A FRAMEWORK FOR THE ALLOCATION OF ITEMS AT MULTIPLE WAREHOUSES

(1)

A FRAMEWORK FOR THE ALLOCATION OF ITEMS

AT MULTIPLE WAREHOUSES

Master thesis, MSc Supply Chain Management

The University of Groningen, Faculty of Economics and Business

24 January 2021

KARLINDE VAN ARENDONK Student number: 2787881 E-mail: k.j.van.arendonk@student.rug.nl Supervisor/ university dr. N.D. van Foreest Co-assessor/ university dr. O.A. Kilic

Supervisor/ field of study J. Melis MSc Case company

(2)

2

ABSTRACT

Customer orders are split in multiple shipments because the items need to be sourced from multiple warehouses. This study explores whether the allocation of items at multiple locations will reduce the number of split shipments. In this light, it is important to determine which items are relevant (i.e. sold and combined frequently) for multiple warehouse allocation, and which items can be allocated at multiple warehouse locations (i.e. item characteristics and warehouse design aligns). A framework is developed that identifies which items need to be allocated at multiple warehouses to reduce the number of split shipments. This framework is implemented and evaluated by an online retailer (case company). This study finds that more than 15% of the split shipments could have been reduced by allocating 550 items at multiple warehouse locations. The framework is a starting point for multiple warehouse allocation of online retailer possessing multiple warehouses.

(3)

3

1. INTRODUCTION

“Jack is looking for a headphone on the internet. When he found one and finalised his online order, he realised that there would be no shipping cost if he added another item. He added an

extra item – a battery, which he did not need in the first place – to his online shopping basket and took advantage of the free shipping. However, Jack was surprised when the headphone

and the battery were delivered as two different packages. This got him questioning…”

1.1 Problem Context

Online retail is growing and the boundaries of the online retailers’ warehouses are getting insight. More extensive assortments and higher inventories must be kept in the warehouses to keep up with customer demands. These developments press on the capacity, causing complications at online retailers’ warehouse (Boysen, de Koster, & Weidinger, 2019). More warehouses are needed to enlarge the capacity and to store the extra inventory needed (Adams, 2020; Lee & Elsayed, 2005)

The breaking apart of customer orders in multiple shipments is a common phenomenon in online retail. Figure 1.1 depicts the situation described above. The order needs to be sourced from different warehouses, resulting in two separate shipments. The last-mile delivery, the shipment towards the customer, is considered the expensive, inefficient, and polluting component of the supply chain process (Gevaers, Van de Voorde, & Vanelslander, 2014).

FIGURE 1.1

(6)

6 Hence, it is of value to reduce the number of shipments. If the items were stored at the same warehouse, the order could have been shipped in a single package. The allocation of items may influence the number of shipments.

1.2 Research Goals

Because of the growth of e-commerce and online retailers, there is a demand for support tools designed specifically for the online retail sector. The aim of this study is to develop a framework that helps online retailers identify which items should (not) be allocated at multiple warehouses to reduce the number of split shipments.

There is limited research on the allocation of items across multiple warehouses, split shipments and related support tools (Holzapfel, Kuhn, & Sternbeck, 2018). Existing literature is mainly focused on the location of multiple warehouse locations and the environmental matters of reducing the number of shipments (Berman, Krass, & Tajbakhsh, 2012; Farahani, Bajgan, Fahimnia, & Kaviani, 2015). The study will contribute to the literature on supply chain management. The results will also be of relevance for supply chain and logistics departments of large online retailers, possessing multiple warehouses.

1.3 Research Questions

The research question of this explorative study is as follows:

How can x% (e.g. 10%) of the split shipments be reduced by allocating items at multiple warehouse locations?

(7)

7 The following sub-questions are formulated to build-up to the answer of the main research question:

1. Which items should be allocated at multiple warehouses?

The goal of this question is to determine which items are interesting for multiple warehouse allocation. It is expected that multiple stock allocation depends on the relevance (e.g. sales, conversion) of the item and that not all items are as relevant.

2. What needs to be considered in the allocation process across multiple warehouses?

The purpose of this question is to identify specific characteristics that determine the allocation of items. It may not be possible to allocate each item at every warehouse. The expected answer to this question is that it may differ on the retailer’s items and the warehouse’s characteristics; these need to align to store an item at a location. Also, item characteristics may determine whether a shipment can be consolidated.

This paper’s remainder is structured as follows: Section 2 outlines the relevant literature. The research methodology is considered in Section 3. In Section 4, the framework is developed in collaboration with experts of a case company. The framework is implemented by the case company and evaluated in Section 5. Finally, Section 6 summarises this study’s results and discusses the limitations and recommendations for future research.

