DEPARTMENT OF TECHNOLOGY MANAGEMENT AND ECONOMICS 

DIVISION OF INNOVATION AND R&D MANAGEMENT  

CHALMERS UNIVERSITY OF TECHNOLOGY 

Gothenburg, Sweden 2025 

www.chalmers.se

Dynamic Operational Policies and 

Performance Evaluation in Hybrid 

Warehouse Optimization 

A Case Study on Iterative Optimization in a Multi-Purpose 

Hybrid Warehouse 

Master’s thesis in Management and Economics of Innovation 

 

 

RYDIN FELIX 

WALLÉN EDVIN 



 

 
 

 
 
 
 

 

 

Dynamic Operational Policies and 

Performance Evaluation in Hybrid 

Warehouse Optimization 

 

A Case Study on Iterative Optimization in a Multi-

Purpose Hybrid Warehouse 

 

RYDIN FELIX 

WALLÉN EDVIN 

 

 

 

 

 

 

Department of Technology Management and Economics 

Division of Innovation and R&D Management 

CHALMERS UNIVERSITY OF TECHNOLOGY 

Gothenburg, Sweden 2025 



 

 

Dynamic Operational Policies and Performance Evaluation in Hybrid Warehouse 

Optimization 

A Case Study on Iterative Optimization in a Multi-Purpose Hybrid Warehouse 

FELIX RYDIN 

EDVIN WALLÉN 

 

 

© FELIX RYDIN, 2025. 

© EDVIN WALLÉN, 2025. 

 

 

 

Department of Technology Management and Economics 

Chalmers University of Technology 

SE-412 96 Gothenburg 

Sweden 

Telephone + 46 (0)31-772 1000 

 

 

 

 

 

 

 

 

 

Cover: 

An aisle of a general conventional warehouse 

Gothenburg, Sweden 2025



 

I 

Dynamic Operational Policies and Performance Evaluation in Hybrid Warehouse 

Optimization 

A Case Study on Iterative Optimization in a Multi-Purpose Hybrid Warehouse 

FELIX RYDIN 

EDVIN WALLÉN 

Department of Technology Management and Economics 

Chalmers University of Technology 

Abstract 

Since warehousing contributes to more than 25% of the supply chain’s total cost and 

largely impacts its efficiency, the importance of optimized and efficient warehouse 

operations are evident. Growing demand, without increasing operational costs, requires 

dynamically integrated optimizations operations which can be used iteratively to 

counter fluctuations in order frequencies. While these challenges are investigated in 

existing literature, there remains a gap on how to perform this in a multiple purpose 

warehouse that is using a conventional warehouse complemented by automated storage 

solutions. Additionally, existing cases mostly discuss one-time solutions, while this 

thesis aims to address how to successfully iterate the optimization process.  

This master’s thesis combines quantitative data analysis with qualitative interviews and 

document analyses to illustrate the current challenges and objectives companies face in 

their warehouses. Different operational policies and performance evaluations methods 

and metrics are discussed to decrease the risk of re-shuffling warehouses without 

increasing their effectiveness. Similarly, different levels of decision making were 

involved in exploring divergences in challenges and objectives. 

The study’s findings suggest a class-based approach separating the conventional and 

automated parts of the warehouse as well as the different product purposes. The 

automated systems are found to be most effective, an approach utilizing volume and 

order appearance frequency is therefore used to assign the right products to the 

automated system. Additionally, by combining slotting, routing and picking policies 

the rest of the conventional warehouse’s locations are classified. Order appearance 

frequency was then used to assign the products to the different classes, ensuring 

minimal travel time and increased efficiency. The performance evaluation metrics and 

methods provide a foundation for decision makers to know when to iterate the 

warehouse optimization. By integrating data instead of using an ad hoc strategy it is 

calculated that the warehouse could save up to 39% on their order picking.   

Keywords: dynamic warehouse optimization, operational warehouse policies, performance 

evaluation methods and metrics, warehouse slotting policy, semi-automated warehouses. 



 

II 

 
 

 
  



 

III 

Acknowledgements  

The master’s thesis was conducted during the spring of 2025 at the Department of 

Technology Management and Economics, under the division of Entrepreneurship and 

Strategy at Chalmers University of Technology. The project was carried out in 

collaboration with a case company within the fire hazard and safety industry.  

Firstly, we would like to express our gratitude towards Kaj Sunesson at Chalmers 

University of Technology. By having the role as both our academic supervisor and 

examiner you have guided us through both practical and theoretical problems and 

challenges throughout the project. Your knowledge and supervision have been very 

appreciated when structuring and planning this thesis, as well as how to execute and 

finish it. Your feedback has ensured a greatly enhanced quality of our report.  

Secondly, we want to thank our case company for giving us the opportunity to put our 

theoretical knowledge into practice. We extend our sincere thanks to our supervisor 

who not only provided expertise and knowledge but also support and direction which 

successfully shaped and concretized the project from start to end. The time and 

resources provided ensured the completion of this thesis.  

Thirdly, we wish to thank all interviewees who shared their invaluable knowledge and 

experience with us throughout the project’s continuation. Your commitment made it 

possible for us to illustrate the scope and scale of the challenges, making your 

participation crucial for completing this project. 

Lastly, we wish to thank our thesis opponents, Wictor Arthur and Arvid Stenberg, for 

critical and constructive feedback. Your outside perspective and thorough suggestions 

provided us with knowledge that strengthened the final quality of the thesis. We are 

grateful for the time and resources you invested in reviewing our report. 

 

 

 

 

 

 



 

 

IV 

Table of Contents

 

List of Figures VI 

List of Tables VII 

Terms and Definitions VIII 

Abbreviations VIII 

Glossary VIII 

1. Introduction 1 

1.1 Background 1 

1.2. Purpose 2 

1.3 Research Questions 2 

2. Theoretical Framework 3 

2.1 Warehouse Optimization 3 

2.2 Review of Theoretical Frameworks for Warehouse Optimization 3 

2.2.1 Views on Warehouse Optimization 3 

2.2.2 Operational Policies 4 

2.2.3 Conventional Compared to Automated Warehouse Optimization 7 

2.2.4 Performance Evaluation 7 

2.3 Decision making process 8 

2.4 Project Specific Theoretical Framework 9 

2.5 Relevant Gaps in Existing Literature 10 

3. Method 12 

3.1 Case Study 12 

3.1.1 Introduction to Case 13 

3.1.2 Case Study in Relation to Research Questions 14 

3.2 Research Strategy 15 

3.2.1 Research Quality 15 

3.3 Research Process 16 

3.4 Data Collection Methods 19 

3.4.1 Literature 19 

3.4.2 Interviews 20 

3.4.3 Document Analysis 21 

3.4.4 Quantitative Data 21 

3.5 Ethical Considerations 22 

3.6 Use of AI 23 

4. Empirical Data 24 

4.1 Technical Structure 24 

4.2 Current layout of the AS/RSs 25 

4.3 Warehouse Management System 27 

4.4 Current Operational Warehouse Policies 27 

4.4.1 Inventory Management 28 

4.4.2 Slotting Policy 29 

4.4.3 Picking and Routing Policy 29 

4.5 Performance Metrics and Evaluation methods 31 

4.6 Current Decision-Making Process 31 

4.6.1 Strategic Decision-making and Planning 31 

4.6.2 Operational Decision Making 32 

4.7 Current Order Flow of the Conventional Warehouse 32 



 

V 

5. Analysis 34 

5.1 Performance Evaluation 34 

5.1.1 Performance Evaluation Metrics in Conventional Warehouses 34 

5.1.2 Performance Evaluation Metrics in Automated Warehouses 36 

5.1.3 Creating method from metrics 37 

5.2 Warehouse optimization 38 

5.2.1 Inventory Management 38 

5.2.2 Picking Strategy 39 

5.2.3 Slotting Policy 40 

5.2.3.1 Multipurpose Consideration 41 

5.2.3.2 Assigning Products to the AS/RSs 41 

5.2.3.3 Conventional W01-Classes 44 

5.2.3.4 Conventional W02-Classes 46 

5.2.4 Routing Policy and Warehouse Areas 46 

5.2.5 Replenishment and Buffer slots 48 

5.2.6 Quantitative Improvements for W01 49 

5.2.7 Quantitative Improvements for W02 51 

5.3 Decision-making Process 52 

5.3.1 Operational Level 53 

5.3.2 Tactical Level 53 

5.3.3 Strategic Level 53 

5.3.3.1 Continuous reshuffling 54 

5.3.3.2 Periodical reshuffling 54 

5.3.3.3 Reshuffling based on triggers 55 

5.3.3.4 Compilation of re-optimization approaches 55 

5.3.3.5 Reshuffling method for specific case 56 

6.1 Evaluations Methods and Performance Metrics 58 

6.2 Warehouse optimization method 60 

6.3 Decision-making 62 

6.3.1 Which decisions are relevant? 62 

6.3.2 Who is responsible for which decisions? 63 

6.3.3 When should decisions be taken and changed? 64 

6.3.4 How does our recommendation impact decision making? 65 

6.4 Generalizability and Contextualization 66 

7. Conclusion and Recommendation 68 

Reference List 69 

Appendix A - Interviews I 

Appendix B – Sample of SKU Data and Calculations IV 

Appendix C – Example Calculation VI 

Appendix D – Optimization Model Summary VII 

Appendix E – Descriptive Statistics VIII 

Appendix F - Balancing of AS/RS IX 

Appendix G - Quantitative Improvements Calculations X 

Appendix H – Dataset Access XIII 

Appendix I - Division of Labor XIV 

 

 



 

VI 

List of Figures 

2.1 The division of picking an order 

2.2 Theoretical framework identifying different parts of the research frontier  

3.1 Theoretical framework highlighting relevant parts for this project 

3.2 The research process 

4.1 Map of current warehouse layout, equipment and interaction with delineated systems 

4.2 Example of AS/RSs layout  

4.3 The operational process  

4.4 The different order types  

4.5 The median time between different racks within one order  

4.6 the percentage of order lines going from one rack to another 

5.1 How the conventional warehouse optimization can be divided into smaller parts 

5.2 The slotting policy framework 

5.3 A potential class based slotting policy for the case company 

5.4 Potential classes overlaid on the layout  

5.5 Routes for proposed classes 

5.6 Number of SKUs in each W01-Class 

5.7 Number of order lines within each W01-Class 

5.8 Complete orders within the different classes 

5.9 Number of order lines for complete orders within the different classes 

5.10 SKUs within the different W02-classes 

5.11 Number of order lines between the different W02 classes  

6.1 The cyclical nature of decision making based on all three level



 

