INVENTORY CENTRALIZATION OR DECENTRALIZATION IN A NOT- FOR-PROFIT CONTEXT

(1)

1

INVENTORY CENTRALIZATION OR DECENTRALIZATION IN A

NOT-FOR-PROFIT CONTEXT

Master‟s Thesis Repair

Master thesis, MSc Supply Chain Management

University of Groningen, Faculty of Economics and Business

August 20, 2020

VINCENT SCHOLING

Student number: 3854043

e-mail: v.scholing.1@student.rug.nl

Supervisor / University of Groningen

Prof. Dr. D.P. Van Donk

Co-assessor / University of Groningen

Dr. O.A. Kilic

Word count: 11.997 words

(2)

2

Abstract

Purpose - This study aims to explore what factors influence inventory centralization and decentralization in a not-for-profit context. Whereas this has been extensively researched in for-profit sectors, knowledge on this subject was limited for not-for-profit organizations.

Methodology - A multiple-case study was conducted at four food bank DCs in The Netherlands. Findings – Inventory (de)centralization in the not-for-profit context is found to be influenced by the division of power among stakeholders, fairness, funding, voluntary labor, accessibility, and organizational goals. These characteristics limit not-for-profit organizations in their (de)centralization decision.

Practical Implications – We provide several insights for not-for-profit organizations regarding (de)centralization, among others; the benefits of decentralization, the importance of fairness, and reduction of supply risks.

Originality/contributions - Previous studies already found that drivers for (de)centralization differ between firm sizes and industries. We contribute by providing the insight that it also differs between for-profit and not-for-profit contexts and how.

(3)

3

1. Introduction

Organizations have to design logistics networks that suit their context‟s characteristics so that they best serve their customers‟ needs. A subject that has been addressed in network design literature is that of inventory (de)centralization. Already known is that trade-offs that influence inventory (de)centralization differ between contexts (Corts et al., 2019). Meanwhile, the interest in operations management research in not-for-profit contexts has risen and has led to a call for papers (Berenguer et al., 2015, 2017). Interestingly, not-for-profit contexts hold multiple fundamentally different characteristics than for-profit contexts. For example, not-for-profit organizations are funded differently than for-profit organizations and can even rely on donations. An example from the Dutch food banks showed that this reliance induced large supply uncertainties during the COVID-19 epidemic and even caused shortages in times their customers needed them most. How did this characteristic influence their choice between centralized and decentralized inventories?

Previous studies focused on identifying factors and trade-offs that drive network design and inventory decisions in for-profit contexts such as holding costs and service levels (i.e. lead time), transportation costs versus fixed costs (Erlebacher & Meller, 2000), and purchasing and fixed costs (Baker, 2007; Berman et al., 2011; Erlebacher & Meller, 2000; Oeser, 2019). However, no knowledge exists about how the inventory (de)centralization decision is influenced by not-for-profit contexts. These contexts are very different from for-profit contexts (Berenguer & Shen, 2019) and characterized by (among other factors) their sources of income (Gallien et al., 2017). Furthermore, the labor force in not-for-profit organizations can consist of volunteers (Sampson, 2006; Wisner et al., 2005). Lastly, fundamental to not-for-profit organizations is that their organizational goal is not profit-maximization (Natarajan & Swaminathan, 2017).

These factors indicate that typical drivers identified in previous literature might not hold or might need to be modified in the context of not-for-profit organizations. For example, reliance on donations takes away possibilities in purchasing such as quantity discounts as a result of economies of scale in orders, which would have led to more centralization in the context of for-profit organizations. Furthermore, reliance on donations could limit organizations‟ investment abilities. Also, receiving free supplies lowers holding costs and modifies the risk of holding too much inventory. These examples show that the not-for-profit context might influence inventory (de)centralization. So far, it remained a question of how these, and possibly different factors influenced inventory (de)centralization in a not-for-profit context. Our example from the food banks shows that not-for-profit organizations cope with the effects of these characteristics regularly. However, their effects are yet to be discovered. This urged the need for research into this subject to ultimately better understand the inventory (de)centralization decision drivers. This led us to the following research question:

“Which factors drive the inventory (de)centralization decision in a not-for-profit context and how?”

(5)

5 we focus on not-for-profit organizations with logistics as one of their core competences (i.e. excluding hospitals, educational organizations, arts, and culture, etc.).

The paper is structured as follows. Section 2 summarizes the influencing factors and trade-offs found in network design literature and identifies characteristics of not-for-profit contexts. Section 3 describes our research design. In section 4 our findings are presented, after which section 5 provides a conclusion and discussion of the results and its implications for literature. Section 6 relates our findings to implications for practice. Section 7 concludes our paper and addresses the limitations and opportunities for future studies.

2. Theoretical framework

This section aims to establish a theoretical framework to understand the theoretical background of our study. First, we start by identifying the trade-offs that lead to (de)centralization of inventories as found by previous research. Second, we identify the differences and similarities between the for-profit and not-for-profit contexts. The last paragraph links the two together and leads up to the conceptual model of our research.

2.1 Drivers of centralization and decentralization

To summarize the literature clearly, we described different trade-offs and distinguished between their implications for both centralization and decentralization. In centralized distribution systems, all demand is satisfied from a single or limited number of facilities, whereas decentralized systems supply demand from multiple, separate, facilities (Wanke & Saliby, 2009).

Inventory costs and service levels Firstly, we should note that holding inventory and supplying customers from that inventory is a key function of many warehouses (Baker, 2007). Therefore, it is important to first understand the role of inventories. Inventories are generally used as a buffer against uncertainty, bridging the gap between supplier lead time and customer lead time (Baker, 2007). Products that are kept in stock can be delivered straight away, greatly reducing customer lead time. However, excessive stock levels can entail high costs for storage, obsolesce, damage, deterioration, shrinkage, insurance, management costs, and costs of capital (Christopher, 2005). This leads to a trade-off between inventory holding costs and service levels. As supplier lead times increase, more stock is necessary to achieve the same customer service level (Lowson, 2002). As a result, for-profit organizations try to cut costs through economies of scale in purchasing to compensate for the additional holding costs, leading to a more centralized network (Balakrishnan & Natarajan, 2014; Hartman & Dror, 2005; Teo et al., 2001).

Risk-pooling Similarly, to reduce inventories without having to compromise service levels, studies in the for-profit context have found advantages of risk pooling, leading to centralized inventory policies.