2. THEORETICAL BACKGROUND

This section is divided into three sub-sections discussing relevant literature on split shipments, multiple warehouses, and the allocation of items at multiple warehouses.

2.1 Split Shipments

(8)

8 warehouse, and (2) the stock level of the items is not sufficient. Large retailers may possess multiple warehouses because the capacity of a single warehouse is not adequate. When the items of a customer order are not stored at the same warehouse, they need to be sourced from multiple locations resulting in multiple shipments. This can be resolved by allocating items at multiple warehouses instead of a single warehouse. However, split shipments still occur if items are stored at multiple warehouse locations (Zhang, Huang, Hu, & Sun, 2017). Due to insufficient stock levels (i.e. out of stocks), a (part of) the customer order may be shipped from a different location or at a different time (Catalán & Fisher, 2012).

Customers prefer the delivery of their order in one package. Research found that customer value decreases when multiple shipments are received for the same order (Kärkkäinen, Ala‐Risku, & Holmström, 2003). They seek convenience in time and effort savings (Korgaonkar, Silverblatt, & Girard, 2006). As follows, customers preferably order all items at the same online retailer when engaging in online shopping. For the retailer split shipments result in inefficiencies in the (delivery) process. Besides multiple packaging materials and transports, it is also time inefficient. Multiple delivery appointment must be made, and various signatures for proof of delivery need to be obtained (Zhang, Sun, Hu, & Zhao, 2019). Also, more last-mile logistical movements are performed, which are considered the most expensive, inefficient, and polluting of the supply chain process (Gevaers et al., 2014).

Split shipments result in waste in the supply chain efficiency, high operational costs and customer dissatisfaction (Co, Miller, & Xu, 2007). Shipment consolidation can increase customer value and a decrease in the supply chain’s inefficiencies. Hence, it is of relevance for both the customer and the retailer. Since stock-outs are hard to predict and sometimes unavoidable, this study focuses on allocating stock at multiple warehouse locations.

2.2 Multiple Warehouses

(9)

9 Multiple warehouses imply more capacity, higher efficiency and more flexibility. The processing capacity of incoming and outgoing goods and stock storage of each warehouse is limited. Multiple warehouse locations accommodate division of capacity (Lee & Elsayed, 2005). If one warehouse is performing at its limits, this can be scaled down by allocating items at the other warehouse. Also, multiple warehouses result in higher efficiency. An example is a warehouse equipped with a special processing line for large items; if this line handles the large items, the other process lines will not be slowed down. Differently designed warehouses, in terms of e.g. processing systems and automatisation, will contribute to efficiency. Lastly, multiple warehouses make the retailer more flexible. Disruptions will have a less severe impact, and customer orders can be sourced from multiple locations (Schmitt et al., 2015; J Melis 2020, personal communication, 6 October).

The growth of e-commerce causes severe complications at the warehouse of online retailers in terms of capacity (Boysen et al., 2019). Adams (2020) found that retailers need more warehouses to store items. Customer demands are getting more diverse and extensive, and they seek for quick delivery windows. Multiple warehouse locations allow for acquiring higher inventories and more extensive assortment storage to fulfil customer needs. Capacity can be transferred, resulting in optimal usage, and warehouses are closer to the customer, resulting in shorter delivery times. Hence, multiple warehouses are beneficial and a strategic decision for the online retailer (Rouwenhorst et al., 2000). The online retailer benefits in capacity management, warehouse efficiency and flexibility by operating from multiple warehouses.

2.3 Allocation of Inventory at Multiple Warehouses

(10)

10 2.3.1 Item characteristics.

If an item’s storage or shipment requirements (e.g. temperatures) are not met, the item may no longer be useful, or its value is diminished (Nahmias, 2011; Coelho & Laporte, 2014). Items have characteristics that need to be considered when allocating them to a warehouse (Holzapfel et al., 2018). For example, consider books and meat, each having different characteristics. Meat needs to be stored at a specific temperature, and books can be put on any shelf. Not all items can be stored in every condition at every location.

Laws and regulations apply for the transport and storage of goods1. For example, perishable food, drugs and pharmaceuticals need to be transported and stored at a specific temperature. Some need to be kept sterile and in a secure place. Items containing hazardous chemicals or explosives need to be handled with care and held in specific storage. Moreover, in exceptional cases, items cannot be transported or stored next to other items. Violation of the rules and regulations is not an option. Besides these rules for shipments and storage, a retailer may also consider the items’ value and packaging (Ploos van Amstel & van Goor, 2003). The characteristics of an item (e.g. size, shape, substance, value) determines the ability to be shipped in combination with other items. For example, a customer orders a washing machine, a fragile glass vase, and a frozen pizza. According to the rules and regulations, these can be combined. However, splitting the order in multiple shipments would make more sense. How to deal with these considerations depends on the retailers’ preferences.