VII 

List of Tables 

2.1. Questions aimed to be answered by each decision level 

3.1 Data needed and relevant data collection method 

3.2. Interviews conducted with employees at the company 

3.3. Received quantitative data 

4.1 Number of lines per lift 

4.2 Number of orders per lift 

4.3 Number of SKUs per lift 

5.1 Relations between operational policies and performance metrics 

5.2. Picking time per order line from lower and higher shelves 

5.3. SKUs and order lines for each category 

5.4. SKUs assigned to AR/RSs and the conventional storage for both subgroups 

5.5. SKUs assigned to each class in W01 

5.6. Number of SKUs and number of picking slots assigned to each class  

5.7. Time to pick one order line under the condition that the order is picked only in that class 

5.8. Advantages and drawbacks of different reshuffling methods 

5.9. Percentage of SKUs changing class a month 

6.1. How the performance of locations and SKUs are evaluated

 

 

  



 

VIII 

Terms and Definitions 

This section includes abbreviations of academia and industry specific abbreviations as well as 

terminology connected to the research area at hand. Moreover, company specific words and 

concepts will be explained to facilitate the comprehension of the report. Lastly, the text written 

in italics shows the Swedish word that has been translated.  

Abbreviations 

KPI: Key performance indicator 

RQ1.1: Research question 1.1 

RQ1.2: Research question 1.2 

RQ2: Research question 2 

SKU: Stock keeping unit  

ERP-system: Enterprise resource planning system 

WMS: Warehouse Management System 

EUR-pallet: A type of pallet with the dimensions … 

SLAP: Storage location assignment problem 

AS/RS: Automated storage and retrieval system 

Glossary

AS/RS: A collection term to describe automated storage and retrieval solutions. An automatic 

rack within the warehouse that makes it possible to store a lot of smaller SKUs in a small area 

by leveraging height. Several different versions exist, the case company's AS/RS works like a 

Ferris wheel where storage shelves, filled with SKUs, rotate on a fixed axle.  

Delivery Specialist: The workers in the warehouse performing picking, packing, and 

reshuffling amongst other similar tasks.  



 

  1 

 

1. Introduction 

The introduction aims to describe the research study’s topic and introduce the purpose of the 

research. After that, three research questions are proposed to fulfill the purpose of the project.  

1.1 Background 

As an integral part of the supply chain, warehouses can accommodate variability caused by 

change in demand or supply from external factors (Gu et al., 2007). Throughout the last decades 

warehouse management has become increasingly crucial, constituting about 25% of a 

company's total logistics cost (Baker and Canessa, 2009). Growing demand for shorter lead 

times has resulted in increased investments in warehouses and increased complexity between 

its operations (Gu et al., 2010).  

To reduce complexity, the concept of warehouse optimization can be broken down into smaller 

parts. Karásek (2013) breaks warehouse design down in three different parts: technical 

structure, warehouse management and operational policies. Similarly, these parts can be broken 

down into even smaller pieces which can highlight deficiencies and potential for optimization. 

In this way the warehouse managers can detect challenges in physical areas, assets like layout 

or system, or in work policies like picking and routing (Gu et al., 2007). Naturally some of 

these aspects are more difficult to change and have larger financial impact than others 

(Rouwenhorst et al., 2000). Focusing on aspects that are easier to change but still can contribute 

to large improvements (e.g. operational policies) is therefore particularly interesting. 

The importance of warehouse optimization has resulted in multiple analytical studies; however, 

most studies are delineated to optimize a specific problem (Gu et al., 2010). For instance, 

existing literature on warehouse optimization predominantly focuses on single-purpose 

warehouse environments, such as distribution or manufacturing settings, and often assumes 

homogeneous automation levels — either fully automated or fully manual operations (Freitas 

et al., 2019). This simplification neglects the integration challenges of hybrid automation 

systems and the co-existence of multiple product purposes within a single warehouse — a 

situation frequently encountered in practice. 

Lastly, warehouse optimization processes are destined to become inefficient due to changing 

conditions (Gu et al., 2010; Karásek, 2013). The warehouse optimization process therefore 

must be made dynamic and iterative for future re-optimization, which is often neglected in the 

current literature. Warehouse re-optimization furthermore implies the importance of 

performance evaluation to determine when to reoptimize.  Gialos and Zeimpekis (2024) 

highlight the need for integrated performance evaluation in the operational policies of 

warehouse optimization. The performance evaluation should not only consider which metrics 

are relevant, but also which methods are useful to attain those metrics. 



 

  2 

 

1.2. Purpose 

This paper aims to contribute to the research frontier by developing a dynamic process to 

facilitate decision-making in the operational policies of warehouse optimization, given a 

warehouse combining conventional and automated storage and multiple product purposes. 

This includes investigating which method and variables that relevantly evaluate the 

operational policies’ performance.  

By integrating performance evaluation into warehouse optimization, the study aims to bridge 

the existing theoretical gap between performance evaluation and operational policies. By 

investigating warehouse optimization when conventional and automated systems are 

combined, as well when warehouses serve multiple product purposes, this study aims to 

broaden the existing knowledge of how warehouse optimization can be adapted to complex 

real world scenarios. 

1.3 Research Questions 

There is a gap within the existing theory on warehouse optimization for an dynamic operational 

policies decision making process that considers multiple product purposes and a combination 

of automated and conventional storage systems. Additionally, in many warehouses the 

alignment between performance evaluation and decision-making is skewed and not integrated. 

Given this and the project’s purpose three research questions have been formulated to structure 

the research. 

RQ1.1. What performance evaluation methods and metrics are most effective for assessing the 

impact of a warehouse optimization? 

RQ1.2. How can a method for operational warehouse policy optimization be constructed that 

considers a combination of conventional and automated storage, and multiple product 

purposes? 

RQ2. How can an integrated decision-making process be designed for optimizing the 

operational policies of a warehouse, considering changing conditions? 

 

  



 

  3 

 

2. Theoretical Framework  

The theoretical framework aims to break down the concept of warehouse optimization and find 

academic gaps in the existing literature. The chapter begins broadly, describing concepts from 

warehouse optimization, and then gets successively narrowed down to the research frontier. 

2.1 Warehouse Optimization  

A warehouse can broadly be defined as material handling stations dedicated to receiving, 

storing and shipping goods (Dotoli et al., 2015). Generally, warehouses can then be divided in 

three categories, distribution warehouses, production warehouses and contract warehouses.   

Warehouse optimization aims to achieve optimal warehouse operations (Karásek, 2013). 

Optimal warehouse operations are achieved when each customer is completely satisfied with 

the order, the due time and all warehouse processes are done in the shortest possible time, with 

minimal cost and optimal utilization of resources under dynamic changing conditions.  The 

general process in a warehouse is defined as receiving, storing, put-away, picking/retrieving 

and shipping goods (Karásek, 2013). 

2.2 Review of Theoretical Frameworks for Warehouse Optimization 

Warehouse design and warehouse optimization are complex issues (Karasek, 2013. Gu et al., 

2010, Rouwenhorst et al., 2000). Many issues to investigate imply the need to break down the 

subject. There are several different ways to break down the subject into different components, 

therefore also different frameworks to conduct warehouse optimization. 

2.2.1 Views on Warehouse Optimization 

Karásek (2013) divides the warehouse optimization problem in three different subgroups. 

Firstly, optimization of the technical structure of a warehouse. The technical structure involves 

the layout design of the warehouse, the choice and dimensions of equipment and the design of 

physical interfaces with neighboring systems. Secondly, optimization of the operational 

policies, which include the inventory management policy, picking policy, routing policy and 

slotting policy. Lastly, optimization of the management system. The warehouse management 

system is used for the coordination and control of all the typical warehouse operations. 

Gu et al. (2010) similarly describes how warehouse design can be divided into five key areas. 

Firstly, the overall structure which defines material flow, department specifications, and inter-

departmental relationships. Secondly the sizing and dimensioning of the warehouse, which 

defines overall warehouse size and space allocation. Thirdly, the department layout focuses on 

the configuration of individual departments within the warehouse. For example, aisle 



 

  4 

 

orientation or automated storage and retrieval systems. Fourthly, the equipment selection, for 

example level of automation, selection of storage and handling equipment. Lastly the 

operational policies act as a framework for the day-to-day work in the warehouse. Once 

selected, these operational policies have an important impact on the overall system and are 

likely not to change often. Gu et al. (2010) mainly mentions storage and order picking policy 

as important examples of operational policies.    

Rouwenhorst et al., (2000) initially proposes three key warehouse characteristics: The 

processes, the resources and the organization. The processes are defined as receiving, storage, 

order picking and shipping. Resources include storage units, systems, equipment and 

personnel. Organization describes how, for example policies are organized for order picking 

or storing and how tasks are assigned. Rouwenhorst et al., (2000) then proposes a warehouse 

design framework based on the characteristics. The framework divides the design process into 

three different levels: the strategic level, tactical level and the operational level. The strategic 

level is concerned with long-term decisions with significant financial impact. For example, 

process flow design, degree of automation and choice of equipment. The tactical level is 

concerned with medium term decisions. Tactical decisions typically concern the dimensions of 

resources. For example, storage system sizes, number of employees, the determination of a 

layout and several organizational issues. Lastly the operational level describes how day-to-day 

decisions will be handled within the constraints set by the higher level (Rouwenhorst et al., 

2000). The main decisions at this level concern assignment and control of people and 

equipment. For example, the allocation of stock keeping units (SKUs) and the allocation of 

tasks.  

