.965 0.893 1.000 D&I 1.000 0.722 1.000 1.000 0.722 1.000 0.722 0.722 C&I 0.929 0.925 1.000 0.919 0.961 0.958 0.955 1.000 F 0.113 0.375 0.842 0.226 0.950 0.459 0.259 1.000 Overall rating 0.759 0.674 0.914 0.742 0.819 0.800 0.631 0.830 Ranking 5 7 1 6 3 4 8 2
Note: N: non-information sharing, C: capacity information sharing, D: demand information sharing, I: inventory information sharing, D&C: capacity and demand information sharing, D&I: demand and inventory information sharing, C&I: capacity and inventory information sharing, F: full information sharing.
sharing upon multi-echelon supply chain performance has become an important issue.
Differing information-sharing scenarios are compared by analyz-ing the resultanalyz-ing performance measures includanalyz-ing: inventory hold-ing costs, shortage costs and order costs of manufacturers, order filling rates of distributors and retailers, customer service levels, and order cycle time. The ranking of scenarios according to perfor-mance measures is often treated in the literature as the problem of multi-criteria classification of elements of one set. Besides the appli-cation of multi-criteria analysis, this problem has been solved by applying different methods such as regression analysis, cluster methods, and factor analysis. This study aims at using a non-para-metric approach, DEA, to estimate the efficiency of information-sharing scenarios in a supply chain with multiple criteria.
Most applications of DEA to multi-criteria analysis have the lim-itations of the existing methodology intrinsic to DEA. The simple effi-ciency score obtained from standard DEA is often misleading. It is difficult to choose the best alternative. In order to rank 5 efficient alternatives we use an aggressive formulation of Doyle and Green’s DEA cross-efficiency model (Doyle and Green, 1997). A comparison of obtained ranks shows that the scenarios were ranked more realis-tically with the cross-efficiency matrix. The results show that the scenario of demand information sharing is the most efficient. Be-sides, the sharing of information on capacity and demand, and full information sharing in general, are good practices.
The previous findings in the literature suggesting sharing as much as information possible to increase benefits, we contrarily advise to share the information as combination. This research can be extended in several ways. Firstly, different types of inven-tory policy can be applied to comparing the efficiency of informa-tion sharing. Secondly, since the results of the simulainforma-tion show that the demand information is the key enabler for information sharing, the demand information, including the interval and quan-tity distribution of customer orders, can be changed to test the sen-sitivity of parameters. Third, the preference of each managerial factor can be further considered, and how the preferences derived from different managerial factors can be further examined in fu-ture works.
References
Austin, T. A., Lee, H. L., & Kopczak, L. R. (1997). Unlocking hidden value in the personal computer supply chain. (Anderson Consulting).
Bourland, K. E., Powell, S. G., & Pyke, D. F. (1996). Exploiting timely demand information to reduce inventories. European Journal of Operational Research, 92, 239–243.
Boussofiane, A., Dyson, G. R., & Thanassoulis, E. (1991). Applied data envelopment analysis. European Journal of Operational Research, 52, 1–15.
Boyd, G. A., & McClelland, J. D. (1999). The impact of environmental constraints on productivity improvement in integrated paper plants. Journal of Environmental Economics and Management, 38, 121–142.
Cachon, G. P., & Fisher, M. (2000). Supply chain inventory management and the value of shared information. Management Science, 46, 1032–1048.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2, 429–444.
Charnes, A., Cooper, W. W., & Lewin, L. M. (1994). Data envelopment analysis: Theory methodology and application. Boston: Kluwer.
Chen, F., Drezner, Z., Ryan, J. K., & Simchi-Levi, D. (2000a). Quantifying the bullwhip effect in a supply chain: The impact of forecasting, lead times, and information. Management Science, 46, 436–443.
Chen, F., Ryan, J. K., & Simchi-Levi, D. (2000b). The impact of exponential smoothing forecasts on the bullwhip effect. Naval Research Logistics, 47, 269–286. Cook, W. D., & Kress, M. (1990). A data envelopment model for aggregating
preference rankings. Management Science, 36, 1302–1310.
Cooper, W. W., Huang, Z., & Li, S. X. (1996). Satisfying DEA models under chance constraints. Annals of Operations Research, 66, 279–295.
Cooper, W. W., Thompson, R. G., & Thrall, R. M. (1996). Extensions and new developments in DEA. Annals of Operations Research, 66, 3–45.