Risk pooling builds on the centralization of inventories (Berman et al., 2011) and creates the advantage

(6)

6 2011). Three other studies argued and found potential to use decentralized systems without the need for higher amounts of safety stock (Croxton & Zinn, 2005; Mahmoud, 1992; Zinn et al., 1989). While the risk-pooling effects in some circumstances might yield benefits, the risk of disruptions increases with it. As the stock is held in a single location, a larger part of the supply chain is affected in case of disruptions (Y. Kim et al., 2015). Therefore, decentralized systems operating under supply uncertainty are found to have less cost variability and equal mean costs compared to centralized systems, also known as the risk-diversification effect (Snyder & Shen, 2006). Thus, decentralized systems are generally more attractive to risk-averse supply chains (Schmitt et al., 2015).

Transportation and distance Another trade-off is that of fixed costs versus transportation costs and distance to customers (Erlebacher & Meller, 2000). For each additional warehouse, the fixed costs increase. Therefore, centralization leads to a reduction of fixed, employee, and capital costs (Abrahamsson, 1993; Croxton & Zinn, 2005; Teo et al., 2001). In contrast, operating from multiple facilities leads to a reduction in physical distance and thus transportation costs (Wanke, 2009). Das & Tyagi (1997) studied the trade-off between inventory costs and transportation costs and found that in settings where transportation costs gain a higher share in total costs than inventory costs, network designs will be better off in a decentralized setting. Also, decentralization can increase the proximities to suppliers, improving the ability to make use of local suppliers while reducing transportation costs (Charles et al., 2016). Furthermore, decentralized systems will have shorter lead times due to closer proximity to customers (Cooper, 1983). The closer proximity has the effect of better customer service and therefore lower costs of lost sales (Abrahamsson, 1993). For products requiring a responsive supply chain (such as perishable products) (Fisher, 1997) these lead-time advantages can lead to a more decentralized configuration.

Interestingly, if the distance is defined in terms of lead time rather than physical distance (Abrahamsson, 1993), this provides a different perspective on the matter. Although centralization goes with a higher average physical distance to customers (Cooper, 1983), Abrahamsson (1993) showed that centralization does not always lead to increased customer lead time if a constant flow of deliveries out from the warehouses is maintained. Combining centralized consolidation techniques with frequent deliveries can thus lead to cost advantages without the disadvantage of longer lead times (i.e. without sacrificing service levels) (Abrahamsson, 1993).

Transportation costs were also found to drive the degree of centralization through the consolidation of goods (Cooper, 1983) and especially provides benefits in terms of cost reductions. Trade-offs between the number of purchases and transportation costs have been widely modeled mathematically in for example One-Warehouse-Multi-Retailer (OMWR) problems (Buzby et al., 1999; Cunha & Melo, 2016; Li et al., 2011; Muckstadt & Roundy, 1987; Solyali & Süral, 2012).

(7)

7 al., 2019). Therefore, decentralized systems require only local controlling capabilities (Corts et al., 2019). Also, decentralized systems are found to stimulate bottom-up plans much better than centralized systems (Y. H. Kim et al., 2014).

Products The last driver found in the literature is the product itself (Fisher, 1997). Due to the above-summarized cost advantages, low-margin products generally will be supplied through a centralized supply chain. Not only margins but also the perishability of products plays a role. Recent research on blood supply chains has shown that centralization can reduce outdate and shortages of perishable goods (Hosseinifard & Abbasi, 2016). Also Corts et al. (2019) suggested that increased inventory rotation can be easier achieved in centralized configurations. Also, the advantages of reduced facility investments may be greater when products require special warehouse investments (e.g. conditioning of food products, cooling, and freezing). Techniques such as cross-docking can combine these benefits, such as consolidation of transport through a centralized cross-docking facility while the inventories are only held on decentralized locations (Apte & Viswanathan, 2000).

(8)

8 TABLE 2.1: Summary of drivers (adapted from Corts et al. (2019))

(9)

9

2.2 Not-for-profit context

From existing literature, the not-for-profit sector can be distinguished from for-profit sectors by multiple characteristics; these are summarized in table 2.2. The first characteristic is the organizational goal. Although it is evident, it cannot be left unaddressed that the organizational goals of not-for-profit organizations are typically not focused on pursuing profit maximization (Natarajan & Swaminathan, 2017). Instead, their economic objectives are subordinate and need to be balanced with multiple, often conflicting, mission objectives from multiple stakeholders. Therefore, their operations and decision making are based on their stakeholders‟ mission objectives rather than profit maximization (Olson et al., 2005; Quarter & Richmond, 2001).

Simultaneously, whereas shareholders and executive board members typically possess a large control of the for-profit organization, power is spread among more stakeholders in not-for-profit organizations (Euske, 2003). The set of stakeholders is wider and spreads out over board members, volunteers, donators, and clients (Berenguer & Shen, 2019). The shift in power moves (network) decision making control to different stakeholders than in for-profit contexts and leads to a different power balance among stakeholders when compared to for-profit contexts.

Similar to the organizational goal that is often not aimed at profit maximization, the way not-for-profit organizations are funded is different from how for-profit organizations are funded. As customers in the not-for-profit context do not always pay for their received services, not-for-profit organizations rely on different sources of income. Instead, these generally rely on fees, funding from governments, or donations (Gallien et al., 2017). This can be either in the form of financial donations or in the form of goods. Uncertainty in both forms of income induces supply risks (i.e. shortages) and limits the effectiveness of the organization (Natarajan & Swaminathan, 2017). Also, not-for-profit organizations can receive donations that the organizations preferably do not receive (e.g. unhealthy food, broken equipment) (Ülkü et al., 2015). As a result, organizations in a not-for-profit context possess limited control over their resources (i.e. inventories and warehouses) and can in some cases only match demand & supply by limiting demand (Tarasuk et al., 2014).

(10)

10 customers being served (Loopstra et al., 2019). Connecting this to the use of donations, a higher density could lead to more (local) donations.

Table 2.2: Summary of not-for-profit characteristics Characteristics Explanation

Organizational goal

The raison d'être, the main reason for the existence of the organization. This is often not profit maximization but is based on the mission objectives of various stakeholders.

Amount of stakeholders

The number of stakeholders with influence. Stakeholders often possess similar levels of influence and therefore power is spread over many stakeholders of a similar size. Funding Not-for-profit organizations are generally funded through fees, funding from

governments, or donations.

Labor Not-for-profit organizations can make use of voluntary labor. However, this goes with problems in workforce planning.