2.3.2 Warehouse design.

A retailer needs to decide on the warehouses’ design, e.g. layout and equipment (Rouwenhorst et al., 2000). These decisions determine the possibilities to allocate items at the warehouse. For example, if a retailer sells frozen food, the warehouse needs to be designed with cooling storage such that the temperature requirement is met. Alternatively, if a retailer sells hazardous chemicals, the warehouse should be equipped with a bunker. These are fixed requirements, based on the authorities’ rules and regulations, which are often related to item characteristics.

Multiple warehouse locations offer the possibility to design each warehouse differently and enlarge capacity (Lee & Elsayed, 2005). For each warehouse tactical decision on e.g.

(11)

11 storage systems, order picking systems, handling equipment and the level of automatisation must be made (Rouwenhorst et al., 2000). For example, if a retailer can equip one warehouse with an order picking system specially designed for smaller items, and the other warehouse with a system specially designed for larger items, the items will be processed at the highest efficiency. Also, the retailer can differentiate between manual or (semi-) automated warehouses. Manual warehouses are more effective in processing large and fragile items and (semi-) automated warehouses are more efficient in processing smaller-sized items in comparison to manual warehouses (Zaerpour, Volbeda, & Gharehgozli, 2019; J Melis 2020, personal communication, 6 October).

2.3.3 Supplier requirements.

It is most cost-efficient for the supplier to deliver all items at the same warehouse (Holzapfel et al., 2018). If a product is allocated at multiple warehouses, it needs to be supplied at multiple locations. This has a consequence for the supplier: the orders need to be split, multiple locations need to be fulfilled, and volumes per warehouse decrease. Shipment costs may increase due to unused transportation capacity, or the supply frequency decreases (Holzapfel et al., 2018). Minimum order quantities and values and full truckloads must be considered. For example, the MOQ of a supplier is 100. Initially, a retailer orders 120 items for one warehouse. After deciding on multiple warehouse allocation, the total quantity ordered is still 120 items, but 80 items are ordered for one warehouse and 40 for the other. If the suppliers’ requirements are not attained, the supplier will charge higher costs, supply less frequent or not deliver the items at all.

(12)

12

3. METHODOLOGY

This section describes the research methodology. In Section 3.1, the research design is determined, and Section 3.2 introduces the case company which was willing to cooperate in the development and evaluation of the framework. Section 3.3 describes how the data is collected.

3.1 Research Design

The aim of this study is to develop a framework that helps online retailers identify which items should (not) be allocated at multiple warehouses to reduce the number of split shipments. A design-based research method is applied to develop this support tool. A real-life practical problem initiates the need for this tool. The framework’s design and development is an iterative process that relates a practical problem with theory (Wang & Hannafin, 2005).

The research cycle is based on the Reeves model and is depicted in Figure 3.1 (Reeves, 2006). First, the practical problem is analysed, and the main variables are defined. Based on the analysis, a framework is constructed which is tested and refined in collaboration with experts. This is an iterative cycle: The experts are asked to give feedback on the framework, the researcher refines the framework based on the obtained feedback, then asks again for suggestions until both the researcher and the experts are satisfied with the completeness of the framework. In the final step of the research cycle, a case company or multiple case companies are needed to implement and evaluate the developed framework. The results and input can be used as input for the beginning of a new research cycle.

FIGURE 3.1

(13)

13

3.2 The Case Company

The case company willing to cooperate in the research, Company X2, is a large online retailer possessing multiple, differently designed, warehouses. As an indication for the size: the company has more than 10 million active customers. It is a representative company in the field of online retail. Company X has experienced the practical problem; many customer orders are split in multiple shipments because items cannot be sourced from the same warehouse. The company and its experts of the supply chain and logistics departments are interested in resolving the problem and willing to participate. In this study, the framework is developed within the company, and data is gathered for testing, implementation and evaluation.

3.3 Data Collection

The data collection can be divided into two phases: First, a qualitative approach is considered to investigate the practical problem, determine the main variables, and design and develop the framework. Secondly, the framework is implemented and evaluated by the case company. This is a combination of a quantitative and a qualitative analysis.