2.2.2 Operational Policies 

The operational policies are vital for a successful optimization (Gu et al., 2010; Karásek, 2013; 

Rouwenhorst et al., 2000). The first main part of the warehouse operational policies is 

inventory management, which involves external processes like purchasing and distribution of 

goods, but also how the warehouse is operated internally, which goods that should be brought 

in and how much of those goods are handled. (Karásek, 2013). 

When goods arrive at the warehouse it needs to be stored away. How and where goods are 

stored are important later when retrieving the goods by order picking (De Koster et al., 2007). 

The storage assignment method or slotting policy is the second main part of the warehouse 

operational policies and involves a set of rules used of which products to assign to which 

location. De Koster et al. (2007) synthesis five main methods of storage assignment: random 

storage, closest open location storage, dedicated storage, full turnover storage and class-based 

storage. With a random policy every incoming product is assigned a random location of all the 

open locations with an equal probability. The random assignment results in a high space 

utilization but longer order picking times. The closest open location policy describes that each 

incoming product is assigned to the closest available open space (De Koster et al., 2007). This 



 

  5 

 

method arguably performs the same as the random storage but with a higher concentration of 

products in the front of the warehouse and a lower concentration at the back.  

A dedicated storage policy on the other hand involves that each product has an assigned slot 

(De Koster et al., 2007). A dedicated policy might decrease space utilization but allows workers 

to become familiar with product locations, and therefore easier retrieval. Bahrami et al. (2019) 

mentions four main ways to prioritize products in a dedicated storage policy. Products can for 

example be ranked based on turnover, the ratio between retrieval operations and space 

requirement, duration of stay or correlation.  

A full turnover storage policy describes that products are distributed over the storage area 

according to their turnover (De Koster et al., 2007). The products with the highest sale rates 

are located at the most convenient places. Whilst full turnover storage allows for efficient 

picking it is sensitive to varying demand rates and changing product assortment. Changes are 

problematic in full turnover storage as each change involves a large amount of reshuffling.  

A class-based storage involves grouping products according to different characteristics (De 

Koster et al., 2007). For example, popularity. Each class is then assigned an area in the 

warehouse where allocation is random within the area. Karásek (2013) extends on the class-

based storage by mentioning that it can be stored either by factors like picking frequency or 

using family grouping where similar products are clustered close to increase efficiency.   

There are different views on which slotting policy is more favorable. In theory a dedicated 

storage policy is most promising in reducing travel time (Bahrami et al., 2019). Some others 

do however consider the dedicated storage policy practically impossible due to the extensive 

need of accurate data, continuous supervision and limited capability to cope with ceaseless 

changes. Aase et al. (2004) found that a class-based storage policy with three classes 

contributes to an improvement in overall picking efficiency 90% as great as dedicated storage 

over a random storage. Overall, class-based storage showed a 12-26% saving in picking time 

compared to a random storage, depending on number of classes and picking list length. 

Furthermore, class partition affects the performance of a class-based storage. With two classes 

Aase et al (2004) found that a 30-70 or 40-60 percentage partition yields the best results. A 30-

70 partition describes that the primary class consists of the 30% most important SKUs, and the 

secondary class consists of the remaining 70% of SKUs. Additionally, different storage 

implications and routing policies impacted the effectiveness of a class-based storage system. 

Lastly, SKU demand distribution also affects the performance of class-based policy. A demand 

distribution where a smaller percentage of the products account for a larger percentage of the 

total demand implies that a class-based storage can contribute to a larger improvement.  

Moreover, there is debate over the optimal number of classes. Aase et al (2004) investigates a 

class-based storage with two to four classes. Four classes contributed to the largest relative 

improvements compared to a dedicated storage policy. Bahrami et al. (2019) however 



 

  6 

 

highlights that recent research has found a traversal relationship between the number of classes 

and the average picking time. The optimal number of classes was found to be between three to 

eight. 

The picking policy is the third main part of operational warehouse policies and manages how 

orders are picked within the warehouse (Gu et al., 2010). Three different policies are brought 

up: wave picking, batch picking, and zone picking. Wave picking entails picking a fraction of 

the daily orders within the corresponding fraction of the day, while batch picking involves 

picking multiple orders in one trip. Zone picking means dividing the warehouse into different 

zones where different workers operate in different zones. Gu et al. (2010) states that the 

effectiveness of each picking policy is dependent on the specific case it is applied to. 

Conversely, Karásek (2013) highlights how the batching policy is superior concerning 

efficiency, while the zoning policy more effectively handles congestion in smaller warehouses. 

The main drawback for the zoning policy is the consolidation of the order after it has been 

picked by multiple warehouse operators.  

Rouwenhorst et al. (2000) elaborate further on the operational policies involving the personnel 

and equipment. Regarding personnel, the authors suggest task allocation based on priorities 

and location, similar to the aforementioned zoning policy. Where picking is dependent on 

equipment, for example with multiple story racks or handling of heavy products the dwell point 

policy is crucial for efficient operation. The dwell point policy concerns where to place the idle 

order picking equipment to minimize travel distance and consequently order picking time 

(Rouwenhorst et al., 2000). Additionally, Gialos and Zeimpekis (2024) differentiate the 

concepts of low-level and high-level picking systems, emphasizing the challenge of slotting 

SKUs three dimensionally. In manual storage operations where equipment like forklifts is used 

it is therefore important to decide not only where to place the SKU in walking distance from 

the consolidation space, but also regarding the vertical distance. Gialos and Zeimpekis (2024) 

advise that heavy SKUs should be placed on the low-level, while high frequency SKUs belong 

on the medium level. The importance of considering the ergonomics for the warehouse 

operators is also stressed by the authors.  

The fourth main part of the warehouse operational policies is routing policy (Karásek (2013). 

There are a few different policy options on how to pick the optimal travel path by choosing a 

routing policy. The policies brought up are S-shape, return policy, mid-point policy, largest 

gap policy and optimal routing. The policies show different ways a delivery specialist can 

travel between different slots in the warehouse to pick a complete order. The different policies 

are efficient in different ways depending on the technical structure of the warehouse, and some 

are dependent on big data sets, while others utilize the routine and experience of the delivery 

specialists (Karásek, 2013). In a multilevel warehouse the routing must be considered three 

dimensionally and it is of importance to have forklifts or similar tools placed within a given 

route to avoid deviance from the route and increased travel distance (Karásek, 2013). 



 

  7 

 

Lastly, Bahrami et al. (2019) highlights that there is a significant statistical correlation between 

storage policy, picking policy and routing policy. A model integration of all parts of warehouse 

operational policies is therefore beneficial.  Gu et al. (2010) extends this by emphasizing the 

positive synergies that may emerge by considering all parts. An effective slotting policy entails 

an effective routing policy and vice versa.  

2.2.3 Conventional Compared to Automated Warehouse Optimization 

Automated storage and retrieval systems (AS/RS) are commonly added to a warehouse's 

technical structure to increase efficiency (Karasék, 2009). The usage of AS/RS has several 

advantages compared to non-automated systems (Roodbergen and Iris, 2009). For example, 

savings in labor costs and floorspace or reduced error rates. However, automated systems are 

often costly and give less flexibility.   

Gao et al. (2017) argues that many automated warehouses still utilize a manual or random 

slotting policy. Likewise, as in a conventional warehouse these methods do not guarantee high 

utilization, especially when more types of goods are present, and the number of goods is high. 

By optimizing slotting policy in automated storages travel time can be minimized by reducing 

travel time for cranes/shuttles, throughput can be increased by reducing congestion, storage 

density can be increased by placing items more efficiently and energy consumption can be 

decreased by optimizing movement patterns.  

Roodbergen and Iris, (2009) presents several challenges to address in automated storage 

optimization. Likewise, as in a conventional storage setup slotting policy, picking policy and 

routing policy needs to be considered. The policies to handle these problems are mostly similar 

to those in a conventional warehouse setup. Overall items that need to be accessed frequently 

need to be placed close to the input/output point to increase efficiency. Slotting policy can 

therefore be dedicated, random, full turnover based, or class based. Regarding picking and 

routing several heuristics can be used. However, aspects such as elevator or crane travel time 

have to be considered as well as warehouse worker travel time.   

2.2.4 Performance Evaluation  

The concept of performance evaluation is an important subject in warehouse optimization since 

it measures the effectiveness of a warehouse design and thereby whether a warehouse design 

optimization is successful or not (Gu et al., 2010). Gialos and Zeimpekis (2024) highlight the 

importance of utilizing effective methods and variables to evaluate and benchmark the 

operational policies. While the methods mainly concern how to optimize the operational 

policies, the variables act as KPIs. 

The methods used are mostly mathematical and analytical models relying on algorithms (Gu 

et al., 2010; Karásek, 2013), but other methods include simulation, field and lab tests, and 

surveys and interviews (Gialos and Zeimpekis, 2024). Each of the methods has its advantages 



 

  8 

 

and disadvantages for specific occasions and KPIs like price, scope and validity must be 

evaluated before deciding method. Gialos and Zeimpekis (2024) highlight that combining 

different methods can better illustrate the warehouse and facilitate decision-making.    

The variables are in this case the KPIs which the methods aim to improve. The most common 

one regarding operational policies and picking optimization is order picking time (Gialos and 

Zeimpekis, 2024), but other KPIs like cost efficiency, throughput time, space utilization, speed 

to fulfill order, order accuracy can also be useful in benchmarking the methods performance 

(Gu et al., 2010; Rouwenhorst et al., 2000). Tompkins et al. (2003) displays how the order 

picking time can be divided, illustrating how time is the most important metric, but that many 

other metrics also impact the overall efficiency. The distribution of time can be seen in figure 

2.1. 

 
Figure 2.1. The division of picking an order 

 

2.3 Decision making process 

Snartland (2023) brings up three decision levels: strategic, tactical and operational. The 

strategic decisions are those causing the most comprehensive, long-term impact and are usually 

made by higher executives within the company. Tactical decisions are instead made by mid 

managers and concerns medium term like ordering quantities throughout a period. Lastly, the 

operational level concerns the delivery specialists working in the warehouse who are making 

daily decisions like allocations of products and picking of orders (Snartland, 2023). Each level 

is crucial for optimizing the warehouse and their focus can be seen in table 2.1. 