Accessibility The distance between facilities and their customers. Minimizing this distance can be an important goal for not-for-profit organizations.

The above paragraphs have shown that not-for-profit contexts are thus fundamentally different from for-profit contexts in five ways. Linking these to the trade-offs for (de)centralization, we suggest that these could influence the trade-offs in various ways. For example, dependence on donations for supplies and resources might lower the costs of capital (Christopher, 2005), therefore influencing the trade-off between inventory costs and service levels. Additionally, lower inventory costs might reduce the need for safety stock reductions through risk-pooling. Therefore, the trade-off between safety stock reductions and the increased risks of centralized inventories (Y. Kim et al., 2015) might be influenced. Furthermore, in the context of not-for-profit organizations opening more locations improves accessibility (Loopstra et al., 2019), but also requires inventories to be kept at more locations. While we have not addressed all the possible influences from the above-mentioned characteristics in this paragraph, the ones mentioned already show that certain relationships could exist. Our research aims to clarify if and exactly how these and possibly other characteristics from the not-for-profit context influence the trade-offs between centralization and decentralization. The resulting conceptual model is shown in figure 2.1.

FIGURE 2.1: Conceptual model

Drivers

 Inventory levels

 Warehousing costs

 Transportation costs

 Service levels

 Management & Control

 Demand & supply patterns

 Product characteristics

(11)

11

3. Methodology

3.1 Research Design

Our research question was of exploratory nature. More specifically, our study aimed to understand which factors influenced the (de)centralization of inventories in not-for-profit contexts and how. A recent call for papers has shown that research on not-for-profit contexts is still in an emerging state. Therefore, our study required a research method that allowed us to identify possible new factors and relationships and simultaneously granted us the ability to understand how and why these relationships existed. Especially the latter requirement induces the need for rich, in-depth qualitative information, something for which case studies are especially suitable (Eisenhardt & Graebner, 2007; Yin, 1994). For example, conducting case studies granted us the ability to study and discuss the trade-offs with managers from a real-life setting. Gathering our data from a real-life setting also allowed us to better relate our findings to practice (Meredith, 1998). Gaining such in-depth information would not have been possible with the use of methods that rely solely on quantitative data, such as survey research (Ketokivi & Choi, 2014). Another important benefit that led us to the use of case studies was the ability to make use of multiple forms and sources of data (Leonard-Barton, 1990). This granted us the ability to improve the generalizability and reliability of our findings through the triangulation of both methods and sources. For example, the use of case studies gave us the ability to include secondary data from the cases‟ websites and would have allowed the use of quantitative data. Initially, we had thought that quantitative data would also be included in our research. For example, data on the inventory levels and demand levels at different locations could have been used to provide insights into the demand and supply patterns. However, our cases preferred to have as little administration as possible, and therefore no quantitative data could be gathered. As a result, the main sources of data were qualitative. We acknowledge that this has limited the reliability of our findings because the information from the interviews could therefore not be supported by quantitative data. Furthermore, the lack of quantitative data also meant that we could not determine the strengths of identified relationships. Also, being dependent on managers for their time and willingness to cooperate could have been a limitation (Meredith, 1998). However, we have found our case managers to be very helpful. The last decision in determining our research method was how many cases to include. Whereas single case studies can be powerful in providing detailed in-depth information, multiple-case studies can provide better generalizability and reliability (Leonard-Barton, 1990). Therefore, we adopted a multiple-case study design (Eisenhardt, 1989; Eisenhardt & Graebner, 2007).

3.2 Case selection

(12)

12 specifically suitable for our research as they included factors that were already identified and also operated in the context in which we were so interested.

The unit of analysis for this study was a food bank distribution center, consisting of one warehouse that delivers to multiple food banks. As we aimed to understand how the context-specific characteristics influenced (de)centralization of inventories, we aimed to include cases with decentralized as well as cases with centralized inventory policies. Including the two different outcomes of the (de)centralization decision would grant the ability to understand why cases operating in the same context opted for their chosen policy and how this decision was affected by the context-specific characteristics. To ensure the accuracy, reliability, and generalizability of our findings we made use of replication logic. We had selected two cases with decentralized policies and two cases with centralized policies. By including similar cases, we used replication logic (Eisenhardt, 1989; Yin, 2009) through literal replication. In addition, we used theoretical replication by including cases that would produce contrary results. All four cases were members of a national food bank association; therefore one could argue that this would have led to too many similarities between cases. However, all cases did operate and managed their inventories as independent foundations.

In selecting our cases, we acquired information about possible cases from case A. From this information, we had chosen to include three other cases to result in a case selection with replication logic. However, during the data collection phase, all chosen as well as all possible cases were found to have decentralized inventory policies. At that point in time, there were no possibilities to search for and include any other cases from different organizations to still ensure literal replication logic. We acknowledge that this limited the generalizability of our findings, as most of our data led to influences toward decentralized policies. As a result, we investigated four similar cases. From our data, we could conclude that selecting a higher number of similar cases would not have led to additional novel insights. As a previous study had shown that (de)centralization decisions differ between firm sizes (Pedersen et al., 2012) we used firm sizes as a control variable. Table 3.1 provides an overview of our cases.

TABLE 3.1: Summary of cases

Case Number of clients

of all customer food banks

Number of customer food banks

Stock keeping Product types

A 2.550 9 Decentralized Food and non-food

B 1.700 17 Decentralized Food and non-food

C 6.000 26 Decentralized Food and non-food

D 5.500 32 Decentralized Food and non-food

3.3 Data collection

(13)

13 To ensure reliability, semi-structured interviews were based on an interview protocol (Yin, 1994). The interview protocol can be found in Appendix I. During meetings with case A to discuss case selection, interviewees from case A validated if managers from other cases would be able to answer the given questions. Their feedback was used to make the interview questions clearer. The interview protocol consisted of three parts, an introduction, in-depth questions, and final questions. It was sent one week in advance so the interviewees could prepare by reading the questions. The introduction included general questions about the organization, such as the organizations‟ sizes, number of households supplied, number of clients, and general information about the interviewee. The in-depth section contained questions relating to the main drivers for the network-design decisions and how these were affected by the not-for-profit context. These questions aimed to provide insights into the rationale that determines why the DCs chose for a centralized or decentralized stock keeping and how this decision was affected by influencing factors. This section contained questions such as “why does the organization hold stock at these locations?”, “What influences the stock allocation process?”, “How does each stakeholder influence the stock allocation process?”, and “Is it a conscious choice to keep inventory at these locations? Could you explain?” Afterward, there was time for the interviewee to discuss matters that he/she felt were relevant and were not highlighted during the interview. All questions were open-ended questions to stimulate rich information. During the interviews, notes were made for further questions and data analyses. Furthermore, permission was asked for recording the interviews for transcribing purposes. All interviews were processed anonymously in the transcriptions after which the recordings were deleted. All transcriptions were sent back to the interviewees for validation and approval to serve as input for our research (Yin, 2009)

In addition to the interviews, secondary data from the organizations‟ websites and internal documents were collected to support answers given during interviews. Secondary archival data such as organizational policies, organizational structures, goals, and conditions for establishing new food banks were used to support findings from interview data and were found on the cases‟ websites. Furthermore, case A supplied hard copies of stock allocation examples and transport schedules. These were similar to those of other cases.

