The current issue and full text archive of this journal is available at www.emeraldinsight.com/0960-0035.

htm

Exploring the benets of vendor managed inventory

Kazim Sari
Department of International Logistics and Transportation, Beykent University, Istanbul, Turkey
Abstract
Purpose The purpose of this paper is to explore the performance increase achieved by vendor managed inventory (VMI) under different levels of outside supply capacity, demand uncertainty, and lead time. Design/methodology/approach The study uses discrete event simulation to explore the performance increase achieved by VMI under different supply chain scenarios. Findings The analysis suggests that when implementing a VMI program, the capacity restrictions of suppliers have to be taken into consideration even though they have not participated in the program. Furthermore, the results also show that unless the retailer provides additional information to the distributor to resolve the uncertainty, higher levels of uncertainty in market demand create signicant reductions in the savings realized by VMI. Finally, this study proves that, regardless of lead time horizons, VMI provides nearly the same level of performance increase as long as the ratio of the retailers lead time to that of the suppliers remains constant. Practical implications This analysis provides a means for practitioners to understand the impact of various environmental and operational factors on the performance increase achieved by VMI so they can better analyze their specic business conditions to prepare their organizations for more successful VMI programs. Originality/value Although there is a range of research focusing on VMI, only a few of those have tried to identify the factors that play an important role in the failure of VMI programs. While extending the current literature, this is the rst study to explore the impact of outside supply capacity on the performance improvements achieved by VMI. Keywords Supply chain management, Vendor relations, Inventory management, Information strategy, Simulation Paper type Research paper

Exploring the benets of VMI

529
Received March 2007 Revised May 2007 Accepted May 2007

1. Introduction Supply chain management (SCM) is:

. . . the efcient management of the end-to-end process, which starts with the design of the product or service and ends with the time when it has been sold, consumed, and nally, discarded by the consumer (Swaminathan and Tayur, 2003, p. 1387).

By coordinating different enterprises along the logistics network or establishing business partnerships, SCM is concerned with nding the best strategy for the whole supply chain (Simchi-Levi et al., 2003, p. 2). One important issue of nding the best strategy for the whole supply chain is sharing product and production information amongst supply chain members. It has been recognized that information sharing at the retailer level produces signicant benets for the supply chain by reducing the bullwhip effect (Lee et al., 1997a; Chen et al., 2000a, b; McCullen and Towill, 2002; Dejonckheere et al., 2004; Ouyang, 2006; Li et al., 2006) and supply chain costs

International Journal of Physical Distribution & Logistics Management Vol. 37 No. 7, 2007 pp. 529-545 q Emerald Group Publishing Limited 0960-0035 DOI 10.1108/09600030710776464

IJPDLM 37,7

530

(Gavirneni et al., 1999; Gavirneni, 2002). However, in spite of these advantages, retailers, most of the time, do not desire to engage in information sharing. This is due to the fact that the primary beneciary from information sharing is the manufacturers, not the retailers (Lee et al., 2000; Yu et al., 2001, 2002; Zhao et al., 2002a, b). Vendor managed inventory (VMI), also known as continuous replenishment or supplier-managed inventory, is one of the most widely discussed partnering initiatives for encouraging collaboration and information sharing among trading partners (Angulo et al., 2004). Popularized in the late 1980s by Wal-Mart and Procter & Gamble (Waller et al., 1999), it was subsequently implemented by many other leading companies from different industries, such as Glaxosmithkline (Danese, 2004), Electrolux Italia (De Toni and Zamolo, 2005), Nestle and Tesco (Watson, 2005), Boeing and Alcoa (Micheau, 2005), etc. It is a supply chain initiative where the vendor decides on the appropriate inventory levels of each of the products and the appropriate inventory policies to maintain those levels. The retailer provides the vendor with access to its real-time inventory level. In this partnership program, the retailer may set certain service-level and/or shelf-space requirements, which are then taken into consideration by the vendor. That is, in a VMI system, the retailers role shifts from managing inventory to simply renting retailing space (Simchi-Levi et al., 2003, p. 154; Mishra and Raghunathan, 2004). VMI offers a competitive advantage for retailers because it results in higher product availability and service level as well as lower inventory monitoring and ordering cost (Waller et al., 1999; Achabal et al., 2000). For vendors, on the other hand, it results in reduced bullwhip effect (Lee et al., 1997b; Disney and Towill, 2003a, b) and better utilization of manufacturing capacity (Waller et al., 1999), as well as better synchronization of replenishment planning (Waller et al., 1999; Cetinkaya and Lee, 2000). While many benets have been identied in the literature, there are also a number of challenges that may exist in practice and that can potentially reduce the benets obtained from VMI or lead to failures in VMI programs. For instance, Spartan Stores, a grocery chain, shut down its VMI effort about one year after due in part to VMI vendors inability to deal with product promotions (Simchi-Levi et al., 2003, p. 161). Similarly, Kmart cut a substantial amount of VMI contracts because they were not satised with the forecasting ability of VMI vendors (Fiddis, 1997). Indeed, a few research studies have been carried out to investigate the underlying reasons of these failures (Aviv, 2002; Ovalle and Marquez, 2003; Angulo et al., 2004; Yao et al., 2007). Most of these research studies are limited to identifying the factors that play an important role in these failures. This may be because they only concentrate on the internal dynamics of VMI. Indeed, the following two factors may be extracted from these earlier studies as an explanation for failed VMI programs: (1) sharing of outdated or inaccurate sales and inventory data due in part to a lack of adequate information technology as well as a lack of mutual trust; and (2) inaccurate demand forecast owing to the fact that under VMI, retailers are excluded from the demand forecasting process. Our research differs from those that have been carried out in previous studies in one major aspect. Beside the internal dynamics of VMI, we have also considered the performance of an outside supplier on the success of VMI programs. For example, consider an outside supplier serving raw materials for VMI partners. The capabilities of this supplier, which is not participated into VMI, may have signicant impacts on the

