Vendor-managed inventory in fresh-food supply chains
Post, Roel
DOI:
10.33612/diss.130028783
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Post, R. (2020). Vendor-managed inventory in fresh-food supply chains. University of Groningen, SOM research school. https://doi.org/10.33612/diss.130028783
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Chapter 5
Conclusion and discussion
VMI is a far-reaching concept that, when introduced in a supply chain, affects many processes and performance indicators. The change in supply chain governance that comes along with VMI alters the decision latitude of both buyers (retailers) and sup-pliers, which also has consequences for their relationship. Furthermore, VMI brings along new opportunities to share information and use this information for planning production and transportation, but also places new demands on the abilities of or-ganizations to do so. This broad range of possibilities to use VMI is even more in-teresting if we consider that each supply chain is unique in terms of its context (i.e., the characteristics of the organizations and products involved). Even within a sin-gle supply chain a broad range of organizations and products can be found, and their characteristics are likely influencing how VMI can and should be adopted. The main purpose of this thesis is to explore how suppliers can adopt VMI and how the context in which they operate affects their decisions and abilities to make VMI a suc-cess. With this purpose in mind, we collected and studied empirical data of a VMI introduction by a large retailer and its suppliers from a different perspective in each chapter of this thesis.
5.1
Main findings and contributions
Over the last 30 years, more than 200 studies that use analytical models to study VMI have been published (Kauremaa et al., 2009; Marqu`es et al., 2010; Govindan,
2013). This thesis builds on this rich body of literature to distill the main compo-nents that can be part of VMI and describes how these compocompo-nents can potentially improve performance of a supply chain. First, an information sharing component that focuses on reducing uncertainty in the replenishment process. Second, a deci-sion transfer component that provides the supplier more decideci-sion freedom by cen-tralizing decisions. Third, an inventory position transfer that reduces the need for buffers by locating inventory closer to demand. And finally, an ownership transfer component, that incentivizes the supplier to act in best interest of the supply chain by aligning decisions with ownership.
The number of studies that use empirical methods to study VMI is far more lim-ited than the number of model-based studies. Only about 30 papers have been iden-tified. Laying our conceptualization of VMI – with its distinct components – over the current empirical research on VMI reveals that empirical understanding of how suppliers can use VMI is still limited. Important details on specific VMI configu-rations are often lost or remain implicit in empirical work, which leaves it unclear what component of VMI was responsible for the observed changes in supply chain performance. For example, was it the additional information shared or the transfer of inventory control that enabled a supplier to improve its service levels?
Furthermore, only a small selection of possible VMI configurations have been studied empirically. Current work that measures how VMI affects service levels and inventory levels in supply chains almost exclusively focuses on large suppliers driv-ing VMI implementations for their buyers where the ownership of the inventories remained with the buyers. Such context naturally leaves the effect of supplier and product characteristics out of scope. Therefore, it remains unclear how suppliers can use the VMI components together and how this is affected by their product and process characteristics.
This thesis addresses this gap in the empirical literature by studying VMI from a new perspective: we consider multiple suppliers that supply products that markedly differ in their characteristics, such as shelf life, volume and profit margins. These dif-ferences also have implications on the production processes and financial motives that drive suppliers’ replenishment decisions. Furthermore, we consider a VMI in-troduction that not only includes a transfer of the control of inventory to suppli-ers, but also a transfer of the ownership of inventory. This combination provides suppliers the freedom to manage their inventory levels without any constraints (or
Conclusion and discussion 111
promises) to satisfy the buyer, which allows a wider range of outcomes of introduc-ing VMI. The methods applied in the thesis also differ from those applied in prior VMI work. First, a new measure for inventory is developed that measures all in-ventory in the supply chain based on the shelf life of the products. This measure provides a unique opportunity to observe how VMI affects the inventory in the sup-ply chain, regardless of where it is located. Measuring all inventory in a supsup-ply channel allows us to focus on a supply chain cost perspective, rather than costs in-curred at specific locations in the supply chain. Second, the thesis explicitly includes both sides of the supplier-retailer dyad in its analyses, which also provides a more complete view on the impact of VMI at a supply chain level. Third, the thesis adopts longitudinal measures that enable observing how the organizations and their perfor-mance evolved up to the introduction of VMI and the three years after.