Three out of four interviews had similar lengths between 30 and 40 minutes. One interview had a duration of almost one hour. Table 3.2 summarizes the roles of the interviewees and interview durations for each case.

TABLE 3.2: Overview of data collection

Case Position of interviewee(s) Interview duration

A DC Board Member, Chair 38:02 Minutes

B DC Manager 32:14 Minutes

C Coordinator Food Procurement Warehouse Manager

59:52 Minutes

D DC Director 40:18 Minutes

3.4 Data analysis

(14)

14 coding). First, the cases were coded individually, after which cross-case analysis was conducted to find similarities and differences between cases. Microsoft Excel was used for this matter. Table 3.3 provides examples of data reductions from our analysis.

TABLE 3.3: Data coding Case and

interviewee

Data reduction (objective codes) Descriptive codes (second-order categories) Link to decision drivers C – Coordinator food procurement

“There is also twice a year what they call a members meeting and all those food banks are included. And of course [food bank x] has a lot of influence because they are by far the largest. But they are all included and they are all listened to.”

Levels of influence

Independence

Division of power

A - Chair “So the motive here for not keeping stock is that every food bank is free to determine when and how much they distribute. [Food banks] manage their own stock, we do not want to influence at all about that. ”

A - Chair These are also statutory

independent foundations, so we are not about that. We are about everyone getting their fair share. That is our goal.”

Independence Fairness

C –

Coordinator food

procurement

“[The food banks] came [to collect] once a week and

nowadays twice since the corona crisis.”

Transshipment frequency

Transportation

A - Chair “Yes or [food was collected] once every 14 days. [The food banks] are not so flexible. If they distribute on Wednesday, it makes no sense to come by on Friday. [So it depends on] opening times and distribution days.”

B – DC manager

“I have a note here and 8 more pallets are coming. But I have no idea when they will come. They may be right around the corner, but it may also be next week. I accepted it a week ago. But it is also possible that it will not come at all..” Supply uncertainty Supply sources Supply sources Funding Funding A- Chair “Each food bank has its own

supply of goods”

(15)

15 and as a DC we receive a modest

contribution from the province to ensure that we can function as a DC.”

Supply Uncertainty D – DC

director

“We've learned in the past that we shouldn't be looking at what might be coming up because once that is late we already have problems and then food banks call us and say: hey! I would get that and this but it's not there.” C-

Warehouse manager

“[Our goal is] to ensure that food is not destroyed but used, and that we help people with that is also very nice. But the real goal was not to see everything go to the trash.”

Objectives of the organization

Organizational goal

A- Chair “[Food banks] often have

multiple distribution points. They have several days when they are [at place x] and then [at place y].”

Number of locations Accessibility

4. Findings

In analyzing the data, we managed to find direct as well as indirect influences of the not-for-profit context on (de)centralization. More specifically, we have found influences from (I) the division of power among stakeholders, (II) fairness, (III) how the case organizations were funded, (IV) the use of voluntary labor, (V) accessibility, and (VI) the organizational goals. The following paragraphs will further elaborate on each finding.

4.1 Division of power

Our findings show that how inventories are managed is influenced by the way organizations are structured, and, more importantly, how power is spread among its (external) stakeholders. More specifically, our data suggest that how power is spread among stakeholders has an indirect effect on decentralization by that it acts as an antecedent of management and control. Our data showed that there were three types of organizations present; local food banks, food bank distribution centers, and an association. Important to note here is that each case included one DC that delivered supplies to several local food banks. Data from conversations prior to our interviews, as well as from the interviews itself showed that all facilities in each of the four cases operated independently and therefore managed their own inventories. However, in each of the four cases, all facilities were members of the national association. The role of the association was to ensure that the service offered at all facilities were roughly the same. For example, the association prescribed rules for accepting new clients and the use of volunteer labor. This was also explained during the interview from case A:

“You could see it as a franchise organization. Independent organizations that all operate according to the principle of the association. This also has to do with criteria that clients must meet to use the food banks. Things like that. They conform to the rules of the association.”

(16)

16 Meanwhile, the roles of the DCs and the local food banks were much more operational and included the responsibility for the distribution of food to clients. However, our data has shown something very interesting about the spread of power between these types of organizations and how this influenced the management and control of inventories. Namely, although each of the facilities operated independently from one another and managed their own inventories, the authority over the DCs‟ inventory policies were spread over all local food banks. For example, the managers from case A explained that annual meetings took place in which the local food banks could determine the DC‟s inventory policies. Also, during these meetings the DCs had to be able to justify their action to the local food banks:

“And [the dc] has a regional consultation platform with all affiliated food banks in which [it] discusses the annual accounts and [its] policies, which is the regional consultation platform. In that sense, the local food banks also have a say in the policies of the DC.” (Chair case A).

Data from case B showed that how much influence each local food bank had during the meetings was determined by their size:

“There is also twice a year what they call a members meeting and all those food banks are included. And of course [food bank x] has a lot of influence because they are by far the largest. But they are all included and they are all listened to.” (DC manager case B).

During the interviews, managers from cases A, B, and C explained that the local food banks possessed so much power because they had created and paid for the existence of the DCs. For example, the warehouse manager from case C explained that originally the DCs did not exist. Instead, all food banks started as individual foundations and collaborated to establish a DC to gain the ability to accept large quantity donations. Therefore, the local food banks wanted to stay in charge of how these donations were further distributed to the local food banks. How these inventories were managed was thus determined by the local food banks. As a result, the DCs had no authority to change their inventory levels themselves and also had no access to information about inventory levels downstream in the supply chain. Trade-offs for centralization and decentralization were thus not made by the DCs, but by their stakeholders. Power over the DCs‟ inventory policies was thus indeed spread over many individual stakeholders, while the DC itself possessed little power. Overall, these cases confirmed that power in not-for-profit organizations is spread among large groups of stakeholders and that this also influences the management and control of inventories.

