143
Towards the Optimal Inventory Review Intervals A simulation study into the effect of inventory review intervals on the supply chain performance of a single- product company Areti Satoglou March 2016

Towards the Optimal Inventory Review Intervals

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Towards the Optimal Inventory Review Intervals

Towards the Optimal Inventory Review Intervals

A simulation study into the effect of inventory review intervals on the supply chain performance of a single-product company

Areti Satoglou March 2016

Page 2: Towards the Optimal Inventory Review Intervals

2

Page 3: Towards the Optimal Inventory Review Intervals

3

Information Master thesis – Management of Technology

Areti Satoglou – 4325516

March 2016

Graduation Committee

Chairman Dr. R.M. (Robert) Verburg

1st supervisor Dr. M.A. (Michel) Oey

2nd supervisor Dr. Ron van Duin (J.H.R.)

External supervisor Dr. Ivo Wenzler (from Accenture)

Laurens van der Drift (from Macomi)

Page 4: Towards the Optimal Inventory Review Intervals

4

Page 5: Towards the Optimal Inventory Review Intervals

5

Contents Abbreviations .......................................................................................................................................... 8

Aknowledgements ................................................................................................................................... 9

Executive summary ............................................................................................................................... 11

1 INTRODUCTION ............................................................................................................................. 15

1.1 Background information........................................................................................................ 15

1.2 Continuous vs periodic replenishment policy ....................................................................... 16

1.3 Presentation of the company ................................................................................................ 16

1.4 Presentation of the situation ................................................................................................ 16

1.5 Positioning of the author’s contribution in the project ........................................................ 17

2 RESEARCH PROBLEM AND APPROACH .......................................................................................... 19

2.1 Knowledge gap and problem statement ............................................................................... 19

2.2 Research objective ................................................................................................................ 20

2.3 Scientific and social relevance ............................................................................................... 20

2.4 Research Questions ............................................................................................................... 20

2.5 Conceptual model of research .............................................................................................. 20

2.6 Research approach ................................................................................................................ 21

2.7 Research methods: literature review, case study design and simulation ............................. 22

2.7.1 Literature review ........................................................................................................... 22

2.7.2 Embedded single case study strategy ........................................................................... 23

2.7.3 Simulation ...................................................................................................................... 23

2.7.4 IDEF0, system theory, black box theory ........................................................................ 27

2.7.5 DEMO methodology ...................................................................................................... 27

2.8 Outline of report.................................................................................................................... 29

2.9 Concluding remarks ............................................................................................................... 30

3 SUPPLY CHAIN ANALYSIS ............................................................................................................... 31

3.1 Research framework for literature review ............................................................................ 31

3.2 Inventory Theory ................................................................................................................... 31

3.2.1 Reasons for focusing on inventory management .......................................................... 32

3.2.2 Why keeping inventories? ............................................................................................. 33

3.2.3 Inventory in the supply chain ........................................................................................ 33

3.2.4 Inventory strategies ....................................................................................................... 34

3.2.5 Business operations model ........................................................................................... 34

Page 6: Towards the Optimal Inventory Review Intervals

6

3.2.6 Introduction to inventory control ................................................................................. 35

3.2.7 Inventory classification .................................................................................................. 37

3.2.8 Inventory control in certain conditions ......................................................................... 39

3.2.9 Inventory control in uncertain conditions: stochastic inventory models..................... 40

3.2.10 The independent variable: the inventory review interval ............................................. 42

3.2.11 The dependent variable: Supply chain performance .................................................... 42

3.2.12 The key performance indicators (KPIs) .......................................................................... 42

3.3 Concluding remarks ............................................................................................................... 43

4 CASE STUDY ANALYSIS ................................................................................................................... 45

4.1 Selection of case study strategy: embedded single case study ............................................ 45

4.2 Case study performance ........................................................................................................ 47