Decision level  Questions aimed to be answered 

Operational How to execute? 



 

  9 

 

Tactical  Where and how? 

Strategic Why and when? 

Table 2.1. Questions aimed to be answered by each decision level  

Aligning with the different levels of decision making, Chen et al. (2010) emphasize the 

importance of choosing to focus on the right policies to create positive synergies. Managers in 

the tactical and strategic level need to predetermine which policies that are of relevance and 

treat them together rather than isolated. Chen et al. (2010) explains that this is due to the close 

interrelation between the different policies such as slotting, routing and picking. If they are 

treated separately positive and negative effects can be neglected. It is therefore argued that all 

these policies should be integrated into one decision making problem, contrary to existing 

literature that treat them all separately despite their close linkage.   

2.4 Project Specific Theoretical Framework 

The frameworks presented above (Gu et al., 2010; Karásek, 2013; Rouwenhorst et al., 2000),  

have several similarities but also complement each other in other ways, this allows for them to 

be synthesized to create a framework that is mutually exclusive and collectively exhaustive. In 

figure 2.2 the three frameworks described above are synthesized to create a general framework 

used in this research. Firstly, the technological structure is delineated. This is based on 

Karásek's (2013) description but complemented by the first four key areas described by Gu et 

al. (2010). Most decisions associated with the designing of the technical structure falls under 

the strategic level of Rouwenhorst et al., (2000) due to the high cost and effort to change any 

of these variables.  

The operational policies is the second delineation. The operational policies involve setting 

policies for the day-to-day operations for a warehouse (Gu et al., 2010; Karásek, 2013). The 

operational policies are in-turn broken down into inventory management, picking policy, 

routing policy and slotting policy. Defining the operational policies mainly concerns the 

tactical level, since the effort to formulate and change these policies implies that it cannot be 

done too often. However, the operational policies also outline how work is carried out on the 

operational level.  

The warehouse management is also delineated and describes the system used to control, 

coordinate and manage the warehouse (Karásek, 2013). The warehouse management system is 

concerned with for example monitoring orders, locations and employees. Overall, the 

warehouse management system enables effective operations and supports day-to-day activity. 

Changes in the management system can however be complex.  



 

  10 

 

In addition, the performance evaluation criterion from Gu et al. (2010), and Gialos and 

Zeimpekis (2024) are integrated into the framework to evaluate the different variables and 

methods within each of the other three subsections. Performing evaluation is not only important 

in the operational work, but also in warehouse design, since a good performance evaluation 

model enables the correct decision regarding the best warehouse design (Gu et al. 2010).  

 

Figure 2.2. Theoretical framework identifying different parts of the research frontier 

2.5 Relevant Gaps in Existing Literature 

Existing academic research has primarily focused on analytical models for specific problems 

rather than integrated design approaches (Gu et al., 2010). There is therefore a need for 

development of integrated analytical and simulation models for better decision-making. There 

is also limited exploration of dynamic and uncertain operating environments (Gu et al., 2010). 

Overall, there are insufficient tools for aiding real-time decision-making during warehouse 

design. 

Rouwenhorst et al. (2000) also highlights a significant focus on isolated problems (e.g., routing, 

batch picking) rather than integrated, holistic approaches in the current research. The focus also 

tends to be on automated systems rather than conventional warehouses. Overall, there is a need 

for more design related methodologies that consider interdependencies between design, 

operations, and performance. Lastly, (Rouwenhorst, et al., 2000) explains that there is also a 

need for more case-based studies to align research with real-world challenges.  

Kofler et al., (2014) also highlights that operational policies regularly have to be remodeled 

due to change in company characteristics like growth or new product launches. The need to 

reoptimize also implies the need to research more dynamic processes and to evaluate the 



 

  11 

 

performance. Gialos and Zeimpekis (2024) elaborate on this and emphasize how the lack of 

considering multiple indicators, in addition to time, makes the methods more static and non-

optimal. Order picking time is the chosen evaluation variable in 56% of the cases, while KPIs 

like number of warehouse operators, ergonomics and costs are evaluated in less than 7% of the 

cases (Gialos and Zeimpekis, 2024).  

Lastly Bahrami et al. (2019) highlights a lack of balance in the research area of warehouse 

optimization between assumption restricted modeling approaches and those based on the 

complex reality of warehouses. These views can be connected to the lack of studies on hybrid 

warehouses that combine automated and conventional storage systems (Freitas et al., 2019). 

Freitas et al. (2019) also highlights the lack of studies on multi-activity warehouses with 

diverse activities. For example, warehouses that combine supply to production lines and supply 

to after-sales.  

Combined knowledge from the existing literature has led to the identification of the following 

research needs within the area warehouse optimization: 

● Need to investigate integrated warehouse optimization processes combining design, 

operations and performance evaluation. 

● Need to study integrated operational policies, rather than separately to leverage 

synergies. 

● Need of research on how warehouse optimization processes can be dynamic and handle 

changing conditions. 

● Further need for case studies aligning research with real world problems.  

● Further need for studies on the complex reality of warehouses rather than assumption-

based modelling approaches. 

● Further need for studies on hybrid and multiple-activity warehouses. 



 

  12 

 

3. Method 

The following section describes how the project will be conducted and why certain methods 

are selected. The section begins by outlining the overall project design, then describing the 

overall research design and lastly handling how and why different methods are used. 

3.1 Case Study 

Case study research is concerned with the complexity and nature of a particular case in question 

(Bell et al., 2022). A case study also implies that the system or situation being studied has a 

definite scope. In this case the definite scope is a single location, as a single warehouse is being 

investigated. The warehouse is meant to be used as a representative or typical case that seeks 

to exemplify how the research can be used in similar situations (Bell et al., 2022). The case 

was selected through purposive sampling. Purposive sampling is when researchers 

intentionally select a case subject based on specific characteristics relevant to the study (Bell 

et al., 2022). Purposive sampling is beneficial in this case since it allows for the selection of a 

particularly interesting subject for the study.  

A case study is favorable in this research project due to several different reasons. Firstly, there 

is an expressed need for case studies within the area of warehouse optimization, to align 

research with real world problems (Rouwenhorst et al., 2000). Secondly a case study facilitates 

the exploration of complex real-world business issues in depth (Bell et al., 2022). Through the 

study of a case, deep and detailed insights can therefore be acquired from a real-world context, 

which is especially relevant when the focus is to generate real world applications rather than 

theoretical models. A case study is therefore favorable to investigate RQ2 as RQ2 is concerned 

with generating a real-world applicable decision-making process for warehouse optimization. 

Thirdly, a case study is favorable to use in qualitative research since it facilitates the use of 

several qualitative methods (e.g. interviews and observations) and therefore credibility. 

Additionally, case studies facilitate theory which to some degree can be generalized to other 

situations and therefore create transferability. 

Bell et al. (2022) emphasize how a combination of qualitative and quantitative research are 

common in case studies to create an in depth understanding. This is applicable to this research 

paper’s case which will be researched using both qualitative and quantitative measures. 

Additionally, it has been chosen to keep the case company anonymous. This is done to ensure 

a more authentic and honest study where the participants dare to highlight challenges and 

problems, thus increasing the reliability of collected data (Godfrey-Faussett, 2022).   



 

  13 

 

3.1.1 Introduction to Case 

The case company is a global provider of fire, flame and gas safety solutions to marine, energy, 

rolling stock and building sectors. Although the company has several locations this study will 

be limited to one. The case company is currently in a high growth phase. This has implied 

increased strain on the warehouse operations. The case company therefore has a need to either 

expand the warehouse operations through additional hires or ensuring more efficient 

operations. The decision to conduct a warehouse efficiency optimization was supported by the 

management team's hypothesis that the warehouse could be organized more efficiently. 

Additionally recent acquisition of barcode readers and their use in the warehouse has ensured 

a strong availability of data and therefore increased the feasibility of a warehouse efficiency 

optimization project. The high growth has also implied limited time to handle the warehouse 

optimization internally, which is why this degree project will be carried out.  

Lastly, due to rigidity and high costs the company seeks a process which optimizes the 

operational policies rather than improving on the technical structure and warehouse 

management. However, the company does not seek a one-time solution, but rather a dynamic 

process that can be iterated in the future based on performance evaluation. The warehouse setup 

of the company is also suitable to answer RQ1.2 since the company deploys a mix of 

conventional storage shelves and AS/RS and has several product purposes as the warehouse 

both distributes finished products to customers, supply production and handle aftermarket 

sales. Figure 3.1 displays our project specific theoretical framework, and which parts are 

relevant for the case.  

 

Figure 3.1. Theoretical framework highlighting relevant parts for this project  



 

  14 

 

3.1.2 Case Study in Relation to Research Questions 

The specific case study performed at the case company is as established before purposely 

chosen to contribute answers and insights to the two formulated research questions. 

RQ1.1. What performance evaluation methods and metrics are most effective for assessing the 

impact of a warehouse optimization? 

The literature highlights the need for research on the integration of performance evaluation 

with warehouse management. Moreover, the literature calls for research either evaluating other 

KPIs than just order picking time or evaluating multiple indicators in relation to each other. 

Similarly, the most prevalently used performance evaluation method is heavily reliant on 

advanced analytical models. Due to their recent introduction of digital tools, the case company 

lacks the capability to use these models and instead must opt for another method. Similarly, 

while picking time is still an important KPI, the need for assessing other performance metrics 

has been implied by the company. This enables the researchers to assess which methods and 

metrics are most effective for optimizing their operational warehouse policies. Furthermore, 

the case company has raised concern that the warehouse might need to be re-optimized in the 

future; there is therefore a need to investigate how performance metrics can be used to diagnose 

the warehouse optimization need and show the result. 

RQ1.2. How can a method for operational warehouse optimization be constructed that 

considers a combination of conventional and automated storage, and multiple product 

purposes? 

Similarly, the existing literature fails to acknowledge multiple purposes and hybrid warehouse 

solutions. Therefore, the objective is to use the metrics and methods found in RQ1.1 to 

construct a method that can optimize these types of warehouses. The case company suits this 

objective well since they utilize conventional and automated storage within their warehouse. 