4.2 Fairness, stock allocation, and risk-pooling

(17)

17 that the underlying reason for such an allocation rule was that all customers food banks aimed to achieve a fair stock allocation and that this rule was perceived as fair. To underline the importance of fair stock allocation, the director from case D explained that the local food banks to which his DC delivered even had access to the DC‟s inventory management system to monitor if the allocation process had been executed fairly:

“Food banks can log in to our system to see what came in that week, what is being distributed that week, and what they should actually receive according to the consumer unit. And also what they actually got from us. They can log in [to our system] and see per food bank [if it is done] correctly.” (DC director case D).

These findings show us that fairness is deemed important in the context of our cases. Of course, we have aimed to understand specifically what caused fairness to become so important in this context. Our data suggested that a possible explanation could lie in that these cases suffered from supply shortages. The reasoning for is that if there are shortages, these shortages should be equally divided among all stakeholders. For example, the board member from case A mentioned that their DC structurally did not receive enough supplies to meet demand. Also, the manager from case B explained that sometimes not enough food was supplied. However, in contrast to cases A and B, cases C and D explained that they usually received more than enough supplies due to their location. Though, these cases used the same allocation rule. Despite our efforts, the data provided no further insights into why also these cases used the same allocation rule and what else could make fairness so important.

Risk-pooling Our data also showed that fairness had implications for our case DCs‟ abilities to make use of risk-pooling. Firstly, it showed that fairness can prohibit the use of risk-pooling. For example, the allocation rule used by the four cases restricted the central DCs‟ abilities to change shipment volumes to ultimately better meet demand at the various locations. None of the DCs were allowed to deviate from the allocation rule determined by the local food banks. However, two out of four cases did mention that there were possibilities to send extra supplies to food banks that suffered from shortages. For example, in case A it was possible for local food banks to voluntarily give a share of their allocated products to another food bank. The manager from Case C explained that did not mind occasionally allocating some extra food to food banks that suffered from shortages. However, it is important to note that case C also had many more supplies to give away and that this would not have much consequence for the allocation to other food banks.

Although the findings did not show that fairness leads to either centralization or decentralization, they did show us that fairness can play an important role in the allocation of stocks that flow through a central warehouse. Specifically, it affected the management and control of inventories as the allocation of stocks was based on fairness. Our data suggested that fairness became so important due to supply shortages. However, our data has also shown that fairness was important for cases that did not experience supply shortages. Therefore our data does not provide enough insights to clearly determine the root cause for making fairness so important. Lastly, our findings show that fairness can prohibit the use of risk-pooling.

4.3 Funding, supply patterns, transportation and service levels

(18)

18 Supply patterns Our data has shown that the dependence on donations induces an unreliable supply pattern. For example, managers from all four cases explained that it is always a question of what and how much will be donated. Even more so, managers from cases B and D explained that deliveries that were agreed on were sometimes canceled or delayed. The chair of case A mentioned that such unreliable supply patterns are caused by their dependence on donations. Because no transactions take place, our cases are dependent on what‟s left:

“That is the difference with if you are a company and you put in orders such as ice cream when you think 'I will sell a lot of ice cream.' No, it is only when it is cold that we get all the ice cream that is leftover and is close to its shelf life.” (Chair case A).

These examples show us that a dependence on donations can cause a very unstable pattern. The dependence on donations can thus be seen as an antecedent of the factor „supply patterns‟. However, data from the interviews also showed that this relationship is moderated by proactivity. For example, whereas case B explained that it is just a „wait and see what is offered‟, cases A and C were very proactive in finding new supply sources. Case A had set up a daily route along all supermarkets in their region which resulted in almost 50 more locations where supplies could be picked up daily. Also, case C was very proactive and had set up contracts with some of the largest food producers in the country. Due to these contracts, the total number of supplied units was increased by roughly 2.5 million in just a few years:

“I said last year we had 5.6 million products, when I came we were at 3. That has more than doubled and we expect it to increase. And that may not be my merit, but it is mainly the contract with the big [food producers].” (Coordinator food procurement case C).

Overall, these examples have shown that a dependence on donations causes supply uncertainty and that the level of proactivity influences just how much supply uncertainty is caused.

Access to local donations A finding that relates directly to decentralization is that of access to local donations. Data from all four cases provided insights into how much local food banks relied on their DC as their main source of supply. Managers from all four cases explained that the reliance differed much between the food banks, but that most food banks received supplies through local donations. This has shown that because there are so many individual food banks, this gave access to local supplies and lowered the dependence on the central DCs. Logically, such a decrease in dependence on a single DC for supplies decreases the risks associated with supply disruptions. Access to local supplies can thus be used as an argument to opt for decentralization in not-for-profit contexts.

(19)

19

“[Transportation] could be more practical. Now, there are food banks that all drive past each other with their vans. If they would work together and have one van drive for a few food banks it could be more efficient.” (DC director case D).

Case C, therefore, had just started a project to explore the options of a centralized transportation system in which all shipments were delivered by the DC itself. However, as they had just started this project, no results were available yet. These findings show that there is a rather interesting contradiction taking place. While decentralization gave access to local donations and sponsors (i.e. higher income), this went along with higher costs due to inefficiencies in their overall transportation system. These findings show us that the way how not-for-profit organizations are funded can have a positive as well as a negative effect simultaneously. For our cases, the sum of the increased income and extra costs associated with inefficiencies had led to decentralization. However, this might be different for other not-for-profit organizations.

Service levels Our data have provided insights into a very interesting finding regarding service levels. Our data suggested that reliance on donations can be a reason for not having service levels. In contrast to for-profit contexts, customers of charity organizations can only be grateful for what they receive. For example, although the managers spoke of meeting demand, managers from two out of four cases (A and B) explained that no actual „demand‟ existed. In answering a question about the ratio between supply and demand, the manager from case B explained the following:

“The food banks don't „need‟ anything at all. The food banks accept what they get. That's the deal we‟ve made. They cannot say: I need 100 boxes of x.” (DC manager case B).