4.2.1 Current and future replenishment policy of company X ............................................... 48

4.2.2 Detailed description of the logic for current replenishment policy (the (s,S) policy) ... 50

4.2.3 Detailed description of the logic for new replenishment policy (Min/Max policy) ...... 52

4.2.4 Data description ............................................................................................................ 53

4.3 Linking the literature review and the case study with the conceptual design of simulation 53

4.4 Concluding remarks ............................................................................................................... 54

5 INVENTORY SIMULATION MODEL ................................................................................................. 55

5.1 Model conceptualization ....................................................................................................... 55

5.1.1 Model objectives ........................................................................................................... 55

5.1.2 Conceptual design of simulation ................................................................................... 55

5.1.3 Description of the conceptual design: system as a black box ....................................... 56

5.2 Specification of the conceptual model .................................................................................. 56

5.2.1 The control variables : IRI and inventory replenishment policy .................................... 56

5.2.2 The mechanisms: the choice of the simulation archetype ........................................... 56

5.2.3 Inputs: data collection and data analysis ...................................................................... 59

5.2.4 Outputs: KPIs expressing finished inventory and service levels .................................... 63

5.2.5 Specification and analysis of the system: Opening the black box ................................. 63

5.3 Verification and Validation of the model .............................................................................. 70

5.3.1 Verification .................................................................................................................... 70

5.3.2 Validation....................................................................................................................... 71

5.4 Simulation and Results .......................................................................................................... 85

5.4.1 Design of experiments ................................................................................................... 85

5.4.2 Execution of experiments .............................................................................................. 88

5.4.3 Results ........................................................................................................................... 90

5.5 Concluding remarks ............................................................................................................. 103

Page 7: Towards the Optimal Inventory Review Intervals

7

6 CONCLUSIONS ............................................................................................................................. 105

6.1 Answering the research questions ...................................................................................... 105

6.1.1 Review of the sub research questions......................................................................... 105

6.1.2 Review of the main research question ........................................................................ 109

6.2 Generalization of results ..................................................................................................... 110

7 LIMITATIONS, RECOMMENDATIONS AND REFLECTION .............................................................. 113

7.1 Limitations ........................................................................................................................... 113

7.2 Recommendations for future research ............................................................................... 114

7.2.1 Future research from a business perspective ............................................................. 114

7.2.2 Future research from an academic perspective .......................................................... 114

7.3 Reflection............................................................................................................................. 115

8 APPENDICES ................................................................................................................................. 117

A: Simulation results ........................................................................................................................ 117

B: “% Delivered on time versus Requested” KPI comparison between the two replenishment

policies regarding ABC classification ............................................................................................... 120

C: “% Delivered on time versus Requested” KPI comparison between the two replenishment

policies regarding XYZ classification ................................................................................................ 123

D: “% Delivered on time versus Requested” KPI comparison between the two replenishment

policies regarding Lead Time classification ..................................................................................... 127

E: “Inventory final product” KPI comparison between the two replenishment policies regarding

ABC classification ............................................................................................................................. 130

F: “Inventory final product” KPI comparison between the two replenishment policies regarding XYZ

classification .................................................................................................................................... 134

G: “Inventory final product” KPI comparison between the two replenishment policies regarding

Lead Time classification ................................................................................................................... 137

9 Bibliography ................................................................................................................................. 139

Page 8: Towards the Optimal Inventory Review Intervals

8

Abbreviations

KPI Key Performance Indicator

IRI Inventory review interval

ATD Actor Transaction Diagram

TRT Transaction Result Table

Page 9: Towards the Optimal Inventory Review Intervals

9

Aknowledgements Executing this graduation project and writing a thesis about the research is the completion of my

master Management of Technology at the Delft University of Technology. I made the conscious

choice of performing this research in combination with an internship at a company in order to gain

work experience besides the execution of my graduation. At Macomi and Accenture I found an

interesting case to work on a project of a Macomi and Accenture’s client, together with a very

interesting company and environment and a group of nice colleagues. Furthermore, I would like to

express my gratitude to those who helped, guided and supervised me in the process of writing this

thesis.