The different perspectives and methods applied in this thesis lead to results that can be considered surprising, not only in relation to prior studies in the academic literature, but also to the organizations involved in the VMI introductions. Chap-ter 2 shows that suppliers on average increase inventories when faced with VMI, which is in sharp contrasts with prior scientific papers and industry reports. On the other hand, after a learning effect (that lasted up to a year), suppliers were able to achieve significantly higher service levels than before the introduction of VMI. Such an improvement is especially remarkable when considering that service levels were already very high in the supply chain under study. The results of this chapter show how the view on VMI in current empirical research is to some extent biased by a very similar empirical context in which those studies took places. That is, empirical research prior to this thesis studied a context where a large and powerful supplier introduced VMI for its buyers, without transferring ownership from the buyer to this initiating supplier. In such a setting this supplier tries to take advantage of cen-tralized production, transportation and inventory planning for its multiple buyers, at the promise of lower inventories and higher service levels at those buyer’s in-ventory positions. In a setting where a retailer initiates VMI for its suppliers, these suppliers do not necessary have multiple buyers (that use VMI). However, VMI does allow the supplier to realize economies of scale through larger production and trans-portation batches at the cost of additional inventory, in particular when inventory is not restricted (because it is owned by the supplier).
processes under VMI and how these changes affect long term performance, we iden-tify the drivers behind the performance changes observed in Chapter 2. On the one hand, the results presented in Chapter 3 show how VMI enables suppliers to more effectively match supply and demand without increasing inventory in the supply chain, which is in line with the dominant view on VMI in empirical work. On the other hand, Chapter 3 also reveals how VMI can motivate suppliers to increase in-ventory levels to improve service levels and enhance economies of scale. In particu-lar, when inventory is not limited by high demand volumes and the perishability of products, suppliers prefer higher inventories to release the strain on their production and transportation planning and enhance economies of scale. The results indicate that suppliers also manage to improve service levels without increasing inventory, especially for products produced in short production cycles because of their perisha-bility or products that are exposed to high demand variaperisha-bility. The latter suggests that the higher potential VMI holds to improve challenging replenishment processes outweighs the higher investment and abilities required to implement VMI in such replenishment processes, which has been a topic of discussion in the academic liter-ature.
The longitudinal case study of eight supplier-retailer dyads presented in Chapter 4 deepens our understanding of the performance effects in the transition to VMI that were observed in Chapter 2 and the different ways in which suppliers use VMI in the long term observed in Chapter 3. Both a supplier’s ability to coordinate the replen-ishment processes and its willingness to cooperate with the retailer affect their per-formance after the introduction of VMI. Moreover, we discern different patterns that show how this ability to coordinate and efforts to cooperate evolve over time, and how these abilities and efforts together help explain changes in performance. If sup-pliers lack coordination abilities at the start of the VMI introduction they are unable to understand how their replenishment process can benefit from VMI (or not), which prevents them from setting goals that match this potential. This can lead to underes-timating the investments required to manage the replenishment processes involved with challenging products, such as highly perishable products. However, because VMI holds a large potential to improve supply chain performance for these prod-ucts, we observe that suppliers can recover in the long term by increasing their co-operation efforts and develop their coordination abilities. This observation explains the large performance dips and the long learning curves after VMI introduction, but
Conclusion and discussion 113
also indicates the large potential of VMI in the long run. On the other hand, for prod-ucts with less challenging replenishment processes we observe overestimation of the VMI performance gains, and therefore over-investment. Such an over-investment can form a risk on the long-term relation between supplier and retailer. For these products suppliers gain mostly from increased inventories that enable economies of scale, rather than costly investments in increased integration of production and transport with the demand information. If such investments are done anyway, this can lead to disappointment when the expected performance improvements do not materialize. Along the more complex dynamic patterns that are observed in this study, Chapter 4 reveals how short-term performance effects after a VMI introduc-tion are not necessarily representative for long term performance effects, but does have predictive value when interpreted in a broader context.
In general, we can conclude that actions of the decision makers we observe in our data do not contradict the modeling assumptions of the analytical studies de-scribed in the the introduction of this thesis. There is at least one work among the diverse studies (Marqu`es et al., 2010) that can explain each observation. However, to find which stream of analytical work is applicable to a given empirical context requires relating the model assumptions to this empirical context, which is often not trivial. The contributions of this thesis can provide decision makers with knowledge of which assumptions best apply to their situation, thus, what models can be used to estimate the outcomes of introducing VMI in this context.