The case company highlights the inefficiencies between their storages, which makes it highly 

relevant as a foundation for answering this RQ. Similarly, the case company’s warehouse has 

multiple purposes. Firstly, the delivery specialists pick sales orders which are directly shipped 

to their customers, either as new sales or as aftermarket spare parts. Secondly, they pick raw 

material orders which are sent to the company’s production department. It is therefore a 

combination of a distribution and production warehouse.   

RQ2. How can an integrated decision-making process be designed for optimizing the 

operational policies of a warehouse, considering changing conditions? 

The case company is currently going through rapid growth and to manage the higher demand 

on their warehouses they have decided to start using digital tools to make their warehouse 

operations more efficient and accurate. Similarly to the identified gaps in the research frontier, 

the case company has lacked an effective and dynamic decision-making process deciding how 



 

  15 

 

and when to iterate their warehouse optimization due to the changing conditions. The digital 

tools were implemented in March 2024, which gives the researchers a unique possibility to 

investigate and design the decision-making processes. In addition, the case company is using 

a hybrid warehouse which is specifically mentioned in RQ2 due to the gap in existing literature.  

3.2 Research Strategy  

The research strategy is based on a deductive approach. A deductive approach begins with 

theory or hypothesis and then tests this through observation, data collection, and analysis, 

working from general to specific (Bell et al., 2022). The deductive approach is suitable in 

answering the research questions since it facilitates the testing of different theories related to 

the subject, and then evaluating their effectiveness using quantifiable metrics and systematic 

evaluation methods. 

Moreover, the research strategy requires a mix of quantitative and qualitative elements to 

answer the research questions. Quantitative research is based on the collection of numerical 

data and qualitative research is based on data compromising of written or spoken words and 

images (Bell et al., 2022). The societal view on the study is that it is continuously changing 

due to human interaction and changes in demand. The study therefore follows a constructivist 

view (Bell et al., 2022). In accordance with Bell et al. (2022) it is therefore important to include 

qualitative research to make the study more dynamic.   

3.2.1 Research Quality 

Qualitative research is sometimes targeted with criticism regarding lack of objectivity and 

being too influenced by the researcher (Bell et al., 2022). Quantitative research on the other 

hand needs to consider questions such as sample stability and context accuracy. To address 

these critics and to enable research quality, different quality criteria has been developed. Bell 

et al. (2022) mentions the quality criteria of reliability, replicability and validity. These are 

however mainly applicable in quantitative research due to the criteria gearing towards numbers. 

To address these concerns the quality criteria of credibility, transferability, dependability and 

confirmability have been developed as an overall assessment of the research trustworthiness 

can be achieved. Since these incorporate the criteria of reliability, replicability and validity, 

trustworthiness will be used to secure the research quality of this project.  

Credibility describes how internal validity will be achieved, which describes how believable 

the findings are (Bell et al., 2022). Many different perceptions of the social world imply many 

different views that have to be incorporated. The ability to interpret findings and include them 

correctly is important to ensure credibility. To ensure credibility in this project interviewees 

will have the opportunity to read the research before publishing to ensure correct interpretation. 

Theory included will also be analyzed by both authors to ensure the common understanding. 

Lastly, to ensure internal validity in quantitative data a strong cause-to-effect relationship 



 

  16 

 

between variables need to be identified. Furthermore, an adequate large sample size will have 

to be used as well as tools to analyze statistical reliability.  

Transferability describes how well the findings apply to other contexts, or external validity 

(Bell et al., 2022). To ensure transferability in the research a rich description of details of 

culture will be made. This will provide others with a database to determine if the research is 

transferable to other contexts.  

Dependability describes the stability and consistency of research findings over time and under 

similar conditions, and therefore parallels reliability (Bell et al., 2022). To account for this 

criterion the paper aims to be as well-structured and transparent as possible. Methodological 

changes and how results are obtained will be well documented to facilitate the readers 

evaluation of the paper.   

Confirmability concerns the objectivity of the research, therefore describing if the investigator 

has allowed their values to intrude on the project to a high degree (Bell et al., 2022). To ensure 

confirmability a detailed record of decisions, methods, and processes followed during the 

research will be maintained. The concept of triangularity will be applied when possible, using 

multiple data sources, researchers, or methods to cross-check findings. Lastly external audits 

are present, both in the form of a peer review at the end of the project and continuous checks 

by the project examiner.  

3.3 Research Process 

The project will be carried out in five different phases to investigate the research questions in 

a structured way. The complete research process is illustrated in Figure 3.2. The different 

phases consist of actions taken to answer the research question.   

 

Figure 3.2. The research process 

In the first phase, relevant literature about warehouse optimization will be reviewed and 

synthesized to create a theoretical framework. The theoretical framework will provide an 

essential foundation for answering the research questions through defining concepts and 

integrating prior research. The theoretical framework will therefore provide the overall 

structure that answers the research questions.  



 

  17 

 

The second phase involves a current state analysis of the case warehouse. Warehouse 

optimization is concerned with the improvement of an existing system, mapping the existing 

system is therefore important as a starting point. Furthermore, profound understanding of 

existing prerequisites and interaction with bordering systems are required to develop the 

integrated decision-making process described in RQ2. In table 3.1 data required to investigate 

the research questions at the case company is listed. There is also a summary of how the data 

will be collected and to which RQ the data is relevant for.   

Type of Data Explanation  Collection method RQ relevance 

Warehouse layout  How is the 

warehouse currently 

organised?  

Document study and 

observations 

RQ1.1, RQ1.2 

and RQ2 

Equipment 

specification 

What is currently 

used in the 

warehouse? (E.g 

forklifts and racks) 

Interviews RQ1.1, RQ1.2 

and RQ2 

Process 

specifications 

What processes are 

there currently and 

how are they carried 

out? (E.g picking 

pattern, restocking) 

Interviews and 

document study 

   

RQ1.1, RQ1.2 

and RQ2 

Inventory 

management policy 

How is inventory 

currently managed? 

(E.g stock keeping 

policy) 

Interview and document 

study  

RQ1.1, RQ1.2 

and RQ2 

Current product 

placements  

Which products are 

there? Where and 

how are products 

currently stored? 

Quantitative data 

extraction from the case 

company's ERP-system. 

RQ1.1 and 

RQ1.2 

Picking times  What time is 

required to pick the 

different specific 

products? 

Quantitative data 

extraction from the case 

company's ERP-system. 

RQ1.1 and 

RQ1.2 

Picking frequencies  How often are 

products used? 

Quantitative data 

extraction from the case 

company's IT-system. 

RQ1.1 and 

RQ1.2 



 

  18 

 

Picking type Picking type for 

different product 

locations - Can the 

product be retrieved 

manually by a 

worker or is a 

forklift needed? 

Interview, document 

study and oservations 

RQ1.1, RQ1.2 

and RQ2 

Perceived 

bottlenecks and 

inefficiencies  

What are the current 

perceived 

bottlenecks and 

inefficiencies? 

Interviews RQ2 

 

 

Current evaluation 

methods and 

performance 

metrics.  

How is evaluation 

currently made? 

Interviews RQ1.1 

Current decision 

making processes 

How are decisions 

currently made? 

Based on what and 

how often?  

Interviews RQ2 

Table 3.1 Data needed and relevant data collection method 

Data for RQ1.1 is collected with a focus on performance evaluation. The quantitative data is 

collected from the company's IT-system and will be used to develop a method to assess which 

performance metrics are most effective in evaluating a company’s warehouse optimization. 

The metrics found in RQ1.1 will then be used in combination with other quantitative data and 

documents to answer RQ1.2. Documented data, like transaction data, current warehouse layout 

and how the case company currently handles orders, is gathered to structure an optimization 

approach, i.e. RQ1.2.  

Data for RQ2 is collected to more broadly comprehend how they currently optimize their 

warehouse and how the decision-making process works. This is mainly qualitative, and the 

current state analysis aims to give foundation for an improved decision-making process.  

The third phase concerns optimization of the operational warehouse policies, which entails 

identifying improvements to the current operational warehouse policies. For RQ2 this will be 

done mainly through interviews and discussions with the staff to detect deficiencies and 

bottlenecks which can be improved. This does not only concern the delivery specialists, but 

also other decision-makers. For RQ1.1 and RQ1.2 this phase includes consolidation and testing 

of quantitative data to improve their current method for optimizing operational policies. In 



 

  19 

 

RQ1.2 the results from RQ1.1 are used to simulate different operational policies, and which 

one gives the optimal solution. The third phase thereby describes how operational warehouse 

policy optimization can be carried out and which evaluation methods and performance metrics 

are most suitable.  

The fourth phase involves contextualizing the study. RQ2 describes that the goal with this 

research is an integrated decision-making process. Therefore, the results of the current state 

analysis and the improvements analysis will be synthesized to create an iterative, integrated 

decision-making process. Similarly, for RQ1.1, the two previous phases will be combined to 

understand which method and metrics are important to assess to improve on the existing system 

continuously. Lastly, for RQ1.2 the results from RQ1.1 in combination with the data from the 

previous two phases will be used to propose a warehouse optimization method. This method 

aims to be generally applicable to similar companies to the case company.   

The last phase is to conclude the findings in a recommendation that can be generalized to firms 

like the case. The recommendation aims to be dynamic and facilitate the decision-making 

process by addressing how and when to reoptimize the warehouse regarding relevant 

performance evaluation methods and metrics.   

3.4 Data Collection Methods  

As stated above both qualitative and quantitative data collection will be carried out. Several 

different methods will be deployed to acquire sufficient data to answer the research questions. 

3.4.1 Literature  

To develop the research methodology and theoretical framework of this project, literature has 

been used. Literature used mainly includes academic articles and books. To construct the 

research methodology and to ensure research quality the book “Business research methods” by 

Bell et al., (2022) has been used.  For the theoretical framework literature has been sourced 

and selected through Google Scholar and Chalmers library. Relevant search words for the 

scope have been used such as “warehouse optimization”, “warehouse slotting”, “slotting policy 

optimization”, “warehouse optimization management” and “warehouse operational policies”.  