benets gained from VMI. For example, the poor performance of the non-participating supplier may lead to a failure in VMI. This factor, to the best of our knowledge, is not explored in supply chain literature. Exploring the impact of non-participant upstream members on the performance of VMI is thus one major contribution of this research study. Moreover, when we investigate the relationship between the benets of information sharing and demand uncertainty, we see that the results of the previous research studies differ considerably. Some show that the value of information sharing is higher at higher levels of demand uncertainty (Lee et al., 2000; Zhang and Zhang, 2007), while others demonstrate that information sharing is more benecial at low levels of demand uncertainty (Waller et al., 1999, Gavirneni, 2002; Lau et al., 2004). The reason behind these differences may be differences in modelling assumptions, the types of supply chain environment studied, or cost structures used. However, in all cases, these differences among the contributions create confusion for SCM practitioners. In this study, we try to explore the impact of demand uncertainty on the performance improvements achieved by VMI. Providing a meaningful explanation of these differences in the literature is another important contribution of this research study. Unlike many prior analytical studies which have very restrictive assumptions for the sake of mathematical tractability (Mishra and Raghunathan, 2004; Lee and Chu, 2005; Yao et al., 2007), we have used a simulation model in this study to investigate the benets of VMI under more realistic circumstances. The simulation approach has been used extensively in the literature for analysing supply chain systems (Waller et al., 1999; Zhao et al., 2002a, b; Angulo et al., 2004; Lau et al., 2004; Zhang and Zhang, 2007). In this study, we considered a four-stage supply chain which consists of four echelons: a manufacturing plant, a warehouse, a distributor, and a retailer. The plant has limited manufacturing capacity and produces a single product. Each enterprise replenishes its inventory from its immediate upstream enterprise. The remainder of this study is organized as follows. The next section reviews the previous literature related to information sharing and VMI with emphasis on the benets of VMI. Section 3 claries the methodology and development of the simulation model. Setting of experimental design is identied in Section 4, followed by simulation results. Conclusions are in the nal section. 2. Related literature There have been a number of analytical and simulation studies that have extensively examined the factors affecting the benets derived from information sharing (Gavirneni et al., 1999; Lee et al., 2000; Yu et al., 2001, 2002; Gavirneni, 2002; Zhao et al., 2002a, b; Simchi-Levi and Zhao, 2003; Lau et al., 2004; Croson and Donohue, 2005; Zhang and Zhang, 2007). Most of these research studies suggest that information sharing yields signicant performance improvements for the supply chain. Moreover, the improvements realized in the performance of a supply chain enterprise through information sharing are also veried by many empirical studies (Zhou and Benton, 2007). Under certain conditions, however, the contribution of information sharing on supply chain performance is substantially higher. Information sharing is more benecial under conditions where downstream level information is shared rather than upstream level information (Lau et al., 2004; Croson and Donohue, 2005); and/or where inventory policies are reformulated to make better use of shared information

Exploring the benets of VMI

531

IJPDLM 37,7

532

(Gavirneni, 2002); and/or where higher levels of manufacturing capacity is available (Gavirneni et al., 1999; Gavirneni, 2002; Simchi-Levi and Zhao, 2003; Lau et al., 2004); and/or where supplier lead time is longer (Lee et al., 2000) and/or; where retailer lead time is shorter (Moinzadeh, 2002). However, when we investigate the relationship between the value of information and demand uncertainty, we see that the results of the previous research studies differ considerably. Some show that value of information sharing is higher at higher levels of demand uncertainty (Lee et al., 2000; Zhang and Zhang, 2007), while others demonstrate that information sharing is more benecial under low levels of demand uncertainty (Waller et al., 1999; Gavirneni, 2002; Lau et al., 2004). On the other hand, we also realize that the benets derived from information sharing are not distributed equally between suppliers and retailers. For example, analytical studies of Lee et al. (2000) and Yu et al. (2001, 2002) in a two stage supply chain environment, with one supplier and one retailer, revealed that retailers benets are insignicant when compared to the benets gained by suppliers. Similarly, Zhao et al. (2002a, b) conducted a simulation study of a supply chain consisting of one capacitated supplier and four retailers and found that retailers needed to be encouraged to participate in information sharing, due to the fact that the benets to retailers were insignicant from information sharing. These studies may explain why supply chain initiatives such as VMI have been popularized in industry as a tool for encouraging retailers to participate in information sharing. In order to better understand the underlying reasons behind the popularity of VMI, we may also need to consider the research carried out by Lee and Chu (2005). In their analytical study, Lee and Chu (2005) tried to identify the appropriate conditions under which VMI is desirable for both parties. Their results indicate that VMI is benecial for both parties if the stock level desired by the supplier at the retailer location is higher than that desired by the retailer. This means that suppliers with their desired stock level at the retailer location can determine the retailers decision about participating in VMI or not. At this point, the perceived higher cost of stockouts for suppliers compared to retailers (Achabal et al., 2000), may explain the popularity of VMI in different industries. Similarly, Yao et al. (2007) conducted an analytical study of a two-stage supply chain to identify the appropriate conditions under which implementing VMI is more desirable. Results from his analytical model indicate that the cost savings realized from VMI are likely to be higher as more and more reduction is achieved in ordering cost of retailers through VMI. Other researchers, on the other hand, investigated the performance of VMI under different conditions. Waller et al. (1999) carried out a simulation study in a two-stage supply chain to analyze the impact of VMI under various levels of demand variability, limited manufacturing capacity, and partial channel coordination. Their results indicate that VMI provides greater inventory reductions and a higher utilization of manufacturing capacity, mainly resulting from more frequent inventory reviews and shorter intervals between deliveries. Aviv (2002) constructed a model of two-level supply chain consisting of a retailer and a supplier. In his model, Aviv (2002) concentrated on the abilities of the supplier and the retailer to explain the uncertainty observed in customer demand. The analytical and subsequent numerical study of Aviv (2002) indicates that the success of VMI programs mainly depends on the relative forecasting skill of the supplier. That is, their results