3.3.1 Phase 1: analyse the problem and develop the framework.

In the analysis of the practical problem, it is essential to consider the problem from different perspectives. Several interviews were conducted. Most interviews were unstructured (e.g. feedback and brainstorm sessions) and lasted 30 up to 60 minutes. During the interviews, the researcher took notes which were summarised and analysed afterwards. The goal of the interviews was to frame the problem and identify the variables used as a starting point for the design and development of the framework. In collaboration with the experts, the framework was developed and tested, several feedback and brainstorm sessions had taken place. The following questions were asked during the development and refinement:

• Does the framework follow a logical sequence?

• Does the framework include all relevant aspects of the process? • Do you have any suggestions for further development?

(14)

14 3.3.2 Phase 2: implement and evaluate the framework.

The framework was implemented and evaluated by the case company. The company provided access to customer order and shipment data of the year 2019. The dataset contained information on ten millions of orders. Data were collected, prepared, and analysed using SQL. SQL is a programming language used to access and manipulate relational databases.

For the evaluation, the framework and the results were distributed and presented. During the presentation the experts were asked to provide opinions and feedback on (1) the framework; (2) the results; (3) applicability, and (4) generalizability. Also, they were asked to propose suggestions for further development. The feedback and opinions are summarised.

4. FRAMEWORK DEVELOPMENT

The goal is to develop a generalisable framework: The problem of split shipments occurs in the online retail sector, so the support tool must be usable by any large online retailer that possess multiple warehouses. It is designed such that the values and requirements are generic and customisable. It is important to verify these with the retailer before implementing the framework since these may differ compared to other retailers.

(15)

15 The framework is divided into five steps:

1. Determine the economic viability of the framework implementation; 2. Determine which items are relevant for multiple warehouse allocation; 3. Assess and control for the internal process of the retailer;

4. Compute the potential costs savings;

5. Provide a suggestion for quantity division.

Step 1-3 determines whether it is interesting to allocate items at multiple warehouses, which items are relevant and which of these items can be allocated at the retailer’s warehouses. Step 4 and 5 are additional steps. Step 4 computes the costs that could have been saved if these items were allocated at multiple warehouses. Step 5 gives a primary suggestion for the quantity division of items at multiple locations. This will be of interest for the retailer if they decide to allocate the items at multiple locations. The steps are described in the following sections. An overview of the framework is outlined in Appendix A.

4.1 Step 1 – Assess Economic Viability

Before implementing the framework, the economic viability is assessed. Multiple warehouse allocation must be beneficial for the retailer. The framework’s potential is determined by analysing the effect of multiple warehouse allocation of one item over a selected time period t.

The item is selected based on (1) the number of customer sales, and (2) the likelihood of being combined. Usually, the selected item is one of the retailer’s most popular items. These are sold frequently over the time period and are likely to be combined. The item is determined in consultation with the retailer. The benefit of allocating this item at multiple warehouse locations is computed by determining how many shipments could have been combined, and how much could have been saved in terms of shipping costs. A rough estimate has been made for all the retailer’s items based on the single item analysis. The potentially saved costs per item for the rough estimate are 5% of costs computed for the single item to control the single

FIGURE 4.1

(16)

16 item’s popularity. For example, if the single item will save 2000 euros, the rough estimate is computed with a potential saving of 100 euros per item. The estimate is used to decide whether the implementation of the framework will be beneficial. It is important to determine what amount of potential shipping costs savings is sufficient for the retailer. The results of the economic viability assessment are evaluated together with the retailer. A go/no-go decision is made by the retailer on the implementation of the framework.

4.2 Step 2 – Determining Relevant Items

This study is interested in the items of the customer orders that are shipped in multiple packages, from multiple warehouses. A list of relevant items can be acquired through filtering. Figure 4.2 depicts Step 2 of the framework.

This study determines the relevant items according to the customer order data over the selected time period t. A distinction between a single-item order and a multi-item order must be made. In this study, multi-item orders (i.e. orders containing two or more items) are

(17)

17 considered since these orders indicate a combination of different items. Secondly, this study is interested in all multi-item orders that are consist of two or more shipments. This indicates whether the customer order is split in multiple shipments. However, this does not indicate that the order is sourced from multiple locations. A customer order that is split in multiple shipments can be shipped from the same warehouse due to e.g. stock-outs (Catalán & Fisher, 2012; Zhang et al., 2017). The multi-item multi-shipment order must be shipped from two or more warehouse locations to control for this event. The items in these multi-item orders, consisting of two or more shipments from at least two warehouses, are relevant for the multiple warehouse allocation.

However, a threshold (T) should be set to limit inefficient allocation at multiple warehouses. First, the number of combinations in multi-item multi-shipment orders per item must be depicted. Some items are likely combined more (or less) frequently compared to others, which will result in a long-tailed distribution (Anderson, 2006). An example of such distribution is depicted in Figure 4.3.