From the University, I would like to thank Michel Oey, my first supervisor, for his involvement in my

project, his willingness to help me, his feedback and the interesting discussions we had during our

meetings. I also want to thank Robert Verburg as Chair of my graduation committee and Ron van

Duin as my second supervisor from the University, for their helpful feedback during the ‘official’

meetings of my project.

From Macomi, I would like to thank Michel Fumarola and Corne Versteegt for supervising my

progress throughout my graduation and also Ivo Wenzler form Accenture, for in the first place

providing me with an internship there. Moreover, I would like to thank Tim Tutenel from Macomi,

for his patience when explaining to me the technical parts that I did not know about or when

exporting for me database queries from S3N. And of course, I would like to thank Laurens van der

Drift for regularly giving useful feedback on my work, for helping me building the simulation model

and for his patience consumed by me when answerig all these questions of mine during those 9

months of internship. (It is true that if I gave him a penny everytime he answered to a question of

mine, by now he would have made a fortune. )

Also, I would like to thank my father Pavlos, mother Ntina and sister Fay, for their interest and full

support in the progress of my graduation. My friends are appreciated for their support and for all our

nice and relaxing activities that helped me taking my mind of the thesis every now and then.

Page 10: Towards the Optimal Inventory Review Intervals

10

Page 11: Towards the Optimal Inventory Review Intervals

11

Executive summary Inventory management is one of the cornerstones of supply chain management as inventory consists

of a key contributor in all supply chains. In the supply chain sector, many inventory management-

related problems have been broadly investigated and discussed over the years. It is of great

importance for the supply chain of a company to make proper decisions when it comes to plan the

order quantity or the time of placing an order, the safety stock that should be keep, the optimal

location for its warehouse or other related decisions. Further, it is understood that defining “how

much to stock” is closely correlated with defining “how much to order”. Hence, each firm follows a

specific replenishment policy according to its needs.

One categorization among inventory replenishment policies is made with respect to the inventory

review interval (IRI). The IRI refers to the frequency of reviewing the inventory to determine when

orders must be placed for replenishment. However, the IRIs differs among policies and companies.

According to this categorization inventory replenishment policies can be either continuous or

periodic. In the continuous review process, the inventory levels are continuously reviewed, and as

soon as the stocks fall below a predetermined level (known as the reorder point or reorder level), a

replenishment order is placed.

This research is conducted from the scope of a specific company’s supply chain, company X that is an

international book seller. Company X is currently one of the many clients of Accenture. Company X is

currently using an inventory replenishment policy similar to the (s,S) policy. However, company X is

in a transition phase aiming to change this current inventory replenishment policy to the so called

Min/Max replenishment policy. The Min/Max inventory replenishment policy is based on the EOQ

calculation. At this project conducted by Accenture, Macomi, a business consulting firm, provides the

simulation capabilities. A discrete event simulation model was built for providing company X with the

optimal EOQs of company X’s suppliers for the new replenishment policy by Macomi with the

author’s contribution. However, it is unknown how the IRI of the current or the new inventory

replenishment policy can influence the supply chain performance of Company X. The author of this

thesis built with Macomi a simulation model and adapted it afterwards in order to conduct this

research. The author contributed in building the simulation model with Macomi and parametrized it

afterwards along with collected data of company X provided by Accenture for her research purposes.

From the aforementioned, the aim of this research is to explore the relationship between the IRI and

supply chain performance of company X. Hence, the main research question that needs to be

answered is: “What are the effects of review intervals on the supply chain performance of company

X?” The following sub questions are intended to be answered in order to answer to the main

research question presented above. Those are formulated below:

1. Why it is important to focus on the IRI and what theories are relevant regarding the IRI?

2. Which inventory replenishment policies and IRI ranges are relevant to company X’s case?

3. How can supply chain performance be defined and measured, in the case of company X?

4. How can one test the effect of IRIs on supply chain performance of company X?

5. How do supply chain performance metrics behave under different IRIs for company X?

First, in order to understand the purpose of inventory management, theoretical drivers of inventory

were gathered. This includes the analysis of the currently used inventory model. Also, measures were