indicate that the performance of the supply chain under VMI could be even worse than the traditionally managed supply chain unless the relative explanation power of the supplier is sufciently good. In a different research study, Ovalle and Marquez (2003) reached similar conclusions as Aviv (2002). They carried out a simulation study by using system dynamics modelling in order to evaluate the performance of various types of collaboration strategies. They presented the simulation results of a four-stage supply chain as described by Sterman (1989) in the article for the bear game. The results of their study indicate that one important problem of VMI programs is the elimination of retailers from inventory and forecasting decisions. Finally, Angulo et al. (2004) explored the impact of sharing inaccurate and delayed information on the performance improvements achieved by VMI in a four-stage supply chain both for stationary and non-stationary demand structures. Statistical analysis of the simulation outputs indicated that the performance of VMI substantially decreases if the shared retail information is not up-to-date. However, they also showed that when the inaccurate information is considered, unless it is highly inaccurate, does not create substantial reductions in the performance of VMI. 3. The supply chain simulation model At the initial stages of this research, we intended to use Microsoft Excel in constructing the simulation model; however, research conducted by Keeling and Pavur (2004) indicated that it might be possible for errors to occur in the random numbers generated by Microsoft Excel. Therefore, in order to eliminate this potential problem, we have used Crystal Ball, an Excel add-in published by Decisioneering. It is a popular risk analysis and forecasting program that uses Monte Carlo simulation in a spreadsheet environment. Two supply chain structures are considered in the model for comparison purposes. The rst structure is a supply chain operated under traditional ways of doing business (TSS) and the second structure is a supply chain model operated under a VMI program. In both systems, an (R, S) inventory control policy is used for replenishment decisions. Here, R indicates the review interval and S indicates the order-up-to level. Review interval (R) is chosen as one week. Order-up-to level, however, is updated at the beginning of each week to reect changes in demand patterns. The simplied simulation logic and the owchart are represented in Figure 1. Under TSS, all supply chain members strive to develop local strategies for optimizing their own organizations without considering the impact of these strategies on the performance of other supply chain members. Moreover, since there is no information sharing between members, upstream stages are unaware of actual demand information at the market place. Here, while creating demand forecasts and inventory plans, supply chain members use only replenishment orders placed by their immediate downstream member. Moreover, each supply chain member uses an installation stock concept (i.e. members consider only their own inventory level) in their production/inventory decisions. The sequence of events followed by a supply chain member under TSS is outlined as follows: . The member receives the delivery from its immediate upstream member, which was ordered L periods ago (the lead-time is L periods). If the member is the plant, L is the production lead time.

Exploring the benets of VMI

533

IJPDLM 37,7

START

Parameters of Demand Structure

Parameters of Capacity (CAP)

534

DEMAND GENERATION

DETERMINATION OF THE PLANT'S CAPACITY

Replenishment Orders

Point of Sales Data

DEMAND FORECAST (Exponential Smoothing)

TSS

SUPPLY CHAIN TYPE

VMI

DEMAND FORECAST (The Retailer and The Distribitor)

CALCULATE ORDER-UP-TO LEVEL

DETERMINE ORDER AMOUNT (The Retailer and The Distribitor) Inventory Level of Retailer Replenishment Orders

CALCULATE ORDER-UP-TO LEVEL (The Retailer and The Distribitor)

DEMAND FORECAST (The Warehouse and The Plant)

DETERMINE ORDER AMOUNT

DETERMINE ORDER AMOUNT (The Warehouse and The Distribitor) SATISFY INCOMING ORDER

REPORT

Average Supply Chain Inventory

Total Supply Chain Cost

Customer Service Level

Figure 1. Flow chart of the simulation model

END

The member observes the order placed by its immediate downstream member. If the member is the retailer, the order is the market demand. The member fulls the customer orders (plus backorders if there are any) by on-hand inventory, and any unfullled customer orders are backordered. The member analyzes the historical replenishment orders placed by its immediate downstream member for forecasting. Based on this demand forecast, the member updates its order-up-to point. If the member is the retailer, historical market demand data is analyzed. The order-up-to point of the member at stage k, S k ; estimated from the observed demand is as follows (Nahmias, 1997, p. 278): Sk F 21 k  bk bk hk  1

Exploring the benets of VMI

535

where F k : is the distribution function of the demand realized by the member at stage k. Similarly, bk and hk are back-order and holding costs of the member at stage k, respectively. Here, parameters of the demand distribution, F k :; are updated at the beginning of each week by using the exponential smoothing method (Nahmias, 1997, p. 74) to reect changes in demand patterns. The member decides how many units to order from its immediate upstream member. The quantity of the order is equal to the difference between the order-up-to level and inventory position. If the member is the plant, a production order is placed. Here, the plant, because of its limited manufacturing capacity, cannot always produce enough to bring its inventory position up to the updated value of S. In these cases, the plant makes full capacity production by backordering the remaining requirement. This modication of order-up-to policy for the case of limited production capacity provides an optimal solution for uncertain demands (Gavirneni et al., 1999; Federgruen and Zipkin, 1986a, b).