Pareto’s law determines the threshold. The distribution illustrates the 80/20 rule. The sum of the number of combinations in multi-item multi-shipment orders is computed. The top 20% of this is a sub-set of most frequently combined items and are expected to have the greatest effect (approximately 80%). The threshold is the cut-off point that discriminates the frequently combined items (20%; green) and the long tail items (80%; yellow). Resulting from this, the most relevant items for multiple warehouse allocation are determined.

FIGURE 4.3

Visualization of the Long Tail Distribution

(18)

18

4.3 Step 3 – The Internal Process

In Step 3 the framework controls for the internal process of the retailer. This step identifies filters to acquire the list of items that can be combined, stored, and shipped from multiple warehouses. The multiple warehouse allocation possibility of the items found in Step 2 is determined by (1) the item characteristics and (2) the warehouse design. Figure 4.4 depicts Step 3 of the framework.

4.3.1 Item characteristics.

It is important to consider the item’s storage and shipment requirements (Nahmias, 2011; Coelho & Laporte, 2014). This study found that it is important to control for two types of item-specific restrictions: (1) rules and regulations; and (2) retailers’ preferences. If an item cannot be combined based on either of the two restrictions (i.e. item’s characteristics do not allow for combination), allocating the item at multiple warehouses will not affect the number of split shipments because it will still be shipped individually. Only items that can be combined are of relevance.

4.3.2 Warehouse design.

Multiple warehouses are beneficial and strategic (Rouwenhorst et al., 2000). Each warehouse can be designed differently, so it is essential to consider the different warehouse designs. The warehouse design determines the ability to effectively process and store items (Zaerpour et al., 2019). The warehouse design and the item characteristics should align, e.g. frozen items must be stored in a warehouse with cooling storage. If one of the warehouses is not designed accordingly and is not equipped with the correct systems, the item cannot be allocated at that warehouse. Only the items that can be allocated at each warehouse are of relevance.

(19)

19

4.4 Step 4 – Compute Potential Savings

This study assumes that the number of split shipments will reduce by allocating items at multiple warehouse locations. This step is added to the framework to gather an idea of the potential costs savings of multiple warehouse allocation.

There are two types of costs: (1) the costs per item (ci), and (2) the costs per shipment

(cs). Figure 4.5 provides an overview.

The costs per item (ci) will differ per warehouse due to the different designs and systems (e.g.

manual or semi-automated processing systems). The difference in the costs between warehouses will be minimal (a matter of cents), and the costs must still be incurred when the item is allocated at the other warehouse. The shipping costs (cs) are assumed to be equal for all

warehouses and can be saved when shipments are combined. When items are allocated at multiple warehouses, and the customer order can be sourced such that the number of split shipments is reduced, the costs of an additional shipment is saved. The potential costs savings are computed as follows:

The total number of combinations in multi-item multi-shipments × costs per shipment

The total number of combinations in multi-item multi-shipment orders are considered since these orders could have been consolidated when the item was allocated at multiple warehouses. However, some shipments could have contained multiple items and thus allocating one item at another warehouse would not affect the number of split shipments. Therefore it is important to control for the number of items in the shipments. If 60% of the shipments contain one item; the re-allocation would have reduced 60% of the split shipments. Hence, the potential savings are 60% of the previously computed amount.

FIGURE 4.5

(20)

20

4.5 Step 5 – Suggestion for Quantity Division

The framework’s final step provides a suggestion for the quantity division of the items at multiple locations. This step considers the items that can be combined and stored at each warehouse (the items resulting from Step 3).

Initially, the items were stored at a single warehouse. According to the framework, some items should be stored at multiple locations. Such that the number of split shipments is reduced and costs are saved. For a suggestion on the quantity division, the total quantity sold over the time period is compared with the total quantity in the multi-item multi-shipment orders. The percentage of the quantity in multi-item multi-shipment orders should be allocated at the non-initial (i.e. the other) warehouse. For example, if 100 items are sold and 10 out of the 100 are in the multi-item multi-shipment orders. The percentage that should be allocated at the other warehouse(s) is 10%. Figure 4.6 visualises the suggestion.

5. IMPLEMENTATION AND EVALUATION

The developed framework is implemented by a case company in Section 5.1. Section 5.2 considers the results and the framework that are evaluated by the experts of the case company. An overview of the implementation can be found in Appendix B.

FIGURE 4.6

(21)

21

5.1 Implementation of the Framework

The framework is implemented at a case company, Company X. As described in Section 3.2, the case company is a large online retailer possessing multiple warehouses. The company provided access to following datasets: (1) customer order data of 2019, (2) shipment data, (3) item-specific data including the warehouse location, and (4) sales data.