Page 12: Towards the Optimal Inventory Review Intervals

12

selected to rate the performance of inventory management practices, the KPIs. This step is achieved

by conducting first a literature review. A top down approach is applied for that purpose starting from

general notions such as supply chain management and narrowing down step by step in a systemic

way to the IRI that is the key concept of this research. Second, an embedded case study strategy was

used, specifically company X‘s case, in order to scope the research.

A discrete event simulation model was built for providing company X with the optimal EOQs of

company X’s suppliers for the new inventory replenishment policies by Macomi with the author’s

contribution. After the model was finished, the author built another version of the model to use it in

order to answer the relevant research questions. The simulation model’s main outputs are

inventories and service levels expressed in Key performance indicators (KPIs). The author’s model is

based on Macomi’s model and is used to conduct this research to identify the impact of the IRI on

supply chain performance.

Hence, first the conceptual design of the simulation was developed using IDEF0. The conceptual

model was perceived as a system. After defining the inputs, controls mechanisms and outputs as

seen in the following figure, the system that was perceived as a black box was opened up using the

DEMO methodology. DEMO was selected as a suitable methodology in order to map the business

processes that are relevant to the inventory management and inventory control of company X.

Figure 1: The conceptual design of the simulation presented as a system diagram.

Afterwards, the model is implemented in the S3N interface. Before using the simulation model to run

the tests, the author verified and validated the model. Subsequently, the design and the execution of

experiments followed. After having executed all the experiments, the author proceeded in exporting

the relevant KPIs for obtaining the results and for being able to analyze them afterwards, for all the

57 products that were selected from the product list. There are two relevant KPIs: “Inventory final

product” and “% Delivered on time versus Requested.

As expected, the results showed that both KPI values decline as we move from smaller to bigger IRIs:

both the finished inventories and the service levels decrease as we move from smaller to bigger IRIs.

However, if one looks at the scale, it is not that strong. Further, looking at both the two different KPI

behaviors, they are consistent in their decrease. However, it is noticed that the decrease is not that

intense, and hence, the influence of the IRIs is not that big. Moreover, it was observed that as long as

Page 13: Towards the Optimal Inventory Review Intervals

13

the IRIs are fluctuating from 1 day to 1 month, the KPI values do not decrease that significantly. The

KPI values drop faster as we move the IRI beyond one month.

Hence, valuable recommendations are made from a business perspective and from a scientific

perspective as well. For instance, from a business perspective, a recommendation for company X

would be to estimate if the Min/Max policy that is proposed to be “a real time” policy is indeed

beneficial: Based on the results of this research, it could be stated that the Min/Max policy works

efficiently with the low levels of inventory that are kept, but still, the review interval of the inventory

does not seem to influence the service level and the inventory KPI. Thus, paying for the real time

policy implementation, from the IRIs perspective does not seem beneficial enough since during one

month period it is not needed to have a real time observation on what is happening to the

inventories. More specifically, the results showed that the influence of the IRIs is not that important

when they are less than one month. In other words, there is no need for investing on expensive

software that helps company X managing and controlling the replenishment of products in real time

because both inventories and service levels decrease with a low rate.

From a scientific perspective, a knowledge gap was tackled as there was not much literature found

regarding the impact of the IRIs on supply chain performance. Further, this research was scoped for

a specific company; company X, in order to answer to the research question “What are the effects of

review intervals on the supply chain performance of company X?” Thus, from a scientific perspective,

it was not clear how supply chain performance could be influenced by different IRIs. Hence, the

author selected a case in order to narrow down the scope of the research. However, limitations that

occurred are presented and discussed. Hence, based on the result interpretations and the limitations

that are observed, someone in the future could perform another case study to test this model for

another company with different supply chain. Nevertheless, one could go further and perform a

more fundamental study and not take just one case of company, but perform a more controlled

experiment in which he is going to turn all the variables that have been thoroughly discussed during

the implementation of the experiments.