To construct a theoretical framework, it is important to critically evaluate literature to know 

what to include in the literature review (Bell et al., 2022). The literature is therefore also 

evaluated based on the trustworthiness criteria. The theoretical framework based on existing 

literature is used as the foundation for this project’s purpose.  



 

  20 

 

3.4.2 Interviews 

Interviews will be conducted throughout the research project to gain valuable insight from 

delivery specialists and decision makers within the case company. The interviews will be a 

combination of semi-structured interviews, with pre-prepared questions, and more loose 

discussions with different employees at the case company. The semi-structured interviews are 

advantages since they allow the interviewer to deviate from the planned questions to delve 

deeper or understand the reasoning from the interviewee, while simultaneously creating an 

agenda of what the interviewer wants answered (Bell et al., 2022). These types of interviews 

facilitate the early process since they narrow the scope and clearly illustrates the case. The 

looser interviews are instead advantageous in the latter part of the study since they allow two-

way communication where both the interviewer and interviewee can ask questions and propose 

ideas. This combines theory and data analysis from the project with experience and routine 

from the employees, resulting in a more comprehensive optimization. The subject of discussion 

of the interviews as well as questions from the semi structured interviews can be found in 

Appendix A.   

The choice of interviewees will be made using snowball sampling, where initial interviewees 

will be asked to refer to further participants with relevant experiences and characteristics (Bell 

et al., 2022). The snowball sampling method has been criticized for being biased and not 

representing the whole population and it is therefore important for the researchers to initiate 

contact with employees from various levels in the organization to create a more holistic view.   

The data gathered from the interviews is presented in table 3.1 above. The overall aim with the 

interviews is to complement the quantitative research performed and qualitatively research the 

current state of the warehouse operation policy and gain insights from delivery specialists. 

With whom and when the interviews were conducted are documented in table 3.2, and more 

information can be found Appendix A. All interviewees are important for answering all our 

RQ’s and illustrating the operational work and decision-making process in different parts of 

the warehouse hierarchy.  

 

Role Department Date 

Manager Delivery Warehouse February 6th 2025 (Semi-

structured) 

VP Supply Chain Finance February 4th 2025  

(Semi structured) 

March 13th 2025  

(Loose structure)  



 

  21 

 

Head of Supply Chain Operational Management 

team 

February 11th 2025  

(Semi structured) 

March 18th  

(Loose structure) 

Delivery Specialist Warehouse March 20th 2025 (Semi-

structured) 

April 15th 2025  

(Loose structure) 

Table 3.2. Interviews conducted with employees at the company 

3.4.3 Document Analysis  

The warehouse layout, current process specifications and current inventory management policy 

are already mapped and documented by the case company. To leverage the data in this report 

a document analysis therefore must be carried out. Access has been granted to these documents. 

The document analysis helps map the current warehouse situation, which is important to 

answer RQ1.1 and RQ1.2 since it sets the prerequisites for the warehouse optimization process. 

However, care must also be taken to ensure that the document specifications also correspond 

to reality.  

3.4.4 Quantitative Data 

The data used in the process optimization has, as established before, been collected since March 

2024 and thereby provides a large enough sampling population to objectively illustrate the 

current processes in the warehouse. To increase the accuracy further, the median, rather than 

mean, picking time will be used since it better represents the typical performance by being less 

affected by extreme values (Smoteks, 2024). 

Access has been granted to data regarding the company’s SKUs and their orders. This data and 

quantitative approach are used mostly to answer RQ1.1 and RQ1.2 since it illustrates their 

current warehouse policies, and which metrics are stored and used to evaluate its performance.  

However, the quantitative data collected can also be fundamental for the decision-making 

process by identifying shifts and trends in the changing environment, thus also being useful to 

complement the more qualitative research in RQ2. 

Data regarding the warehouse and its SKUs has been extracted from the case company’s ERP-

system in the form of six different spreadsheets containing requested data from the researchers. 

The spreadsheets contents are described in depth in table 3.3. 



 

  22 

 

Name Description 

Inventory on hand with current 

location 

Contains all SKUs and where in the warehouse they are 

located. Additionally, the current available inventory of 

each SKU can be observed.   

Warehouse location and profile This spreadsheet describes how the different racks 

currently are being used, e.g if they are meant to be picked 

from or if they are used as buffer places.  

Work line details Raw data This spreadsheet contains every order since the 

introduction of the hand held units. Each order, include 

number of lines, order of lines, time to perform line, SKUs 

picked, quantity picked, and from which shelf the SKUs 

got picked.  

Picking place with picking 

height 

This data classifies if each location in the warehouse can 

be reached when picking from the floor or if a forklift has 

to be used. 

Utlagringsdata W01 Weland 

AUTO1 

This data highlights each order in the AS/RS called 

AUTO1 since May 2024. It includes SKUs and their 

location within the AS/RS. Since AUTO1 contains two 

lifts this also highlights in which lift the SKU is located. 

Utlagringsdata W02 Weland 

AUTO2 

This data highlights each order in the AS/RS named 

AUTO2 since May 2024. It includes SKUs and their 

location within the AS/RS. 

Table 3.3. Received quantitative data 

3.5 Ethical Considerations  

When involving other people in the report through interviews and surveys it is crucial to 

consider ethics and potential implications. The ethics of business research can be broken down 

into four main principles: harm to participants, lack of informed consent, invasion of privacy, 

and deception (Diener and Crandall, 1978). To minimize the risk of harming research 

participants these ethical principles will be taken into consideration. Before the interviews the 

participants will be well informed about the research’s purpose to decide whether to participate 

or not and to avoid complications the participants will remain anonymous. Moreover, it is 

important for the researchers to present their intentions clearly for the participants. Each 

participant should be handled with respect and their preferences should be accounted for to not 



 

  23 

 

invade their privacy. Lastly, to not deceive the participants the results of the study should not 

be used for any other purpose than what has been communicated to participants before.  

3.6 Use of AI 

Careful considerations have been made of the use of AI in this research. AI tools were utilized 

to gain an understanding of the subject and related fields, as well as to get an overview of which 

articles were important and how the field has developed over time. Additionally, AI has been 

used to summarize and bring forward important concepts from articles, thus showing their 

relevance to this project. AI has not been used to generate any written material nor any other 

related material used in this report. Lastly, AI has been used to facilitate programming in 

Python.     

  



 

  24 

 

4. Empirical Data 

The following chapter utilizes the previously constructed theoretical framework to outline the 

current warehouse design, decision making process and current benchmarking procedure. 

Outlining the current warehouse state and decision-making process is important in answering 

RQ2 and RQ1.2 since it determines the prerequisites for an improved operational warehouse 

process, thereby providing context for the improved decision-making process. Similarly, 

investigating the current benchmarking procedure is important to answer RQ1.1 since needs 

and organizational context determine how to build an effective evaluation method.   

4.1 Technical Structure 

As Karasek (2013) describes, the technical structure of a warehouse has a significant impact 

on the other warehouse optimization tasks. Understanding fixed prerequisites of the technical 

structure is therefore beneficial since it frames the decision-making process in what can be 

dynamically changed and what has to remain fixed.  

Figure 4.1 illustrates the technical structure, i.e. layout, of the case warehouse. The figure is 

based on documentation from the case company and observations made by the authors but has 

been simplified for anonymity. Although simplified, it still preserves the key elements of the 

original data. The different letter symbolizes different racks. The placement of the racks is as 

established before fixed, but the SKUs’ slots are changeable to optimize the operational work. 

The company has two stand-up trucks which are utilized to reach SKUs on the higher levels. 

The racks have two to eight different levels. How many levels can be reached whilst picking 

from the floor varies from zero to three. The different slots are named according to letter 

number letter code (e.g. A1A). Each slot (e.g. A1A) can hold up to three EUR-pallets, and 

several SKUs are often consolidated on one EUR-pallet. The slots can either be full height or 

half height. Other important remarks are that the A and S rack is currently used for finished 

goods and spare parts respectively. Additionally, AUTO1 and AUTO2 are AS/RSs, used for 

handling smaller SKUs. Lastly, it is important to highlight that the A rack is built differently 

than the other racks and does not fit EUR-pallets. Instead, it is built to fit larger cardboard 

boxes of finished goods.    



 

  25 

 

 

Figure 4.1. Map of the current warehouse layout, equipment and interaction with delineated systems 

4.2 Current layout of the AS/RSs 

The automated part of the warehouse consists of two towers, where one is concerned with spare 

parts (AUTO2) and the other to raw materials and sales orders (AUTO1). AUTO1 consists of 

two lifts, called H1 and H2, which are meant to work together to facilitate the picking of orders. 

The objective is that when the delivery specialists retrieve SKUs from one lift, the other lift 

should automatically bring forward the shelf with other SKUs to be picked in that order. The 

two lifts have a total of 112 shelves which can be customized to hold specific sizes and 

quantities of SKUs. Figure 4.2 shows how these may look and are based on observational and 

documented studies made during the project.  



 

  26 

 

 

Figure 4.2. Example of AS/RS layout 

For maximum efficiency the two lifts should be balanced, which means that the same number 

of lines should be picked. Table 4.1 shows how many order lines that are picked from each of 

the lifts. Every unique SKU in an order has its own order line, e.g. an order with three different 

SKUs has three order lines. Table 4.2 displays the total number of orders that make use of each 

lift or both. Lastly, the number of SKUs in each lift is shown in table 4.3. Together the tables 

highlight how AUTO1 has been operating since May 2024.  

  Number of order lines Percentage of order lines 

H1 20864 24,4% 

H2 64718 75,6% 

Table 4.1 Number of lines per lift 

  Number of orders Percentage of orders 

Both 5186 36,5% 

Only H1 3434 24,2% 

Only H2 5576 39,3% 

Table 4.2 Number of orders per lift 

 

 

 

 



 

  27 

 

  Number of SKUs Percentage of SKUs in Auto1 

Both 9 0,8% 

Only H1 491 43,1% 

Only H2 639 56,1% 

Table 4.3 Number of SKUs per lift 

The head of supply chain (Personal communication: March 18, 2025) explains that the goal 

earlier was to place SKUs that often occurred in the same order on the same shelf to create a 

prepared product assembly slot. However, this proved to demand a lot of time and resources to 

maintain and has not worked accordingly. Consequently, this has caused a lot of idle spaces 

within AS/RSs that are occupied by either low frequency goods or completely empty.   