The managers from cases A and B explained that there were no actual demand and service levels. Instead, the local food banks could only be satisfied with everything that was supplied by the DC. This was similar to data from case A, where the chair explained that also the DC could only be satisfied with what suppliers delivered to the DC:

“We are dependent on what is left. We don't have to pay for it, we always have to say thank you. We are not a customer in the sense of… a customer who pays. We are damn glad that the supplier will bring it to our doorstep.” (Chair case A).

These examples show us that especially in charity organizations where supplies are donated service levels do not always exist. Instead, the aim would logically be to supply as much aid or supplies as possible.

4.4 Voluntary labor

(20)

20

“If you ask people who say to be willing to become volunteers, the first thing [those people] think about is handing out packages to the clients. At least those people think about tasks that involve interaction directly with the client. [But] the rest of the week we are busy with all kinds of other things and [those people] do not know that. [...] [those people] can combine those tasks and then they do it. But if you say that those people can only prepare pallets for another food bank, then there are few who do [volunteer work].” (DC director case D).

However, the data from case B contrasted with the data from cases A and D. Whereas cases A en D showed that volunteer tasks could be combined, the manager from case B explained that tasks could not be combined. Instead, he explained that tasks in the DC are too different from tasks in food banks; whereas tasks in a DC involve carrying heavy products, tasks in a food bank are much more social (e.g. having a chat with a client). The manager explained that those tasks would need very different types of volunteers and therefore the tasks could not be combined. While the tasks in all cases were the same and cases A and D showed that sharing volunteers was indeed possible, the managers seemed to have had a different opinion. Our data has not provided more insights as to what was seen as the barriers for combining workforces, other than the one just mentioned. Our data thus showed that the use of volunteer labor can lead to centralization through the ability to combine workforces and tasks, but that this is not always the case.

4.5 Accessibility

Our data provided limited insights into the effects of accessibility. However, it did show that organizations can have accessibility preferences and that such preferences can lead to decentralization. For example, secondary data from the association‟s website showed us that the association preferred to have one member food bank per municipality. This preference resulted in a total number of 170 member food banks throughout the country. Also, during the interview with case A it was mentioned that local food banks often set up several distribution points in their region to minimize the distance to clients. This was also mentioned by the coordinator food procurement from case C. Although data from the other cases did not include insights regarding accessibility, these findings do show us that preferences regarding accessibility can lead to decentralization.

4.6 Organizational goal

Last but not least, our data provided limited insights into the effect of organizational goals. As earlier mentioned, our cases were quite specific. However, because their organizational goal was so specific, this has shown us how it had influences one of the existing drivers of (de)centralization. Namely, the goal of all four cases was to reduce food waste. Therefore, all four cases focused on foods that were at the end of their expiration dates. While it is logical, it should not be ignored that the goal of the organization ensured that products had to flow through the supply chain as quickly as possible. The effect of this is that there was little time to store products centrally and that stocks sometimes even had to reach the end customers within 1 day. As a result, managers from all case DCs wanted to store as few supplies as possible and wanted to send all supplies to the local food banks as quickly as possible. Because managers wanted to send all stocks to the local food banks as quickly as possible and to keep as little stock as possible in the DC, this automatically leads to decentralization. The goal of the organization has thus influenced the product characteristics, which has led to decentralization.

5. Discussion and conclusion

(21)

21 upcoming interest in operations research in the not-for-profit sector (Berenguer et al., 2015, 2017), it was yet to be studied what influenced the trade-offs in the not-for-profit sector. From previous literature, we had identified trade-offs between the following factors that influenced (de)centralization: inventory levels and warehousing costs (Baker, 2007), transportation costs (Erlebacher & Meller, 2000), service levels (Lowson, 2002), management & control (Abrahamsson, 1993; Cooper, 1983; Croxton & Zinn, 2005; Das & Tyagi, 1997), demand & supply patterns (Berman et al., 2011), and product characteristics (Hosseinifard & Abbasi, 2016). We have also built upon recent work that identified the game-changing characteristics of this specific context (Berenguer & Shen, 2019; Euske, 2003). In answering our research question, we provide insights into multiple direct as well as indirect influences from these factors on existing trade-offs and factors for (de)centralization. We contribute to the existing literature by providing new insights regarding the influences of (I) the division of power among stakeholders, (II) fairness, (III) funding, (IV) voluntary labor, (V) accessibility, and (VI) organizational goals. Their implications for existing trade-offs and factors are discussed in the following paragraphs.

5.1 Division of power and management & control

An important trade-off for centralization and decentralization with several (dis)advantages for each outcome (Abrahamsson, 1993; Y. H. Kim et al., 2014; Pedersen et al., 2012; Teo et al., 2001) is that of determining who possesses authority over the management and control of inventories. Previous studies have found benefits from centralizing the management and control in increased visibility (Teo et al., 2001), better lead time precision and thus a better information flow towards customers (Abrahamsson, 1993; Corts et al., 2019). However, it was also known that decentralization of management and control would be more beneficial as local management possesses better capabilities for making operational decisions than top management (Pedersen et al., 2012). Furthermore, decentralization would require less strategic alignment and planning (Axsäer, 2005) and would stimulate bottom-up plans (Y. H. Kim et al., 2014). However, how these trade-offs were influenced by the not-for-profit context were striking. Euske (2003) already found that power (in a general sense) in not-for-profit contexts is typically spread over many individual stakeholders. Our findings strongly confirmed that this also holds for the division of power over inventories. However, most striking is that our findings showed that the power to make such trade-offs actually lies outside the focal not-for-profit organizations and is spread over many individual stakeholders. For example, the DCs in our cases possessed no power to change their own inventory policies. Therefore, the case DCs were not able to make trade-offs as above mentioned as a basis for their inventory policies. Instead, their inventory policies were fully dependent on the preferences of their stakeholders and the ones making the trade-offs were actually the large group of stakeholders. Although we had expected that the large group of stakeholders would have influenced these trade-offs, we had not expected that external stakeholders could possess just so much influence. Following our findings, we propose that:

P1: Power over inventory management and control in not-for-profit contexts is spread over large groups of stakeholders, whereas the focal organizations possess very little power. Therefore, also the trade-offs for (de)centralization are made by the large groups of stakeholders rather than by the focal organizations.

(22)

22 sponsors. Our extension is very similar to the work of Charles et al. (2016), who in their model for efficient humanitarian supply networks proposed that networks that use local suppliers should use a rather high number of decentralized warehouses. Therefore, our finding lets us propose that:

P2: For not-for-profit contexts, decentralized management & control over inventories improves the access to local supplies and local sponsors.