Under VMI, on the other hand, the retailer provides the distributor with access to its real-time inventory level as well as its point-of-sales (pos) data. In return, the distributor takes the responsibility of managing the inventories of the retailer. That is, under VMI, the distributor does not only need to take its own inventories into account while making inventory plans, but also the inventories of the retailer. Therefore, under this structure, the distributor uses the echelon stock concept (Clark and Scarf, 1960; Axsater and Rosling, 1993) in replenishment planning. All other echelons of the supply chain (the plant and the warehouse), on the other hand, operated in the same way as in TSS. In order to compute the echelon order-up-to levels of the retailer and the distributor, the heuristic developed by Shang and Song (2003) is used. Again, in this supply chain, the exponential smoothing method is used to update the order-up-to level at each week. Although the literature suggests that VMI may result in signicant reductions in retailers lead times (Waller et al., 1999), we do not include this reduction in our simulation model. Instead, we assume that the retailers lead times remain constant after VMI implementation. Indeed, here we have concentrated on information sharing and centralized decision-making aspects of VMI only. Therefore, we may surely state that the results obtained from this study constitute a lower bound for the benets obtained from VMI.

IJPDLM 37,7

The cost structures for the supply chain members in the simulation model are assumed to be as follows; the unit back-order costs per week for the plant, the warehouse, the distributor, and the retailer are $5, 11, 18, and 25, respectively. The unit inventory costs per week for the plant, the warehouse, the distributor, and the retailer are $0.25, 0.50, 0.75, and 1.00, respectively. 3.1 Retailers demand structure Although normal distribution is more widely used in supply chain research studies, the g distribution is used here to represent the customer demand realized by the retailer. This is due to the fact that there are some limitations in the normal distribution in representing demand structures. For example, normal distribution allows the occurrence of negative customer demand. Therefore, in order to avoid this unrealistic situation, some restrictive assumptions have to be included in the model (Waller et al., 1999; Zhao et al. 2002a, b; Lau et al., 2004). The g distribution, on the other hand, does not have such problems because it allows only non-negative values. Moreover, the g distribution is exible in that it can represent a wide variety of demand structures. Keaton (1995), for instance, states that choosing g distribution is an effective choice to represent the demand patterns. There are two parameters of the g distribution. These are shape a and scale b parameters. The mean and the variance of the distribution can be expressed as ab and ab 2 ;, respectively. In the simulation model, we assume that the shape parameter of the demand distribution is 15 a 15: The scale parameter b; on the other hand, is assumed to be a stochastic variable in the form of equation (1).   2p t 2 bt 20 season sin 52 In equation (2), bt is the scale parameter of the g distribution in week t. The variability in the scale parameter of the demand distribution allows us to generate both seasonal and non-seasonal customer demands. For example, while assigning zero to the season constant produces non-seasonal demand pattern, assigning non-zero values results in seasonality in customer demand. A representative histogram of the market demand for the selected parameters is generated in Figure 2 to clarify the distribution of the market demand to the readers. Three demand structures representing different combinations of seasonality are used in this study. These are customer demand with no seasonality (SDV), customer demand with medium level of seasonality (MDV), and customer demand with high level of seasonality (HDV). The values of the season constant for each demand structure are determined as 0, 2, and, 4, respectively. The values of the season constant
0.25

536

Probability

0.2 0.15 0.1 0.05 117 157 197 237 277 317 357 397 437 477 517 557 597 0

Figure 2. Histogram of the market demand when a 15 and bt 20

Market Demand

are selected in such a way that both non-seasonal and seasonal customer demands with different strengths are generated. For example, while SDV represent the non-seasonal customer demand, MDV and HDV represent the demand structures with seasonal swings of the size of approximately 10 and 20 percent of average demand, respectively. 3.2 Verication and validation of the simulation In order to verify that the simulation program performs as intended, the conceptual model is divided into three parts: demand generation and determination of total manufacturing capacity, forecasting and production/inventory planning, and order fulllment and reporting. Each part is designed separately so that more efcient and effective debugging is made possible. Moreover, the combined simulation model is also traced and tested with the results calculated manually. Later, in order to validate the simulation output, the random demand variables generated in the simulation model are plotted on a scatter diagram. Then, it is validated that the intended demand structure is generated. The supply chain model above was simulated for 1,128 weeks. The initial parameters of the forecasting model were estimated with the rst 400 weeks of simulation run, which were removed later from the output analysis to eliminate the worm-up period effect. Therefore, the rest of the data was used for effective simulation output analysis. In order to reduce the impact of random variations, the same random numbers were used to simulate both systems. That is, same customer demand was generated for both types of supply chain system. In addition to this variance reduction technique, 15 replications for each combination of the independent variables were conducted. 4. Experimental design Four independent factors are considered in the experimental design. These are; type of supply chain (SCTYPE), available production capacity of plant (CAP), uncertainty in customer demand (DV), and replenishment lead times (L). The number of levels of these variables and their values are listed in Table I. Factor SCTYPE refers to the way the supply chain is operated. Specically, this factor indicates whether the supply chain is operated under TSS or VMI. The factor CAP is expressed as the ratio of the plants total capacity to the total market demand. Total capacity of the plant is distributed to each week, equally. The factor L denotes the replenishment lead times between each member of the supply chain. Finally, the factor DV indicates the level of uncertainty seen in market demand. Three kinds of factors are used as dependent variables in the experimental design in order to evaluate benets gained from VMI. These factors are average inventory level in the supply chain (INV), total cost for the entire supply chain (TSC) and customer service level of the retailer (CSL). TSC is the sum of the inventory holding costs of all members in
Levels 3 HDV 7 1.30