Page 14: Towards the Optimal Inventory Review Intervals

14

Page 15: Towards the Optimal Inventory Review Intervals

15

1 INTRODUCTION This chapter start with providing the reader wih background information on relevent theories

regarding supply chain and invenory management and inventory replenishment policies.

Subsequently, an introduction to the company that is considered in this thesis takes place along

with the backgroud information regarding the situation and the positioning of the author of this

research and his contribution in the existing project of this company.

1.1 Background information Supply chain management

In today’s complex market place the competition is towards supply chains rather than individual

companies. Hence, the need of efficient supply chain management is evident. Supply chain

management is commonly defined as the management of the flow of goods, information and

services. It includes the movement and storage of raw materials, work-in-process inventory, and

finished goods from point of origin to point of consumption. Supply chain management has also

been defined as the "design, planning, execution, control, and monitoring of supply chain activities

with the objective of creating net value, building a competitive infrastructure, leveraging worldwide

logistics, synchronizing supply with demand and measuring performance globally.” (APICS, n.d.)

In addition, when managing and developing the supply chain, performance measurement of the

entire supply chain and its processes is essential and also helpful in the continuous improvement of

SCM. Performance measurement provides information for managers and decision makers that

enable them identifying which strategies to follow and facilitating the understanding of the situation

in a market of complexity and uncertainty. Further, performance measurement assists in directing

management attention, revising company goals, and re-engineering business processes. (Chan,

2003)

Inventory management

Inventory management is one of the cornerstones of supply chain management as inventory consists

of a key contributor in all supply chains. Logistic costs, holding inventory costs are examples of costs

that all companies want to control while providing satisfactory services to their clients. Thus,

effective management of inventories is considered to be a vital function of management plays a

crucial role in basic engineering management topics such as quality management and lean

manufacturing. In general, successful inventory management involves creating a purchasing plan that

will ensure that items are available when they are needed. This means that the inventory level

should be neither too low nor too high.

Moreover, successfully managing the inventory means keeping track of existing inventory and its use.

Two common inventory-management strategies are the just-in-time method, where companies plan

to receive items as they are needed rather than maintaining high inventory levels, and materials

requirement planning (MRP) which schedules material deliveries based on sales forecasts. In order to

control the imbalances between supply and demand, companies use various inventory management

methods. Usually, demand is uncertain and that is why firms should always be able to fulfill customer

needs and have an adequate inventory level that is not so high to cause excessive costs to the firm

and at the same time not so low to prevent a stock-out situation.

In the supply chain sector, many inventory management-related problems have been broadly

investigated and discussed over the years. It is of great importance for the supply chain of a company

Page 16: Towards the Optimal Inventory Review Intervals

16

to make proper decisions when it comes to plan the order quantity or the time of placing an order,

the safety stock that should be keep, the optimal location for its warehouse or other related

decisions. Further, it is understood that defining “how much to stock” is closely correlated with

defining “how much to order”. Hence, each firm follows a specific replenishment policy according to

its needs.

1.2 Continuous vs periodic replenishment policy Inventory Replenishment policies and inventory management are close related. Planning of the

inventory establishes the optimal inventory level for each company to maintain in order to balance

costs and service levels for demand fulfillment. In general, reordering or replenishment process

needs to define a review period for reordering and an ordering quantity. Then it needs the inventory

parameters to determine whether an order for replenishment should be placed at the time of review

or not. Based on how the review period and order quantities are defined, there are a few options to

drive the reordering.

Hence, one categorization among inventory replenishment policies is made with respect to the

inventory review interval (IRI). This refers to the frequency of reviewing the inventory to determine

when orders must be placed for replenishment. However, the IRI differs from inventory

replenishment policy to inventory replenishment policy and from company to company. According to

this categorization inventory replenishment policies can be either continuous or periodic. In the

continuous review process, the inventory levels are continuously reviewed, and as soon as the stocks

fall below a predetermined level (known as the reorder point or reorder level), a replenishment

order is placed. Regarding the periodic review, the inventory levels are reviewed at a set frequency.