4.3 Warehouse Management System 

The warehouse management system (WMS) is used for coordination and control of warehouse 

activities (Karasek, 2013). Understanding of how the WMS works is therefore important in 

understanding how flow of people, machines and goods is handled. Understanding the WMS 

helps answer RQ2 since the WMS needs to be integrated in a dynamic decision-making 

process. Lastly the WMS is used to gather data, the WMS therefore sets the prerequisites for 

which evaluation methods and metrics can be used, which is an important aspect to be handled 

in RQ1.1 and RQ1.2. 

The vice president of Supply Chain (Personal communication: February 4, 2025) at the case 

company explains that the case company uses an ERP-system to keep track of their order data. 

The data used in this project is gathered from that system. Handheld units with barcode readers 

are used in the daily work and interact with the ERP-system. Power BI is used to forecast and 

plan future demand and number of work hours needed to complete predicted orders.  

4.4 Current Operational Warehouse Policies 

The optimization of operational policies of a warehouse consists of four main areas: Inventory 

management, slotting policy, picking policy and routing policy (Karasek, 2013). All these areas 

are important to consider in RQ1.1. RQ1.2 and RQ2 since they are parts of the warehouse 

processes. The case company’s current warehouse process is illustrated in figure 4.3 with the 

relevant areas of operational warehouse policy highlighted. The process is taken from the 

company’s documentation but is simplified and connected to this study’s theoretical 

framework to remain comprehensive and anonymous.  



 

  28 

 

 

Figure 4.3. The Operational warehouse process 

4.4.1 Inventory Management 

The purchasing department is responsible for inventory management, i.e.  replenishment of 

materials and stock keeping levels (Manager delivery, personal communication, February 6, 

2025). Decisions are therefore mostly made outside the warehousing department’s control in 

the case company, forcing the warehousing department to adapt. However, some policies are 

deployed that directly affect the warehousing situation. Firstly, a FIFO policy is employed. 

This implies that if a product has two dedicated places goods should be picked from where the 

product was first place. The FIFO policy can imply inefficiencies, as goods are not always 

retrieved from the most convenient place. Secondly besides dedicated product places there are 

also buffer slots. Buffer slots are used to store excessive volume of products when an 

extraordinary amount arrives.  

In addition, the case company divides their products into two different abstract warehouses. 

W01, W02. Although everything is picked from the same physical warehouse, W01 concerns 

the products picked for sales orders and to production, while W02 concerns spare parts. Each 

of the warehouses contains both a conventional part and an AS/RS. By separating them from 

each other the case company can put different prices on the same products depending on which 

warehouse it is in. Spare parts are sold with higher margin (VP of supply chain, Personal 



 

  29 

 

communication: March 13, 2025). W02 is replenished by picking SKUs from W01 and 

transferring them to W02 both digitally and physically, there is therefore no need for different 

buffer slots for the same SKU.  

4.4.2 Slotting Policy 

The manager delivery (personal communication, February 6, 2025) describes that the case 

company currently employs a fixed position warehouse slotting policy. Each SKU is assigned 

one or more slots in the warehouse. It is desirable that the SKUs are assigned to one slot close 

to the floor where products can be conveniently retrieved, and one or more less convenient 

slots where excess storage is located. That is, however, not always the case. Additionally, if 

the SKU is small and picked frequently, it often also has a slot within one of the two AS/RSs. 

Although, one current drawback is that when one SKU has been placed in the AS/RSs once it 

usually stays there and keeps its place despite it not being optimal (Delivery specialist, personal 

communication: April 15, 2025). When new goods arrive, they are therefore often located in 

buffer slots and replenishment of picking slots is often done from buffer slots. Assigned slots 

are recorded in the ERP-system and slot names are marked in the warehouse. The current 

system therefore represents neither of the five different slotting policies described by De Koster 

et al. (2007) but is rather a combination of several methods. 

Determining a product's fixed locations is currently ad hoc and experience based (Manager 

delivery, personal communication, February 6, 2025). Implying that a product's allocated slot 

is changed when a delivery specialist raises concern that it would be more efficient to change 

the slots. There are, however, some main guidelines in the slotting policy. Raw material is most 

commonly assigned to the AS/RSs or another particular shelf, due to their high pick frequency 

and small size.  

4.4.3 Picking and Routing Policy  

Orders are released to the warehouse operators throughout the day (Manager delivery, personal 

communication, February 6, 2025). Picking in the warehouse is done by order. However, the 

orders that require picking from both the conventional and automated storage are divided into 

two different orders, which later are combined when it is time to pack the order. To increase 

efficiency SKUs from the bottom levels are picked first and when that is done stand-up trucks 

are used to reach the higher levels of the warehouse.   

The head of supply chain (personal communication, February 11, 2025) at the case company 

explains that there are two different types of orders. Firstly, orders within W01 or warehouse 

one. These are orders either going directly to customers via sales orders or raw material orders 

going to the internal production of the case company. Secondly, there are the orders within the 

W02. These are orders of spare parts which are picked directly for customers. W01 consists of 

3367 different SKUs, while W02 consists of 591.  



 

  30 

 

Orders within the W01 and the W02 are always picked separately as WMS-system classifies 

these as different works (Head of supply chain, personal communication, February 11, 2025). 

Additionally, all products in the automated area are also handled in separate works. This 

implies that there are four different picking types depending on purpose and part of the storage. 

The order types are illustrated in figure 4.4. The separate works imply that W01 SKUs located 

in the automated storage are never picked with W01 SKUs located in the conventional storage 

and so on.  

 

Figure 4.4. The different order types. 

Handheld units are used to keep track of orders (Manager delivery, personal communication, 

February 6, 2025). The delivery specialist picks one order at the time. When a delivery 

specialist starts an order, a list of lines is received. Each line contains an SKU, where it is 

located and how many of the SKU that should be gathered. The delivery specialist then marks 

the line as completed when the product is gathered and moves on to the next product. The route 

between the different order lines shown in the handheld units follows a descending alphabetical 

order. After all lines in the order are complete the delivery specialist either drops off the order 

at the assigned location in the packing space or at the assigned location for production. The 

delivery specialist then starts a new order.  

The handheld units are also used to gather data. Time between the finished lines can be 

extracted. This is the time it takes for a worker to go from one rack to another and pick the 

ordered quantity from that rack. For example, it measures the time to go from A1A to B1A and 

pick a quantity of X SKUs. In this way the time to walk between the racks and picking X 

quantity of the product is included each time, making the time dependent on three variables: 

distance, ease of picking, and quantity.   



 

  31 

 

4.5 Performance Metrics and Evaluation methods 

As a performance metric, the case company benchmarks towards 500 orders a day (Manager 

delivery, personal communication, February 6, 2025). Using 500 orders per day as a 

benchmark and planning method to forecast demand has this far worked fairly well, however 

there are some challenges too. One order could vary in size from just a few SKUs to more than 

50. This greatly impacts the picking time of the orders and thereby also the continuous 

planning. Another challenge is delays due to lack of supply. The manager delivery (personal 

communication, February 6, 2025) mentions how many orders cannot be finished the day they 

are supposed to due to not having enough stock. Inconsistencies between stock on hand and 

assigned orders highlights to suboptimal communications between outbound- and inbound 

logistics planning.  

The case company currently uses Power BI as a tool to evaluate their warehouse processes in 

relation to their performance metrics (Manager delivery, personal communication, February 6, 

2025). This evaluation method makes it possible to use historic data to forecast future demand 

and in that way distribute the orders over time. This evaluation method has thus far been 

perceived as effective in measuring performance and has facilitated planning of resources, 

although it does not explicitly optimize the warehouse’s operations. Furthermore, it is 

hypothesized that the order packing is a bottleneck (Manager delivery, personal 

communication, February 6, 2025). Using orders per day as a warehouse performance metric 

might therefore be misleading as it is affected by bottlenecks further down in the process.  

4.6 Current Decision-Making Process 

The decision-making process regarding the case company’s operational policies can be viewed 

in multiple levels, from more strategic decisions regarding how to optimize warehouse’s order 

picking time, to more ad hoc decisions in the warehouse deciding where to put specific SKUs. 

In regard to RQ2, it is important to navigate how these different levels of decision-making 

affect each other and the overall warehouse processes.   

4.6.1 Strategic Decision-making and Planning 

The warehouse’s main objective is to be able to fulfill the order needs each day. By using tools 

like Power BI, the decision-making process is facilitated by giving the case company the ability 

to plan how to distribute orders effectively over time. As of now the planning process follows 

a cycle where day 0 is the current day and day 1 is the day after etc. (Manager delivery, personal 

communication, February 6, 2025). The goal is to complete all orders for day 0 without spilling 

over to the next day.  

As mentioned before the case company is currently experiencing growth, leading to an 

increased number of orders. The strategic decision is to not increase the number of delivery 



 

  32 

 

specialists, but rather trying to optimize their operational policies. This shows how the three 

levels of decision-making currently act with the strategic level making a decision to optimize 

the warehouse, the tactical level has to decide how to do it, and the operational level lastly has 

to execute on the decision.  

4.6.2 Operational Decision Making  

Within the warehouse, delivery specialists are given big autonomy in decision-making 

regarding where to place SKUs. Most of the current movements are made due to the 

introduction of new SKUs and limited space on the racks rather than trying to make the 

warehouse more efficient (Manager delivery, personal communication, February 6, 2025). 

These changes are usually made ad hoc and based on the experience of the delivery specialists. 

This entails that SKUs sometimes are moved to slots based on existing space rather than an 

optimal spot (Delivery specialist, personal communication: March 20, 2025). There is a 

possibility for the delivery specialist to file suggestions for improvement to other levels of 

decision making, but it is utilized rather seldom and with little effect.   