Exploring the benets of VMI

537

Independent factors SCTYPE DV L CAP

1 TSS SDV 1 1.10

2 VMI MDV 4 1.20

5 Table I. Independent factors of the experimental design

1.40

1.50

IJPDLM 37,7

the supply chain and backorder cost of the retailer. Here, we include the back order cost of the retailer only, because all other back order costs are internal costs within the entire supply chain and they are not actually incurred. Factor CSL is the percentage of customer demand satised by the retailer through the available inventory. 5. Simulation output analysis The output from the simulation experiments is analyzed using the MANOVA procedure of the SPSS. The MANOVA procedure is more appropriate for our model because it considers the correlation between the dependent variables in the experimental design. For more detailed information about MANOVA see the paper by Hair et al. (1998, p. 331). Selected MANOVA results are presented in Table II. MANOVA results in Table II show that at 5 percent signicance level, SCTYPE has signicant impacts on all three-performance factors, which indicates that VMI has substantial inuences on the performance of the supply chain. The performance of each type of supply chain is presented in Table III.
Dependent variables INV F-value Pr . F 3,818.611 1,371.075 39,476.629 6.572 21.995 248.518 157.211 18.197 33.560 58.013 0.110 47.666 2.325 4.107 1.021 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.9989 0.0000 0.0988 0.0000 0.3450

538

CSL Source SCTYPE CAP L DV SCTYPE *CAP SCTYPE *L CAP *L SCTYPE *CAP *L SCTYPE *DV CAP *DV SCTYPE *CAP *DV L *DV SCTYPE *L *DV CAP *L *DV SCTYPE *CAP *L *DV F-value 85.379 26.536 444.785 1,130.268 10.854 28.735 7.782 0.245 20.900 8.321 0.604 85.479 2.218 3.293 0.316 Pr . F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.9820 0.0000 0.0000 0.7746 0.0000 0.0287 0.0009 0.9953

TSC (a) F-value Pr . F 127.424 3.455 4,090.439 570.928 4.727 1.148 5.366 1.123 40.252 4.413 0.846 72.943 1.223 5.294 0.260 0.0000 0.0000 0.0000 0.0000 0.0000 0.3180 0.0001 0.2821 0.0000 0.0000 0.5622 0.0000 0.2821 0.0000 0.9985

Table II. Selected MANOVA results

Note: aBased on residual analysis, log transformation of TSC was made to satisfy the assumptions of ANOVA

Performance measures CSL (percentage) TSC ($) Table III. Performance of each type of supply chain INV (unit)

SCTYPE TSS VMI TSS VMI TSS VMI

Average 94.52 96.08 701,761 577,899 1,956 1694

95 percent condence interval Lower bound Upper bound 94.35 95.95 685,510 547,198 1,949 1686 94.70 96.21 718,013 608,189 1,963 1701

Examination of Table III reveals that compared to TSS, the reduction in average inventory level in the supply chain is 13.4 percent. Similarly, the reduction in total supply chain cost varies from 6.5 to 43.3 percent with an average around 17.6 percent. The results also indicate that VMI signicantly increases the retailers service level. For example, while the customer service level under TSS is 94.3 percent on the average, it is around 96 per cent under VMI. Therefore, these results lead us to conclude that VMI produces substantial increases in the retailers customer service level while decreasing average inventory level and the total cost of the entire supply chain. Actually, these ndings are simple and intuitively expected for us, so we will not concentrate on them further. Instead, we will concentrate on how a non-participating members manufacturing capacity (CAP) affects the benets of VMI as well as the uncertainty in customer demand (DV) and lead times (L). For this purpose, performance of VMI under various capacity levels (CAP), demand uncertainty (DV), and lead times (L) are produced in Figure 3. 5.1 Capacity of non-participated member MANOVA results in Table II show that at 5 percent signicance level, the interaction effect between CAP and SCTYPE has signicant impacts on all three dependent factors. This means that capacity level of non-participated member has a signicant inuence on the performance of VMI for all three performance measures. Examination of Figure 3 shows that the reduction amount in total supply chain cost is signicantly higher at larger levels of capacity at the plant. For example, the reduction amount in total supply chain cost increases from 14.6 to 21 percent, on the average when capacity ratio (CAP) increases from 1.10 to 1.50. In addition, when we consider the average inventory level of the supply chain we see a very similar situation with supply chain cost. Finally, when the customer service level is considered, we see that under all levels of manufacturing capacity, VMI produces higher ll rates than TSS. Therefore, these ndings show us that VMI produces substantially higher benets at higher levels of the manufacturing capacity. This result is very interesting.
24 20 16 12 8 4 0 1 25 20 15 10 5 0 1.10 1.20 1.30 CAP 1.40 1.50 4 L 7 98 97 96 95 94 93 92 24 20 16 12 8 4 0 SDV MDV DV HDV 98 97 96 95 94 93 92 Reduction in TSC and INV (%) Reduction in TSC and INV (%)

Exploring the benets of VMI

539

CSL (%)

97 96 95 94 93 92

Reduction in TSC and INV(%)

CSL (%)

TSC INV CSL of TSS CSL of VMI

CSL (%)