This could be the end of each day, week, year etc. At the time of the review, if the stock levels are

below the pre-determined level, then an order for replenishment is placed, otherwise it is ignored

until the next cycle.

1.3 Presentation of the company This research is conducted from the scope of a specific company’s supply chain. The company

selected will not be named due to confidentiality issues. Thus in this thesis the chosen company will

be referred to as company X. Company X was selected as a case study, thus, data from company X

were provided to the author in order to conduct this research. Company X is a multinational

publishing and education company. It is one of the largest education companies and book publishers

in the world. Furthermore, the company has a limited portfolio of products. The focus of this thesis is

only on books as product type. Hence, in this report the company will be referred to as a single

product- company, meaning the books.

1.4 Presentation of the situation Company X is currently one of the many clients of Accenture. Accenture is a management

consulting, technology services and outsourcing company. Company X’s supply chain consists of a

central warehouse in UK, 16 Litho-suppliers worldwide and 2 digital suppliers based in UK. Accenture

has been asked from company X to propose the optimal Economic Order Quantities (EOQs) for the

18 suppliers. EOQ is a calculation that determines the most cost effective quantity to order or

produce by finding the point at which the combination of order cost and carrying cost is the least.

Company X is currently using an inventory replenishment policy similar to the inventory

replenishment policy called the (s,S) policy, that will be discussed in the following chapters. However,

Page 17: Towards the Optimal Inventory Review Intervals

17

company X is in a transition phase aiming to change this current inventory replenishment policy to

the so called Min/Max replenishment policy. The Min/Max inventory replenishment policy is based

on the EOQ calculation. At this project conducted by Accenture, Macomi, a business consulting firm,

provides the simulation capabilities. A discrete event simulation model was built for providing

company X with the optimal EOQs of company X’s suppliers for the new replenishment policy by

Macomi with the author’s contribution. However, it is unknown how the IRI of the current or the

new inventory replenishment policy can influence the supply chain performance of Company X. The

author of this thesis built the simulation model with Macomi and adapted it afterwards in order to

conduct this research: The author contributed in building the simulation model with Macomi and

parametrized it afterwards along with collected data of company X provided by Accenture for her

research purposes.

1.5 Positioning of the author’s contribution in the project With respect to the requirement of the University, the author conducted this thesis externally at

Accenture. For the purposes of the research, the author was involved in the project of company X.

The author contributed in building the simulation model with Macomi and parametrized it

afterwards along with collected data of company X provided by Accenture for her research purposes.

Page 18: Towards the Optimal Inventory Review Intervals

18

Page 19: Towards the Optimal Inventory Review Intervals

19

2 RESEARCH PROBLEM AND APPROACH This chapter presents the knowledge gap along with the formulation of the problem statement.

Hence, the research objective stems from the identification of the problem at stake and further, the

scientific and societal relevance regarding this topic are tackled. Subsequently, the research

questions are formulated, a preliminary conceptual model is presented and the research approach

and methods that are necessary to conduct this research are introduced. In the end of the chapter,

the outline of the report is shown along with some concluding remarks.