VMI is of course not the only way to centralize responsibilities in the supply chain. Instead of transferring decisions to the supplier, many retailers have moved in the opposite direction: transferring responsibilities from the supplier to their own centralized planning departments. An example of this is factory gate pricing (le Blanc et al., 2006), where the retailer is in charge of collecting the products from the supplier’s facilities and pays the supplier an after production price. Both ways to centralize supply chain decisions have their advantages and disadvantages. A limitation of VMI is that in many situations downstream inventory points remain in control of the retailer, for example inventory in the stores. If the retailer is ac-tively influencing the upstream demand with inventory and pricing decisions at these downstream inventory locations, this may limit the suppliers’ ability to take control over the supply chain and introduce ambiguity about the costs such actions induce. If these inventory and pricing decisions are important instruments that the
retailer is not willing to give up, it could be more effective to centralize inventory control at the retailer. On the other hand, the extent to which centralized control from the retailer can take the supplier’s upstream processes into account is limited in a similar way. Especially if the retailer fails to take into account the strong hetero-geneity among suppliers’ shown in this thesis, centralized decisions from the retailer may be sub-optimal on a supply chain level. VMI is less subjected to this risk as it can benefit more from suppliers’ expertise and involvement. Most importantly, both centralizing decisions at the supplier or retailer does not make good communication between supplier and retailer obsolete. The shortcomings of centralization methods can be overcome by learning from each others processes and sharing information about decisions. As has been suggested for VMI before, such shared investment in the supply chain may actually be the main driver behind the successful outcomes of either method (Cachon and Fisher, 1997).
5.2
Limitations and generalizability of the research
find-ings
As with all empirical research, generalizability of the results of the studies in this thesis is to some extent limited by the context in which they were conducted. All supplier-retailer dyads that we observed in this thesis operate in the context of a modern supply chain of a European retailer with relatively short lead times, high levels of information sharing, low inventories and high service levels. Moreover, the fact that this retailer chose to (and was able to) drive the VMI implementation is likely to impose some selection bias on the suppliers involved in this study and the power relations on the dyads. In addition, the inventory measure applied in Chap-ters 2 and 3 of the thesis forms a proxy of the inventory in the supply chain based on shelf lives, and assumes that other factors, such as the production and transporta-tion lead times remain constant. Therefore, trends in this factors would remain un-observed in our models and thus may bias our results. However, continuous checks with the retailer’s account managers and the eight suppliers we have been in close contact with during the studies did not reveal such a trend.
Notwithstanding these limitations, the empirical context and the inventory mea-sures adopted also have important advantages that allowed us to explore long-existing
Conclusion and discussion 115
gaps in the academic literature. First, the focus on a single, retailer-driven introduc-tion of VMI enabled the collecintroduc-tion of rich qualitative and quantitative empirical data from both the supplier and retailer perspective. This longitudinal and multidisci-plinary approach allowed us to connect findings of prior research based on analyt-ical modeling – that often focuses strongly on the detailed ways in which suppliers can change their replenishment operations – with insights from empirical research relying on qualitative measures – that often take a more abstract managerial perspec-tive.
Second, the controlled transaction data set enables us to conduct a more detailed analysis and obtain cleaner results that would not be possible with data from many dissimilar firms and/or VMI introductions. In this respect, this thesis falls in the realm of prior studies that have used real-life data from a single supply chain (see, for example, Fisher and Ittner (1999), van Donselaar et al. (2010), or Dong et al. (2014)). Moreover, because VMI is such a multifaceted concept, the rich empirical context offered plenty of interesting mechanisms to study, such as, for example, the different ways suppliers can use VMI described in Chapter 2 and 3 and the contrasting collab-oration dynamics observed between suppliers in Chapter 4. Moreover, the sample contains a wide range of suppliers and products that can be compared in relation to the implementation of VMI.