Doyle, J. R. (1995). Multi-attribute choice for the lazy decision maker: Let the alternatives decide. Organizational Behavior and Human Decision Processing, 62, 87–100.
Doyle, J. R., & Green, R. H. (1994). Efficiency and cross-efficiency in DEA: Derivations, meanings and uses. Journal of Operational Research Society, 45, 567–578.
Ernst, R., & Kamrad, B. (1997). Allocation of warehouse inventory with electronic data interchange and fixed order intervals. European Journal of Operational Research, 103, 117–128.
Färe, R., Grosskopf, S., Lovell, C. A. K., & Pasurka, C. (1989). Multilateral productivity comparisons when some outputs are undesirable: A nonparametric approach. Review of Economics and Statistics, 71, 90–98.
Forrester, J. W. (1998). Industrial dynamics. Harvard Business Review, July–August, 37–66.
Green, R. H., Doyle, J. R., & Cook, W. D. (1996). Preference voting and project ranking using DEA and cross-evaluation. European Journal of Operational Research, 90, 429–444.
Haynes, K. E., Ratick, S., & Cummings-Saxton, J. (1994). Toward a pollution abatement monitoring policy: Measurements, model mechanics, and data requirements. Environmental Professional, 16, 292–303.
Jahanshahloo, G. R., Hadi Vencheh, A., Foroughi, A. A., & Kazemi Matin, R. (2004). Inputs/outputs estimation in DEA when some factors are undesirable. Applied Mathematics and Computation, 156, 19–32.
Korhonen, P. J., & Luptacik, M. (2004). Eco-efficiency analysis of power plants: An extension of data envelopment analysis. European Journal of Operational Research, 154, 437–446.
Lee, H. L., Padmanabhan, V., & Whang, S. (1997). Information distortion in a supply chain: The bullwhip effect. Management Science, 43, 546–558.
Lee, H. L., So, K. C., & Tang, C. S. (2000). The value of information sharing in a two-level supply chain. Management Science, 46, 626–643.
Lovell, C. A. K., Pastor, J. T., & Turner, J. A. (1995). Measuring macroeconomic performance in the OECD – a comparison of European and Non-European countries. European Journal of Operational Research, 87, 507–518.
Metters, R. (1997). Quantifying the bullwhip effect in supply chain. Journal of Operations Management, 15, 89–100.
Sahin, F., & Robinson, E. P. (2002). Flow coordination and information sharing in supply chains: Review, implications, and directions for future research. Decision Science, 33, 505–536.
Salminen, P., Hokkanen, J., & Lahdelma, R. (1998). Comparing multi-criteria methods in the context of environmental problems. European Journal of Operational Research, 104, 485–496.
Seiford, L. M., & Zhu, J. (2002). Modeling undesirable factors in efficiency evaluation. European Journal of Operational Research, 142, 16–20.
Seiford, L. M., & Thrall, R. M. (1990). Recent developments in DEA: The mathematical programming approach to frontier analysis. Journal of Econometrics, 46, 7–38.
Sexton, T. R., Silkman, R. H., & Hogan, A. J. (1986). Data envelopment analysis: Critique and extensions. In R. H. Silkman (Ed.), Measuring efficiency: An assessment of data envelopment analysis. San Francisco: Jossey-Bass.
Simchi-Levi, D., Kaminsky, P., & Simchi-Levi, E. (2000). Designing and managing the supply chain: Concepts, strategies and case studies. New York: McGraw-Hill. Strader, T. J., Lin, F., & Shaw, M. J. (1999). The impact of information sharing on order
fulfillment in divergent differentiation supply chain. Journal of Global Information System, 7, 17–25.
Thonemann, U. W. (2002). Improving supply-chain performance by sharing advance demand information. European Journal of Operational Research, 142, 81–107. Zhao, X., Xie, J., & Leung, J. (2002a). The impact of information sharing and ordering
coordination on supply chain performance. Supply Chain Management: An International Journal, 7, 24–40.
Zhao, X., Xie, J., & Leung, J. (2002b). The impact of forecasting model selection on the value of information sharing in a supply chain. European Journal of Operational Research, 142, 321–344.
Zofío, J. L., & Prieto, A. M. (2001). Environmental efficiency and regulatory standards: the case of CO2 emissions from OECD countries. Resource and Energy Economics, 23, 63–83.