5.1.1 Step 1 – Assess economic viability.

Before implementing the framework and determining which items are relevant for multiple warehouse allocation, the economic viability is assessed. The goal of the case company is to save at least 1 million euros by multiple warehouse allocation. The economic viability analysis was conducted by means of a single item; the potential costs savings of allocating this item at multiple warehouses were estimated. The item was selected in consultation with the case company. The item, a headphone, was one of the retailers’ the most popular items and sold more than 10.000 times in 2019. Because of the popularity and number of sales, it is assumed that the item was likely to be combined. Resulting from the analysis, the allocation of this item at multiple warehouses could have saved up to 4.000 euros. Accordingly, it was estimated that the company could save up to 2 million euros by implementing the framework (see Appendix C for the computations). This is more than the predetermined amount, so the case company gave us a ‘go’ for further implementation.

5.1.2 Step 2 – Determining relevant items.

This study considers all customer orders of 2019. After analysing the orders, quantities, warehouse locations and shipments, the number of multi-item multi-shipment orders (N3) in

(22)

22

FIGURE 5.1

Number of Combinations and Threshold

Figure 5.1 depicts the distribution of the number of combinations in multi-item multi-shipment orders per item. As expected, the distribution is long-tailed. The most popular item is combined 8500 times. The sub-set of 20% is composed of 680 items and are combined at least 1000 times. Expected is that these 680 items have the greatest effect, and thus are the most relevant items for multiple warehouse allocation.

5.1.3 Step 3 – The internal process.

In this step, the framework controls for the internal process of the retailer. It is assessed whether the 680 items can be combined, stored at and shipped from both warehouses. The item characteristics and the warehouse design need to be considered. Figure 5.2 outlines an overview of the restrictions of the case company.

The case company’s assortment does not contain items that cannot be combined based on rules and regulations. Nevertheless, the company does have preferences in terms of the size of an item. Large and extra-large products cannot be combined in a shipment. Therefore, this study only focusses on extra small, small and medium-sized items. This relates to warehouse design. Both warehouses are equipped with processing lines for extra small, small and medium-sized items. However, the warehouses are designed differently. One warehouse is equipped with a

FIGURE 5.2 Restrictions

Type Based on

Item-specific Size group Warehouse-specific Bunker

(23)

23 bunker for hazardous chemicals and explosive, the other warehouse is assembled with a high-value cage for sensitive-to-theft items. Besides special storages, processing systems are also important. Releases for example, these are kept on stock before it is available to the public. Customer can pre-order the item, but the retailer is not allowed to deliver it yet even though it is on stock. The processor must be designed such that the items are not shipped until the release date. The company has such system only in one warehouse. So the items with the corresponding characteristics can solely be allocated to a single warehouse. The warehouse design and item characteristics are interrelated. After controlling for the item characteristics and warehouse design, 550 items are relevant for multiple warehouse allocation.

5.1.4 Step 4 – Compute potential savings.

In this step, the potential cost savings are computed. The company’s shipping costs are 4 euros per shipment regardless of the warehouse it is shipped from. In total 550 items can be allocated at multiple warehouses. The savings are computed based on the number of item multi-shipment orders that could have been consolidated. The number of combinations in multi-item multi-shipment orders for the 550 items ranges from 1000 up to 8500 combinations. The sum of these combinations is 1 million. Hence, if the 550 items were allocated at multiple warehouse locations, 1 million orders would not have been split in multiple shipments. The potential costs savings can be computed as:

1 million orders × 4 euro = 4 million euro

However, this computation does not yet control for the number of items in the shipment. More than 70% of the shipments contained 1 item, meaning 70% of the shipments can be reduced by allocating the items at multiple warehouse locations. The potential costs savings will be 70% of 4 million euros. Allocating 550 items at multiple warehouse locations would have saved the case company approximately 2.8 million euros in 2019, and by the allocation of these items more than 15% of the split shipments are reduced.

5.1.5 Step 5 – Suggestion for quantity division.

(24)

24 division ranges from 4% up to 60%, but on average the quantity division is 20%. Meaning 20% of the total quantity sold were in a multi-item order shipped from multiple locations. If this 20% had been allocated at the not-initial (i.e. the other) warehouse, the number of split shipments would have been reduced because the items could have been sourced from multiple warehouse locations.

If the company decides not to allocate items at multiple locations, the following is of interest. Four items are combined with items of the other warehouse for more than 50% of the time. If these items were initially allocated at the other warehouse, a total of 25.000 euros in shipping costs would have been saved. It is recommended that the company re-evaluates the initial stock location of these items.