Figure 3. Overall performance of VMI

IJPDLM 37,7

540

That is, in this supply chain, the plant does not participate in the VMI program, but its manufacturing capacity has a signicant inuence on the performance of VMI. The result obtained here is consistent with previous research studies (Gavirneni et al., 1999; Lee et al., 2000; Gavirneni, 2002; Simchi-Levi and Zhao, 2003; Lau et al., 2004). That is, in these earlier studies, researchers suggest that benets obtained from information sharing substantially decrease with respect to in-house capacity restriction. Our research, however, extended the results obtained from these earlier studies by indicating that capacity restriction of non-participated suppliers also substantially reduces the benets derived from VMI. 5.2 Uncertainty in customer demand MANOVA results in Table II show that at 5 percent signicance level, the interaction effect between DV and SCTYPE has a signicant impact on all performance factors. This indicates that uncertainty in customer demand has a signicant inuence on the performance of VMI for all three performance measures. Examination of Figure 3 shows that the reduction in total supply chain cost and the average inventory level achieved by the VMI program reduce as the level of demand uncertainty increases. For example, we see from Figure 3 that when there is no seasonality in customer demand, the reduction in total supply chain cost and average inventory level are realized as 24 and 16 percent, respectively. However, when there is a high-degree of seasonality, these savings decreased to 15 percent for the total supply chain cost and 9.3 percent for the average inventory level. Moreover, when we consider the other performance criteria, we see that under all levels of demand uncertainty, VMI produces higher level of service. The results obtained here show us that the uncertainty in customer demand substantially reduces the benets obtained from VMI, as also indicated in some earlier studies (Waller et al., 1999; Gavirneni, 2002; Lau et al., 2004). In contrast with these results, the ndings of other researchers (Lee et al., 2000; Zhang and Zhang, 2007) show that information sharing is more benecial at higher levels of demand uncertainty. The underlying reasons for this apparent contradiction between these studies may be differences in modelling assumptions, the supply chain environment studied, or cost structures used. However, among these reasons the most inuential ones are related to the type of information sharing strategy implemented. While some types of information sharing strategies involve almost all relevant descriptions about uncertainty of probable future demand that exists in the system, some others convey only partial information to help supply chain partners in resolving uncertainty. Namely, answers of the following two questions play a critical role in determining the value of information sharing under different levels of demand uncertainty. These are: (1) Which types of information are shared among supply chain partners? (2) Under what conditions are these information shared? Here, let us consider two information sharing strategies. In the rst strategy, information of a retailer about an upcoming promotion is fully transferred to the upstream members along with sales and inventory data. In the second strategy, however, only inventory and sales data are shared. Of course, the benets obtainable from these two strategies under different levels of demand uncertainty vary. That is, the benets obtained from the former strategy, under the market conditions where

much of the uncertainty stems from product promotions, are substantially higher. On the contrary, the value of the second strategy substantially decreases as uncertainty comes from product promotions increases. As a result, it can be said that information sharing is more benecial at higher levels of demand uncertainty as long as the type of information shared can help supply chain members in resolving the uncertainty observed in customer demand. Otherwise, higher levels of demand uncertainty will denitely reduce the benets obtained from information sharing as has also been seen in this research. 5.3 Lead times MANOVA results in Table II show that at 5 percent signicance level, the interaction effect between L and SCTYPE has a signicant impact on the customer service level and supply chain inventory, but it does not have any signicant impact on the total supply chain cost. This means that while reduction in average inventory level and increase in customer service level realized from VMI change with lead times, reduction of supply chain cost gained from VMI does not change with lead times. Examination of Figure 3 shows that the inventory reduction obtained from VMI is higher at shorter lead times, but by very small amounts. For example, while the inventory reduction is 13.90 percent at a lead time of one week, it is 13.10 percent at a longer lead time of seven weeks. Similarly, we also see that VMI provides a little higher performance in ll rates at shorter lead times. That is, while it produces 1.95-points higher ll rate than TSS when the lead times are one week, it provides 1.14-points higher ll rate when the lead times are seven weeks. However, these slight performance improvements in service level and average inventory level are not reected in total supply chain cost as can be seen in Table II and Figure 3. Therefore, from these results, we can conclude that regardless of the lead time horizon, VMI produces nearly the same level of performance increase. Current literature on information sharing indicates that performance increase achieved by information sharing under longer supplier lead time is quite substantial and it increases with respect to the supplier lead time (Lee et al., 2000). Similarly, the literature also suggests that performance increase achieved by information sharing is lower under longer retailer lead time (Moinzadeh, 2002). This paper extends the current literature and provides valuable information for SCM practitioners by indicating that VMI or information sharing provides nearly the same level of performance increase in supply chains where lead times are short or long, as long as the ratio of the retailers lead time to that of the suppliers remains constant. 6. Conclusions This paper explores the benets of VMI in supply chains involving four stages by providing important insights into the performance of VMI under various supply chain scenarios. One major contribution of this research is to explore the inuence of capacity restriction of an outside supplier on the performance of VMI. Moreover, the performance increase achieved by VMI under various supply chain scenarios, which are characterized by demand uncertainty and lead times, are also explored extensively. Through comprehensive simulation experiment and subsequent statistical analysis of the simulation outputs, we make the following important ndings:

Exploring the benets of VMI

541

IJPDLM 37,7

542

(1) We examine that there are substantial decreases in the performance increase obtained from VMI as the uncertainty in customer demand increases, unless the shared information provides additional information to the vendor to resolve uncertainty in customer demand. (2) We examine that the performance of outside supplier plays an important role for the success of VMI programs. Our simulation study indicates that as the manufacturing capacity of the outside supplier decreases, the benets obtained from VMI also decreases substantially. This nding, interestingly, shows that success of VMI programs does not depend only on internal dynamics of VMI, but also on external factors. In addition, we also realize that capacity restriction of a non-participated supplier is more crucial when the demand uncertainty in customer demand is high. This is because the higher demand uncertainty in customer demand also reduces the benets derived from VMI. (3) Finally, we examine the impact of lead times on the performance of VMI. We observe that, regardless of lead time horizons, VMI provides nearly the same level of performance increase as long as the ratio of the retailers lead time to that of the suppliers remains constant. In conclusion, managers in a supply chain enterprise have to make careful benet/cost analysis while making their decisions regarding VMI implementation under conditions where the performance of upstream members is poor and/or uncertainty in customer demand is high. Although this study provides some insights into VMI and its relationship with outside suppliers as well as demand uncertainty, there are some limitations that we have to state. First, in the simulation model, we assumed that the retailers lead time remains constant after VMI implementation; however, VMI may, in fact, signicantly reduce the retailers lead time. Second, we considered a serial supply chain structure with one member at each echelon. This is only a simplied case and in future researches, modelling more realistic supply chain structures will better explain the results obtained from this research. Third, we assume that the members in the supply chain use modied or simple order-up to policy to make their inventory decisions; however, there are other inventory or production policies that can be included in the model. Forth, the cost structure used in the model only represents one special case.
References Achabal, D.D., McIntyre, S.H., Smith, S.A. and Kalyanam, K. (2000), A decision support system for vendor managed inventory, Journal of Retailing, Vol. 76 No. 4, pp. 430-54. Angulo, A., Nachtmann, H. and Waller, M.A. (2004), Supply chain information sharing in a vendor managed inventory partnership, Journal of Business Logistics, Vol. 25 No. 1, pp. 101-20. Aviv, Y. (2002), Gaining benets from joint forecasting and replenishment process: the case of auto-correlated demand, Manufacturing & Service Operations Management, Vol. 4 No. 1, pp. 55-74. Axsater, S. and Rosling, K. (1993), Notes: installation vs echelon stock policies for multilevel inventory control, Management Science, Vol. 39 No. 10, pp. 1274-80. Cetinkaya, S. and Lee, C.Y. (2000), Stock replenishment and shipment for vendor-managed inventory systems, Management Science, Vol. 46 No. 2, pp. 217-32.

Chen, F., Drezner, Z., Ryan, J.K. and Simchi-Levi, D. (2000a), Quantifying the bullwhip effect in a simple supply chain: the impact of forecasting, lead times and information, Management Science, Vol. 46 No. 3, pp. 436-43. Chen, F., Drezner, Z., Ryan, J.K. and Simchi-Levi, D. (2000b), The impact of exponential smoothing forecasts on the bullwhip effect, Naval Research Logistics, Vol. 47 No. 4, pp. 269-86. Clark, A.J. and Scarf, H. (1960), Optimal policies for a multi-echelon inventory problem, Management Science, Vol. 6 No. 4, pp. 475-90. Croson, R. and Donohue, K. (2005), Upstream versus downstream information and its impact on the bullwhip effect, System Dynamics Review, Vol. 21 No. 3, pp. 249-60. Danese, P. (2004), Beyond vendor managed inventory: the Glaxosmithkline case, Supply Chain Forum: An International Journal, Vol. 5 No. 2, pp. 32-9. De Toni, A.F. and Zamolo, E. (2005), From a traditional replenishment system to vendor-managed inventory: a case study from the household electrical appliances sector, International Journal of Production Economics, Vol. 96 No. 1, pp. 63-79. Dejonckheere, J., Disney, S.M., Lambrecht, M.R. and Towill, D.R. (2004), The impact of information enrichment on the bullwhip effect in supply chains: a control engineering perspective, European Journal of Operational Research, Vol. 153 No. 3, pp. 727-50. Disney, S.M. and Towill, D.R. (2003a), Vendor-managed inventory (VMI) and bullwhip reduction in a two level supply chain, International Journal of Operations & Production Management, Vol. 23 No. 6, pp. 625-51. Disney, S.M. and Towill, D.R. (2003b), The effect of vendor managed inventory dynamics on the bullwhip effect in supply chains, International Journal of Production Economics, Vol. 85 No. 2, pp. 199-215. Federgruen, A. and Zipkin, P. (1986a), An inventory model with limited production capacity and uncertain demands I: the average-cost criterion, Mathematics of Operations Research, Vol. 11 No. 2, pp. 193-207. Federgruen, A. and Zipkin, P. (1986b), An inventory model with limited production capacity and uncertain demands II: the discounted-cost criterion, Mathematics of Operations Research, Vol. 11 No. 2, pp. 208-15. Fiddis, C. (1997), Manufacturer-retailer Relationships in the Food and Drink Industry: Strategies and Tactics in the Battle for Power, Financial Times Retail & Customer Publications, London. Gavirneni, S. (2002), Information ows in capacitated supply chains with xed ordering costs, Management Science, Vol. 48 No. 5, pp. 644-51. Gavirneni, S., Kapuscinski, R. and Tayur, S. (1999), Value of information in capacitated supply chains, Management Science, Vol. 45 No. 1, pp. 16-24. Hair, J., Anderson, R.E., Tatham, R.L. and Black, W.C. (1998), Multivariate Data Analysis, 5th ed., Prentice-Hall, Englewood Cliffs, NJ. Keaton, M. (1995), Inventory control under gamma demand and stochastic lead time, Journal of Business Logistics, Vol. 16, pp. 107-31. Keeling, K.B. and Pavur, R.C. (2004), Numerical accuracy issues in using excel for simulation studies, Proceedings of the 2004 Winter Simulation Conference, pp. 1513-8. Lau, J.S.K., Huang, G.Q. and Mak, K.L. (2004), Impact of information sharing on inventory replenishment in divergent supply chains, International Journal of Production Research, Vol. 42 No. 5, pp. 919-41.