5.2 Fairness, management & control and risk-pooling

Management and control In the existing literature, fairness was not yet identified as an important factor for the centralization and decentralization of inventories. In contrast, our findings have shown that fairness was actually very important for the management and control of inventories in the not-for-profit context from our cases. Although our findings do not indicate that fairness directly leads to centralization or decentralization, they did show a large impact on (de)centralization through the trade-offs in the management and control of inventories. More specifically, our findings indicated that fairness plays an important role in the allocation of inventories from a central DC. This makes sense when we consider that power over the allocation policy was spread over many individual stakeholders, which is similar to the finding from Euske (2003) who found that it is typical for not-for-profit contexts that power in a general sense is spread over many individual stakeholders. In such contexts one stakeholder can't enforce its will on another, therefore making equality between stakeholders a credible result. Although not many studies on inventory management have focused on fairness, we were able to find one study that focused on fair stock allocation for limited supplies (Berenguer et al., 2017). That study was similar to ours in that both studies involved data in which supplies were too limited to meet demand. Our findings suggested that this negative ratio between supply and demand is what made fairness so important and that fairness has caused inventories to be equally allocated to all end customers. Therefore, we propose that the ratio between supply and demand antecedes fairness and that fairness itself acts as a mediator to management & control:

P3a: In not-for-profit contexts, a negative ratio between supply and demand enables fairness to be an important factor for the management and control of inventories.

Risk-pooling Similar to proposition 3a, our findings have shown that fairness can prohibit the use of risk-pooling. Risk-pooling is based on the centralization of inventories (Berman et al., 2011) by combining multiple variable demand patterns into one stable demand pattern. Our findings have shown that fairness has caused a situation in which inventories were allocated equally to all customer locations and that therefore the individual demand patterns had no influence on stock allocation. Therefore, fairness prohibited the use of risk-pooling. This finding shows us that fairness can act as a barrier for risk-pooling and that typical benefits resulting from risk-pooling such as cost reductions through reduced safety stocks (Maister, 1976; Oeser, 2019; Teo et al., 2001; Zinn et al., 1989) cannot be achieved. As a result, the centralization of inventories yields fewer benefits for not-for-profit contexts where fairness is deemed important. Therefore, we propose that:

P3b: In not-for-profit contexts, fairness can act as a barrier for using risk-pooling in centralized supply chains.

5.3 Funding

(23)

23 not-for-profit organizations rely on different sources of income. Instead, these generally rely on fees, funding from governments, or donations (Gallien et al., 2017). Typical for food banks (Tarasuk et al., 2014), our findings have shown that demand could not always be met with the number of donated supplies. Although it was already known that uncertainty in these three forms of income induces supply risks and limits the effectiveness of an organization (Natarajan & Swaminathan, 2017; Tarasuk et al., 2014), we link this finding to the field of research on centralization and decentralization by suggesting that the way not-for-profit organizations are funded acts as an antecedent of supply patterns. However, our findings also suggested that how much uncertainty is induced is influenced by an organization‟s proactivity to search for supply sources. Therefore we propose that:

P4a: How not-for-profit organizations are funded precedes supply patterns.

P4b: Proactivity moderates the level of supply uncertainty induced by how not-for-profit organizations are funded.

Risk-pooling Our findings have indicated that decentralized management of inventories can lead to better access to local donations. This effect has already been explained for proposition 2. However, we now link this finding to research on risk-pooling and risk management. Previous studies have found that risk-pooling can lead to great cost reductions (Fleischmann, 2016; Teo et al., 2001). In contrast, other researchers have argued that this has to weigh up to the increased effects of possible disruptions because disruptions will affect a larger part of the supply chain if it is centralized (Y. Kim et al., 2015). Decentralized inventories are therefore generally more attractive to risk-averse supply chains (Schmitt et al., 2015; Snyder & Shen, 2006). Our findings complement these findings and show that decentralized not-for-profit supply chains are less sensitive to disruptions through access to local sponsors. Therefore, we propose that:

P5: Risk-averse not-for-profit organizations are most likely to be decentralized through access to local sponsors.

Lastly, our findings have shown that a reliance on donations can act as a barrier for using service levels. Therefore we propose that:

P6: Funding of not-for-profit organizations can precede service levels by acting as a barrier if organizations rely on donations.

5.4 Voluntary labor and exchanging volunteers

Not-for-profit contexts are much different from for-profit contexts in that they allow the use of voluntary labor. Sampson (2006) has shown that this often causes difficulties for workforce planning. They have shown that expanding the pool of volunteers is difficult, causing the need to balance labor shortages among tasks. While previous researchers have found why expanding the labor force is difficult (Anderson & Moore, 1978; Farmer & Fedor, 1999), our findings have shown how not-for-profit organizations cope with this problem through centralization. More specifically, our findings have shown that organizations are centralizing their activities so that they can better balance labor capacity over tasks. Therefore, we propose that:

(24)

24

5.5 Accessibility

Although our data provided limited information about the effects of accessibility, it did show that goals for accessibility can lead to decentralization. Especially the trade-off between fixed costs and transportation costs (Erlebacher & Meller, 2000) is influenced by this. Whereas centralized supply chains benefit from reduced fixed warehousing costs (Abrahamsson, 1993; Croxton & Zinn, 2005; Teo et al., 2001), decentralized supply chains benefit from reduced distance to customers (Wanke, 2009). Our findings suggest that goals for accessibility can increase the focus on distance reduction and thus moderates the relationship between distance (embedded in transportation costs) and (de)centralization. Therefore, we propose that:

P7: Accessibility positively moderates the relationship between the factor transportation costs (i.e. distance) and decentralization.

5.6 Organizational goals and product characteristics

Not-for-profit organizations are fundamentally different than for-profit organizations by their organizational goals (Natarajan & Swaminathan, 2017). Previous studies have shown that the goals of not-for-profit organizations are based on mission objectives other than profit maximization (Olson et al., 2005; Quarter & Richmond, 2001). Our findings have shown that these goals influence can influence the factors that lead to (de)centralization. In our cases, it has led to the use of perishable products and thus resulted in decentralization through the need for a fast supply chain. However, we acknowledge that cases with different goals would not have shown this relationship. This lets us propose that:

P8: The organizational goal of not-for-profit organizations can act as an enabler for product characteristics as a factor in the trade-offs that lead to (de)centralization.