Since the delivery specialists make most of the decisions regarding where to optimally place 

each SKU one could consider them responsible for the tactical decisions as well since they 

both decide where to locate the SKUs and execute on that (Delivery specialist, personal 

communication: March 20, 2025). In addition, there is no communication from functions like 

purchasing expired or expiring products. Instead, the delivery specialists will receive a signal 

once they scan an expiring product and then find out. This results in expiring and expired 

products taking up good locations in the conventional and automatic part of the warehouse, 

thus wasting both time and space.  

Lastly, the different levels of decision-making experience have some collaborative challenges. 

The VP of supply chain (Personal communication: March 13, 2025) raises the concern that the 

delivery specialists carry too much knowledge and information by themselves. This is due to 

most of the operational decisions concerning the warehouse are made by the delivery 

specialists without any information and data going to the other levels. Implementing the 

scanners was seen as a way to transfer the operational knowledge to the tactical and strategic 

level to facilitate more comprehensive decision making in the warehouse, but as of now it has 

not worked flawlessly  

4.7 Current Order Flow of the Conventional Warehouse 

Using the data collected two matrices have been constructed to illustrate how the warehouse 

and the orders within it currently works. Figure 4.5 shows the median time it takes for a 

delivery specialist to move from a given rack to another rack and retrieve the line for that order. 

Similarly, figure 4.6 shows the percentages of orders going from one shelf to another. Together 

the two figures display which racks are used most often and how quickly the delivery specialists 



 

  33 

 

can move between them. The data depicted also highlights which routes that take longer time 

than expected and which routes that may be underutilized. Additionally, the matrices show the 

incongruities in the retrieved data from the case companies ERP. As mentioned earlier, the 

orders from the AS/RS and conventional warehouse should be separated into different orders, 

but as evident from the matrices, this is not always the case. Similarly, the scanners should 

follow an alphabetical order, but there are exceptions in this data too. The figures are made 

using the quantitative dataset presented in the appendix.  

 
Figure 4.5. The median time between different racks 

 
Figure 4.6. The percentage of order lines going from one rack to another  



 

  34 

 

5. Analysis 

The analysis is divided into three parts, where the first part aims to explain how performance 

evaluation metrics and methods can be utilized to optimize the warehouse, therefore answering 

RQ1.1. The second part uses the findings from the first part to analyze how an operational 

warehouse optimization method can be created to account for multiple product purposes and 

balancing of automation and conventional storage; thereby answering RQ1.2. Lastly the third 

part analyses how the method can be made into a dynamic and iterative decision-making 

process, thereby answering RQ2. 

5.1 Performance Evaluation 

Given the findings from empirical research, the following chapter aims to analyze bottlenecks 

and where there is potential for optimization within the warehouse. The chapter is structured 

like the theoretical framework to emphasize each strategic operation and what measures have 

been taken to optimize it. Note the big emphasis on how and why actions are taken, since the 

objective is to develop a coherent and reusable method, rather than solving a problem once. 

5.1.1 Performance Evaluation Metrics in Conventional Warehouses 

As earlier studies have brought up, the main performance evaluation metric being used when 

optimizing a warehouse is time it takes to perform the order picking operation (Gialos and 

Zeimpekis, 2024). Reducing time and in that way, costs is also of great essence in this project 

and the case company aims to evaluate the performance evaluation method using time saved. 

However, overall warehouse efficiency is dependent on several factors which also can be 

divided into smaller parts. Figure 5.1 displays these connections and how they relate.  

 
Figure 5.1. How the conventional warehouse optimization can be divided into smaller parts 



 

  35 

 

As figure 5.1 shows, the overall efficiency of the warehouse is dependent on how quick an 

order can be picked and the number of orders that must be picked within that time. Similarly, 

the time to pick one order is dependent on how fast each line can be picked and how many 

lines each order has. Additionally, time to pick one line is dependent on the distance the 

delivery specialist must travel plus the ease of picking that SKU multiplied with the quantity 

of SKUs to be picked. Moreover, in bigger warehouses the time to find an SKU also must be 

considered. This is most evident when multiple SKUs share one shelf. As one of the case 

company’s biggest inefficiencies is picking errors (Delivery Specialist, personal 

communication: March 20, 2025), this must be addressed. The picking ambiguity could be 

minimized by giving the SKUs enough space and clear labeling throughout the warehouse. 

The distribution between the activities of the order-picking can vary (Tompkins et al., 2003). 

Figure 2.1 illustrates a typical distribution of picking an SKU but neglects the quantity of given 

SKU that must be picked. Figure 5.1 combined with the typical distribution between activities 

as presented in the theory section allows the conclusion that the travel distance should be 

reduced to optimize the warehouse, but the bigger the quantity of one SKU to pick is, the 

smaller the percentage of the order picking travel time becomes. Therefore, to optimize the 

operational policies, focus should be on how many orders each SKU is a part of rather than the 

picked quantity of each SKU since this will yield the biggest impact on total order picking 

time.  

As the number of lines and orders are dependent on demand they are difficult to explicitly 

influence from the warehouse, but as the manager delivery  (Personal communication: 

February 6, 2025) mentioned, the case company are using benchmarking as a tool to plan 

number of orders per day and if the total order time is reduced by using a more effective 

operational policies, they could meet a higher demand per day. This will implicitly result in 

being able to meet demand and manage more orders per day. Importantly, the demand is 

currently bigger than the case company can supply, which means that by being able to complete 

more orders they can satisfy a bigger demand, thereby increasing their revenue.  

By using benchmarking, significant factors affecting the warehouse are neglected. 

Additionally, as established before the number of lines per order has a great variance. A more 

precise metric to use would therefore be time per line, which in turn is dependent on factors 

illustrated above. Starting narrowly on smaller challenges will allow continuous improvements 

without causing too big risks. Smaller changes in performance metrics are also more likely to 

be accepted by the workers than a sudden big change in their current benchmarking (Kirk, 

2022). As the case company’s goal is to increase the warehouse efficiency without having to 

employ more workers, it is crucial to keep the existing worker satisfied and not make it seem 

like they are being exploited and overworked.  

The chosen metric will be applied to all levels of the operational policies and will be used to 

evaluate the effectiveness of each suggested method. The operational policies are divided 



 

  36 

 

similarly to the theoretical framework, where different methods to optimize it will be evaluated. 

Time to pick one line is as illustrated by figure 5.1 dependent on three factors and each one of 

those correspond to a different part of the operational policies. Table 5.1 shows how each 

operational policy relates to each performance metric. 

Operational Policy Performance Metric 

Inventory Management Space utilization 

Picking Policy  Distance to walk 

Time to pick each SKU 

Quantity of SKUs 

Slotting Policy Distance to walk 

Time to pick each SKU 

Space utilization 

Routing Policy Distance to walk 

Table 5.1 Relations between operational policies and performance metrics 

The synergies within the warehouse are important to highlight and the performance metric of 

space utilization is useful to notice them. For example, this far the total order picking time has 

been used as an evaluation metric, but that neglects the time it would take to replenish the 

SKUs in the warehouse. If order picking time is the only metric used, the importance of buffer 

places would be disregarded. Initially, this would not be noticed in the order picking time, but 

once the SKUs need to be replenished the inflexibility and inefficiencies of the solution would 

be seen. 

5.1.2 Performance Evaluation Metrics in Automated Warehouses 

The performance metrics for the automated parts of the warehouse is relatively similar to those 

of the conventional warehouse. However, instead of walking distance the focus should be on 

shortening the lift travel distance to save time. Since the case company has two lifts in their 

AS/RS meant to work together, the picking time is not only dependent on how fast one lift can 

bring an SKU, but also how fast the other lift can bring the other SKUs in that order. The goal 

should therefore be that when the SKUs are picked from one of the lifts, the other one should 

retrieve the next SKU, resulting in reduced waiting time. 

The case company highlights how the unbalanced lifts reduce the effectiveness of the 

automated storage. About 70% of orders from the automated storage is handled by one lift, 

resulting in unnecessary stall time on the other lift. Additionally, only about 70 orders are 

completed using the AS/RS daily, compared to the 500 daily orders they should be capable of 



 

  37 

 

picking. Therefore, it is not only important to pick a shelf with short lift travel time, but also in 

which lift the SKU should be placed. Connecting with Gu et al. (2010) and Rouwenhorst et al. 

(2000) this therefore becomes a multivariable case where the main goal is to reduce picking 

time, but metrics like travel distance, SKU allocation and limiting space also play a big part in 

how to optimize the automated storage. This directly translates to the metric of space utilization 

which further complicates the case. When filling the AS/RSs one could either try to fit as many 

SKUs as possible or fewer, but the most frequently picked ones. Giving the SKUs too little 

space in the AS/RSs would entail higher rates of replenishment, thus reducing the overall 

efficiency of that solution. Conversely, having too few SKUs in the AS/RSs would mean that 

most of the order is picked in the conventional warehouse, thus not utilizing the AS/RSs 

effectively. It is therefore of importance to consider these tradeoffs in metrics and evaluate the 

total change and not each metric separately.   

5.1.3 Creating method from metrics 

To analyze the current slotting policy a framework was constructed using aforementioned 

performance evaluation metrics. The framework divides the current product into four different 

categories depending on importance and slot quality. The framework is illustrated in figure 5.2. 

SKUs that currently are important and have a good placement in the warehouse are categorized 

as prime pricks, implying that those products already have an adequate placement. Likewise 

unimportant products with lesser places are dust collectors, implying that these products are 

rather irrelevant and that less optimal places are adequate. More important products with bad 

places are defined as lost essentials. Irrelevant products with good locations are defined as 

premium waste. The framework illustrates that lost essentials need better places whilst 

premium waste SKUs can be moved to worse places.  

 

Figure 5.2. The slotting policy framework 



 

  38 

 

The objective with this framework is to identify the lost essentials and premium waste. After 

doing that the SKUs can be sorted after order frequency to their new positions, reshuffling the 

lost essentials and premium wastes. This will result in frequently picked SKUs being placed 

on better locations, thus increasing warehouse efficiency.  

The framework requires a defined breakpoint between the x-axis and the y-axis. i.e. a limit 

between more and less important products and better and worse locations. In this case the 

distribution between the number of orders per product shows heavy skewness. Imp