Exploring the benets of VMI

543

IJPDLM 37,7

544

Lee, C.C. and Chu, W.H.J. (2005), Who should control inventory in a supply chain, European Journal of Operational Research, Vol. 164 No. 1, pp. 158-72. Lee, H., Padmanabhan, V. and Whang, S. (1997a), Information distortion in a supply chain: the bullwhip effect, Management Science, Vol. 43 No. 4, pp. 546-58. Lee, H., Padmanabhan, V. and Whang, S. (1997b), The bullwhip effect in supply chains, Sloan Management Review, Vol. 38 No. 3, pp. 93-102. Lee, H., So, K.C. and Tang, C.S. (2000), The value of information sharing in a two-level supply chain, Management Science, Vol. 46 No. 5, pp. 626-64. Li, J., Sikora, R., Shaw, M.J. and Tan, G.W. (2006), A strategic analysis of inter organizational information sharing, Decision Support Systems, Vol. 42 No. 1, pp. 251-66. McCullen, P. and Towill, D.R. (2002), Diagnosis and reduction of bullwhip in supply chains, Supply Chain Management: An International Journal, Vol. 7 No. 3, pp. 164-79. Micheau, V.A. (2005), How Boeing and Alcoa implemented a successful vendor managed inventory program, Journal of Business Forecasting, Vol. 24 No. 1, pp. 17-19. Mishra, B.K. and Raghunathan, S. (2004), Retailer- vs vendor-managed inventory and brand competition, Management Science, Vol. 50 No. 4, pp. 445-57. Moinzadeh, K. (2002), A multi-echelon inventory system with information exchange, Management Science, Vol. 48 No. 3, pp. 414-26. Nahmias, S. (1997), Production and Operations Analysis, Irwin/McGraw-Hill, Homewood, IL. Ovalle, O.R. and Marquez, A.C. (2003), The effectiveness of using e-collaboration tools in the supply chain: an assessment study with system dynamics, Journal of Purchasing & Supply Management, Vol. 9 No. 4, pp. 151-63. Ouyang, Y. (2006), The effect of information sharing on supply chain stability and the bullwhip effect, European Journal of Operational Research, Vol. 182 No. 3, pp. 1107-21. Shang, K.H. and Song, J.S. (2003), Newsvendor bounds and heuristic for optimal policies in serial supply chains, Management Science, Vol. 49 No. 5, pp. 618-38. Simchi-Levi, D. and Zhao, Y. (2003), The value of information sharing in a two stage supply chain with production capacity constraints, Naval Research Logistics, Vol. 50 No. 8, pp. 888-916. Simchi-Levi, D., Kaminsky, P. and Simchi-Levi, E. (2003), Designing and Managing the Supply Chain, 2nd ed., McGraw-Hill, Homewood, IL. Sterman, J. (1989), Modeling managerial behavior: misperceptions of feedback in a dynamic decision making experiment, Management Science, Vol. 35 No. 3, pp. 321-39. Swaminathan, J.M. and Tayur, S.R. (2003), Models for supply chain in e-business, Management Science, Vol. 49 No. 10, pp. 1387-406. Waller, M.A., Johnson, M.E. and Davis, T. (1999), Vendor-managed inventory in the retail supply chain, Journal of Business Logistics, Vol. 20 No. 1, pp. 183-203. Watson, E. (2005), Nestle switches to vendor managed inventory with Tesco, Food Manufacture, Vol. 80 No. 8, p. 20. Yao, Y., Evers, P.T. and Dresner, M.E. (2007), Supply chain integration in vendor-managed inventory, Decision Support Systems, Vol. 43 No. 2, pp. 663-74. Yu, Z., Yan, H. and Cheng, T.C.E. (2001), Benets of information sharing with supply chain partnerships, Industrial Management & Data Systems, Vol. 101 No. 3, pp. 114-9. Yu, Z., Yan, H. and Cheng, T.C.E. (2002), Modelling the benets of information sharing-based partnerships in a two-level supply chain, Journal of Operational Research Society, Vol. 53 No. 4, pp. 436-46.

Zhang, C. and Zhang, C. (2007), Design and simulation of demand information sharing in a supply chain, Simulation Modelling Practice and Theory, Vol. 15 No. 1, pp. 32-46. Zhao, X., Xie, J. and Zhang, W.J. (2002a), The impact of information sharing and ordering co-ordination on supply chain performance, Supply Chain Management: An International Journal, Vol. 7 No. 1, pp. 24-40. Zhao, X., Xie, J. and Leung, J. (2002b), The impact of forecasting model selection on the value of information sharing in a supply chain, European Journal of Operational Research, Vol. 142 No. 2, pp. 321-44. Zhou, H. and Benton, W.C. Jr (2007), Supply chain practice and information sharing, Journal of Operations Management. Further reading Mason-Jones, R. and Towill, D.R. (2000), Coping with uncertainty: reducing bullwhip behaviour in global supply chains, Supply Chain Forum: An International Journal, Vol. 1 No. 1, pp. 40-5. Corresponding author Kazim Sari can be contacted at: kazims@beykent.edu.tr

Exploring the benets of VMI

545

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.