Overall it is recommended that the case company allocates the 550 items at multiple warehouses both warehouses. Of the total stock, 80% should be allocated at the initial warehouse, and 20% at the other warehouse. By allocating the 550 items at both warehouses, more than 15% of the split shipments could have been reduced, and potentially 2.8 million euros could have been saved in shipping costs. This is a starting point for multiple warehouse allocation; several iterations of the framework implementation need to occur for the optimal solution.

5.2 Evaluation of the Framework

In consultation with the case company, the framework and the results are evaluated. The experts were asked to give opinions and feedback on the framework, the results, the applicability and the generalizability. The feedback and suggestions for further development from the case company’s experts can be utilised as input a follow-up study.

(25)

25 The case company is delighted with the results. The potential costs savings was 0.8 million more than expected based on the economic viability assessment. However, the savings indicate some points for further development. The estimate savings of 2.8 million euros are solely savings, other costs e.g. supply and warehousing costs are not considered. The follow-up framework should consider costs for both the retailer as for the sfollow-upplier. Also, the scope of the framework is limited. Further development should incorporate the suppliers’ perspective and requirements. The experts suggest considering the initial stock location in further development. It is assumed that the initial location is not changed over the time period. However, some items were re-allocated in 2019.

The framework is a starting point for the allocation of items. Once the framework is implemented and items are re-allocated, the framework can be applied again to evaluate the split shipments orders and presumably other items are of relevance. This study outlines one consecutive research cycle (Reeves, 2006). For optimal allocation, the framework should be applied several times. This indicates a potential further research suggestion, in the future a mathematical optimisation model should be designed such that a single application will be sufficient.

The case company will utilise the framework and the results as input for new projects on multiple warehouses and stock allocation.

6. CONCLUSION

6.1 Summary

(26)

26 characteristics (e.g. cooling) the item cannot be allocated at that warehouse. This study finds that item characteristics and warehouse design are interrelated.

This study developed a framework that identifies which items need to be allocated at multiple warehouses to reduce the number of split shipments. A practical problem initially indicated the need for the development of the framework. Secondly, there was limited literature on the allocation of items at multiple warehouses and related support tools. A design-based approach was chosen to develop a support tool and relate the practical problem with theory. A case company implemented the developed framework. The results indicate that more than 15% of the split shipments could have been reduced by allocating 550 items at multiple warehouse locations. One of the recommendations for the case company is to allocate these 550 items at multiple locations with the following quantity division: 80% at the initial warehouse, and 20% at the non-initial (i.e. the other) warehouse. This is a starting point. For optimal allocation, several implementations need to take place.

6.2 Implications

There is limited research on the allocation of stock at multiple warehouses and related support tools. This study adds to this scarce research. The results outline the important variables that need to be considered in the allocation process, and a support tool is developed that follows a logical structure.

(27)

27 6.3 Limitations and Future Research

If the developed framework is used for future research, some limitations should be considered. This framework is developed in collaboration with a case company, so the variables considered are of interest to this specific company. However, it is questionable if other companies value these variables similarly. Besides, the framework is evaluated by the experts of the same company. This may result in biased evaluations and limits the generalizability of the framework.

(28)

28

8. REFERENCES

Adams, K. (2020, June 05). 12 ways the coronavirus will change warehousing. Retrieved from SHD Logistics:

https://www.shdlogistics.com/covid-19/12-ways-coronavirus-will-change-warehousing

Anderson, C. (2006). The long tail : why the future of business is selling less of more. Hyperion. Berman, O., Krass, D., & Tajbakhsh, M. (2012). A coordinated location-inventory model. European

Journal of Operational Research, 217(3), 500-508.

Boysen, N., de Koster, R., & Weidinger, F. (2019). Warehousing in the e-commerce era: A survey.

European Journal of Operational Research, 277(2), 396-411.

Catalán, A., & Fisher, M. (2012). Assortment Allocation to Distribution Centers to Minimize Split Customer Orders. SSRN Electronic Journal.

Co, H. C., Miller, R. H., & Xu, X. (2007). Clustering of skus to reduce split delivery cost and improve on-time delivery in online merchandising. California Journal of Operations Management,

6(1), 45-51.

Coelho, L. C., & Laporte, G. (2014). Optimal joint replenishment, delivery and inventory

management policies for perishable products. Computers & Operations Research, 47, 42-52. Farahani, R., Bajgan, H., Fahimnia, B., & Kaviani, M. (2015). Location-inventory problem in supply

chains: A modelling review. International Journal of Production Research, 53(12), 3769-3788.

Gevaers, R., Van de Voorde, E., & Vanelslander, T. (2014). Cost Modelling and Simulation of Last-mile Characteristics in an Innovative B2C Supply Chain Environment with Implications on Urban Areas and Cities. Procedia - Social and Behavioral Sciences, 125, 398-411.