Third, as outlined in the introduction of this thesis, the number of VMI configu-rations that exist in practice is very high. In a sample of suppliers that operate under diverse VMI configurations, inferences on the specific ways in which VMI affects supply chain performance of these suppliers is nearly impossible. Studying suppli-ers that deliver to the same retailer, under the same contracts, and with access to the same information allows for a ’fair’ comparison between suppliers with respect to the way they use VMI to change their replenishment processes. Furthermore, be-cause all of the retailer’s fresh food suppliers were faced with a change to VMI, our study was insensitive to selection bias with respect to suppliers that are likely to adopt VMI.
Finally, and perhaps most importantly, our research findings strongly built on theoretical concepts identified in the conceptual or analytical modeling literature on VMI. By relating our observations in the particular empirical context to insights from those studies we feel that many of the new insights offered by this thesis can be tran-scended to other contexts. Understanding why a supplier can use VMI in a particular
way – and then studying how VMI is actually used – is crucial in understanding the performance outcomes of any VMI introduction in practice.
The gap between insights from analytical modeling and empirical observations from industry as addressed in this thesis is not unique to research on VMI, but is signaled in Operations Management literature in general (Singhal et al., 2014). Con-cerns have long been raised that the gap between theory and practice is widening rather than narrowing (Wagner, 2002). This is even more remarkable when consid-ering that the amount and the quality of data used in supply chains is increasing rapidly. To start bridging this relevancy gap, many authors have pointed towards multi-method research as a potential solution (Singhal and Singhal, 2012; Sodhi and Tang, 2014; Choi et al., 2016). Our research underlines the need for this type of re-search. The extensive knowledge of the supply chain context that produces the data is crucial to properly identify models for quantitative analyses and interpret the out-comes. Such knowledge can only be obtained when also deploying more qualitative interview and observatory methods. However, because such an approach is time consuming, this imposes natural limitations on the scope of the research in terms of organizations observed.
5.3
Directions for future research
The introduction of this thesis describes the distinct components of VMI, all of which have individually been studied by means of analytical modeling, but also shows that few empirical studies have measured how the introduction of different components of VMI affects the performance of a supply chain. This thesis is the first to empirically study an introduction of VMI where a large retailer transfers both the control and ownership of inventory – on top of the sharing of information that already took place before introducing VMI. Future work could study the performance effects of other VMI configurations, for example, introducing a transfer of ownership of inventory in a setting where control of inventory already resided with the supplier.
Because suppliers typically supply multiple products using the same resources, the effects that we measure on a product level in Chapter 3 may interact on a supplier level. Exploratory analysis on the data used for the studies in this thesis revealed that that such effects may indeed affect suppliers replenishment actions under VMI. If a supplier supplies, for example, a combination of products with long and short shelf
Conclusion and discussion 117
lives, the higher responsiveness required to deliver shorter shelf in an efficient way may dictate the replenishment frequencies for the assortment with longer shelf lives too. In many cases, this product that imposes the highest requirements on respon-siveness not only dictates the replenishment frequency, but also the configuration of the entire production process, planning methods and the supporting IT. There-fore, future work may explore the impact of organization-level factors on the way supplier’s can and want to adapt to VMI.
Most empirical research thus far has focused on firm level inventories, rather than those in the entire supply chain. Chapter 3 shows how shelf life registrations, that retailers and suppliers increasingly store in data platforms, can provide new valu-able insights on the supply chain level inventories. Future work could exploit such measures to study other aspect of supply chain improvement efforts, such as more general information sharing or Sales & Operations Planning efforts. Furthermore, combing supply chain level inventory measures with firm specific performance met-rics can be used to compare how inventory observations at a supply chain level relate to the firm-level inventories. Such metrics could provide additional insights into how centralization of inventory decisions or information sharing affects the distri-bution of inventory over the different inventory positions in the supply chain and, accordingly, contributes to supply chain performance.
Finally, Chapter 4 shows how suppliers’ prior level operational maturity and goals and their development over time can have predictive value for VMI outcomes in the long term. Future work can benefit from expanding the analysis to multiple supply chain settings and in other supply chain collaboration efforts. Moreover, the study showed that, even though all suppliers could have access to the same informa-tion from the retailer, the performance difference between suppliers was explained to a large extent by how suppliers were willing and able to use the information. Fu-ture survey-based studies can benefit from these insights when constructing their instruments. In general, studies that connect the more qualitative insights on how suppliers are able to use information with the more developed quantitative knowl-edge of the (potential) value of information in supply chain processes would greatly benefit our understanding.