6. Managerial implications

(25)

25

7. Limitations and further research

Similar to any other studies, our study has its limitations. Firstly, we have only included food banks in our research. We acknowledge that these were quite specific case organizations and therefore our findings hold limited generalizability for other not-for-profit organizations. For example, the organizational goal of reducing food waste caused led us to the relationship between the organizational goal and product characteristics. However, for organizations with non-perishable products or different organizational goals this finding logically does not hold. We could have overcome this by adding cases from different contexts. Another limitation is that our cases were all very similar. For example, the cases were all related to one another and operated with the same objectives and contexts. Therefore, the data we gathered from each of the cases provided similar results. This made it harder for us to conduct critical cross-case analysis. We acknowledge that this is partly due to how we selected cases. To select cases, we used data that we received from one of the cases without comparing it with other data. Instead, we assumed that their data would be reliable because the cases were related to one another. However, this turned out not to be the case. Also, no qualitative data was used to support our findings. This limited our findings in several ways. For example, subjective data from interviews could not be supported by objective data. Also, the strengths of identified relationships could not be tested. Another limitation of our study lies in the collection of data. Due to circumstances, the data could not be collected together with a fellow student. This may have affected how much information we were able to extract from these cases. Moreover, the data was not always detailed enough to understand how and why relationships exist. For example, the data only provided limited insights regarding accessibility. We suggest future research to focus on cases from other not-for-profit contexts (e.g. cases that are funded differently, have different organizational goals, or distribute different products). Including cases from different contexts could lead to new insights into our identified relationships. Also, future research could focus on cases that operate on a global scale. Previous research had already shown that the drivers for (de)centralization are influenced by an organization‟s size. Therefore, cases that operate on a global scale could lead to different insights. We also suggest future research to include quantitative data (e.g. information on inventory levels and demand patterns) to support qualitative data.

References

Abrahamsson, M. (1993). Time-Based Distribution. The International Journal of Logistics

Management, 4(2), 75–84.

Anderson, J. C., & Moore, L. F. (1978). country‟s. Journal of Voluntary Action Research, 7, 120–129. Apte, U. M., & Viswanathan, S. (2000). Effective Cross Docking for Improving Distribution

Efficiencies. International Journal of Logistics Research and Applications, 3(3), 291–302. https://doi.org/10.1080/713682769

Axsäer, S. (2005). A simple decision rule for decentralized two-echelon inventory control.

International Journal of Production Economics, 93–94, 53–59.

https://doi.org/10.1016/j.ijpe.2004.06.005

Bacon, C. M., & Baker, G. A. (2017). The rise of food banks and the challenge of matching food assistance with potential need : towards a spatially specific , rapid assessment approach.

Agriculture and Human Values, 34(4), 899–919. https://doi.org/10.1007/s10460-017-9783-y

Baker, P. (2007). An exploratory framework of the role of inventory and warehousing in international supply chains. The International Journal of Logistics Management, 18(1), 64–80.

(26)

26 Balakrishnan, A., & Natarajan, H. P. (2014). Integrated procurement planning in multi-division firms.

Production and Operations Management, 23(10), 1795–1810.

https://doi.org/10.1111/poms.12096

Berenguer, G., Keskinocak, P., Shanthikumar, J. G., Swaminathan, J. M., & Van Wassenhove, L. (2015). Call for Papers: Special Issue of Production and Operations Management : Not for Profit Operations Management . Production and Operations Management, 24(2), 352–353.

https://doi.org/10.1111/poms.12339

Berenguer, G., Keskinocak, P., Shanthikumar, J. G., Swaminathan, J., & Van Wassenhove, L. N. (2017). A Prologue to the Special Issue on Not-for-Profit Operations Management. Production

and Operations Management, 26(6), 973–975. https://doi.org/10.1111/poms.12730

Berenguer, G., & Shen, Z.-J. (Max). (2019). Challenges and Strategies in Managing Nonprofit Operations: An Operations Management Perspective. Manufacturing & Service Operations

Management, Articles i(April 2020), 1–18. https://doi.org/10.1287/msom.2018.0758

Berman, O., Krass, D., & Mahdi Tajbakhsh, M. (2011). On the benefits of risk pooling in inventory management. Production and Operations Management, 20(1), 57–71.

https://doi.org/10.1111/j.1937-5956.2010.01134.x

Buzby, B. R., Campbell, G. M., & Webb, I. (1999). Cyclical Schedules for One-Warehouse ,

Multi-Retailer Systems with Dynamic Demands. 50(8), 850–856.

Charles, A., Lauras, M., Van Wassenhove, L. N., & Dupont, L. (2016). Designing an efficient humanitarian supply network. Journal of Operations Management, 47–48, 58–70. https://doi.org/10.1016/j.jom.2016.05.012

Christopher, M. (2005). Logistics & Supply Chain Management. In International Commerce Review (Vol. 7, Issue 2). https://doi.org/10.1007/s12146-007-0019-8

Cooper, M. C. (1983). Freight Consolidation and Warehouse Location Strategies in Physical Distribution Systems. In Journal of Business Logistics (Vol. 4, Issue 2, pp. 53–74).

Corts, N. F., Herbert-Hansen, Z. N. L., Larsen, S. B., & Khalid, W. (2019). The degree of inventory centralization for food manufacturers. Production Engineering, 13(1), 21–32.

https://doi.org/10.1007/s11740-018-00872-1

Croxton, K. L., & Zinn, W. (2005). Inventory Considerations in Network Design. Journal of Business

Logistics, 26(1), 149–168.

Cunha, J. O., & Melo, R. A. (2016). On reformulations for the one-warehouse multi-retailer problem.

Annals of Operations Research, 238(1–2), 99–122. https://doi.org/10.1007/s10479-015-2073-4

Das, C., & Tyagi, R. (1997). Role of inventory and transportation costs in determining the optimal degree of centralization. Transportation Research Part E: Logistics and Transportation Review,

33(3), 171–179. https://doi.org/10.1016/S1366-5545(97)00019-7

Eisenhardt, K. M. (1989). Building Theories from Case Study Research. The Academy of Management

Review, 14(4), 532–550.

Eisenhardt, K. M., & Graebner, M. E. (2007). Theory Building from Cases : Opportunities and Challenges. Academy of Management Journal, 50(1), 25–32.

https://doi.org/10.1177/0170840613495019

INVENTORY CENTRALIZATION OR DECENTRALIZATION IN A NOT- FOR-PROFIT CONTEXT