Holzapfel, A., Kuhn, H., & Sternbeck, M. (2018). Product allocation to different types of distribution center in retail logistics networks. European Journal of Operational Research, 264(3), 948-966.

Kärkkäinen, M., Ala‐Risku, T., & Holmström, J. (2003). Increasing customer value and decreasing distribution costs with merge‐in‐transit. International Journal of Physical Distribution &

Logistics Management, 33(2), 132-148.

Korgaonkar, P., Silverblatt, R., & Girard, T. (2006). Online retailing, product classifications, and consumer preferences. Internet Research, 16(3), 267-288.

Lee, M., & Elsayed, E. (2005). Optimization of warehouse storage capacity under a dedicated storage policy. International Journal of Production Research, 43(9), 1785-1805.

Nahmias, S. (2011). Perishable Inventory Systems. Springer Publishing.

Ploos van Amstel, W., & van Goor, A. R. (2003). Distribution logistics and product characteristics. In A. R. van Goor, M. J. Ploos van Amstel, & W. Ploos van Amstel, European distribution and

supply chain logistics (pp. 107-137). Stenfert Kroese.

Reeves, T. (2006). Design research from a technology perspective. In J. van den Akker, K.

(29)

29 Ricker, F., & Kalakota, R. (1999). Order fulfillment: the hidden key to e-commerce success. Supply

chain management review, 11(3), 60-70.

Rouwenhorst, B., Reuter, B., Stockrahm, V., van Houtum, G., Mantel, R., & Zijm, W. (2000). Warehouse design and control: Framework and literature review. European Journal of

Operational Research, 122(3), 515-533.

Schmitt, A. J., Sun, S. A., Snyder, L. V., & Shen, Z.-J. M. (2015). Centralization versus

decentralization: Risk pooling, risk diversification, and supply chain disruptions. Omega, 52, 201-212.

Wang, F., & Hannafin, M. J. (2005). Design-based research and technology-enhanced learning environments. Educational Technology Research and Development, 53(4), 5-23.

Zaerpour, N., Volbeda, R., & Gharehgozli, A. (2019). Automated or manual storage systems: do throughput and storage capacity matter? INFOR: Information Systems and Operational

Research, 57(1), 99-120.

Zhang, Y., Huang, M., Hu, X., & Sun, L. (2017). Package consolidation approach to the split-order fulfillment problem of online supermarkets. Journal of the Operational Research Society,

69(1), 127-141.

(30)

30

9. APPENDIX

A. An Overview of the Framework

(31)

31

B. Application of the Framework

(32)

32

C. Economic Viability of the Case Company

In consultation with the case company, the following item is chosen: a headphone. In 2019 the headphone was ordered by 10.000 customers. For these orders, 11.000 packages were shipped. The number of packages does not align with the number of orders because some orders need to be shipped from multiple warehouse locations. Figure E.1 depicts a customer order that needs to be sourced from multiple warehouses. Out of the 11.000 packages, 8.000 packages contained only the headphone. Hence, 3.000 packages were of combined orders. Out of the 3.000 packages, 2.000 contained a headphone and were shipped from the headphone’s warehouse (WH1). An additional 100 packages were shipped from the headphone’s warehouse (WH1) due to stock-outs. From the other warehouse (WH2), the warehouse where the headphone is not stored, 900 packages are shipped. The costs of a shipment are 4 euros. The total shipping costs for this scenario is 12.000 euros.

FIGURE E.1

Customer Order Sourced From Multiple Warehouses

(33)

33

FIGURE E.2

Reduce Number of Shipments By Multiple Warehouse Allocation

By allocating the headphone at multiple warehouses, the number of shipments can be reduced by 33% and potentially 4.000 euros will be saved.

Go/no-go Decision

Based on the analysis of the headphone, a rough estimate is made for all the items of the retailer. Out of all retailer’s items, x%3_{of the items can be allocated at multiple warehouses, taking into} account the item characteristics and warehouse design. The split shipments of 10.000 items can be reduced. For the headphone, multiple warehouse allocation could have saved up to 4.000 euros. However, not all products are sold as frequent as the headphone. Thus fewer shipments can be reduced. Based on estimation, 5% of the headphone savings can be saved for each product, meaning 200 euros. The expected costs savings are 2.000.000 euros. Savings this amount of money means that the framework is economically viable for the case company. Resulting in a go: the retailer is willing to cooperate, provide data and implement the framework.

A FRAMEWORK FOR THE ALLOCATION OF ITEMS AT MULTIPLE WAREHOUSES