2.1 Knowledge gap and problem statement Relationship between IRI and supply chain performance

Inventory replenishment policies have been thoroughly investigated along with the effects that they have on the supply chain performance. Example papers can be found in the literature. For instance, in a paper, (Lau, et al., 2008) the effects of information sharing and early order commitment on the performance of four inventory policies used by retailers in a supply chain of one capacitated supplier and four retailers, are investigated. However, while conducting this preliminary literature research, it was observed that there is little literature on this topic and mostly papers that are recently published. In a recent article (Shang, et al., 2015), a periodic inventory review system is considered that aims to minimizing the average costs per period. Shang et al find the optimal reorder intervals for their research by decomposing the total costs into each facility and then construct a lower bound to the allocated facility cost. Subsequently they use these lower bounds to reach bounds for the optimal order intervals. Moreover, in an older article (Axsater, 1993)a system with a periodic review order-up-to-S policy is considered and provides procedures for the evaluation of holding and shortage costs. Further, Liu et al (2012) studied the optimization of the (S, T) policy where T is the replenishment interval and S is the order-up-to level. In their research they identify properties based on which they develop algorithms to calculate the optimal policy for cases of either continuous or discrete demand. The IRI refers to the frequency regarding how often to review the inventory in order to determine when to place the next order. In other words, a review period is the length of the interval between two consecutive inventory reviews. From a literature review perspective, a reason why there is not enough literature on the topic is because the focus is usually on determining the optimal order quantity instead of the optimal review interval: If one determines the optimal order quantity he can define what the review period will be. A simple calculation for that would be to divide the demand by the order quantity to estimate the number of times needed to place an order and thus, one can calculate what will be the order period. Moreover, it is observed in both literature and real life cases that this frequency varies among replenishment policies and companies. With respect to the IRI, as mentioned before, there are two kinds of inventory reviews: continuous review and periodic review. From the aforementioned, it is intended to explore the relationship between the IRI and supply chain performance of company X. Hence, the overall problem that is intended to be explored in this research can be formulated as follows: “There is not a clear picture regarding the effects of the IRI on supply chain performance.” Given the fact that company X is chosen for performing a case study, this problem will be tackled in the frame of Company X.

Page 20: Towards the Optimal Inventory Review Intervals

20

2.2 Research objective Based on the problem statement the following research objective is formulated:

“The research objective is to investigate the impact that the IRI has, on the supply chain performance

of Company X.”

2.3 Scientific and social relevance To begin with, from a scientific perspective, a knowledge gap will be investigated and the research

will shed light on the relationship between the review interval and supply chain performance on a

real life case. Specifically, the result of conducting a preliminary literature review showed that this

topic is not researched much due to the very few literature that there exists. As mentioned, it was

assumed while reviewing literature that the focus usually is on determining the optimal order

quantity instead of first destemming the review intervals. Hence, the outcome of the research

intends to contribute in linking the notions of supply chain performance and IRIs.

The social relevance lies in the added value in practice for company X, the company chosen for the

research conduction. More specifically, the exports of the simulation model that is built, aims to give

useful recommendations to Company X regarding the dependence of the IRI on supply chain

performance for both the current and the new replenishment policies of company X. Hence, the

research product could be used as an advising report, regarding which review interval would best fit

the company’s profile with respect to supply chain performance metrics (KPIs). In addition, the

research product aims to be valuable for Accenture in practice, as the stakeholders of the project

from the side of Accenture will able to enrich their advice to company X and in the future apply these

insights to other similar projects.

2.4 Research Questions The main research question follows from the problem definition and from the aforementioned

research objective and is formulated as follows:

“What are the effects of review intervals on the supply chain performance of company X?”

In order to answer this research question, the following sub questions have to be answered. These

questions provide structure for the research.

1. Why it is important to focus on the IRI and what theories are relevant regarding the IRI?

2. Which inventory replenishment policies and IRI ranges are relevant to company X’s case?

3. How can supply chain performance be defined and measured, in the case of company X?

4. How can one test the effect of IRIs on supply chain performance of company X?

5. How do supply chain performance metrics behave under different IRIs for company X?

2.5 Conceptual model of research The basic conceptual model of this research is causal and it can be depicted by the following simple

diagram in Figure 2:

Page 21: Towards the Optimal Inventory Review Intervals

21

Figure 2: The basic conceptual model

As illustrated in the figure, the dependent variable is the supply chain performance and the

independent variable is the IRI. The causal relationship among those is the interest of this research

that is framed within the case study conducted in company X.

2.6 Research approach Ostrom (2008) defines the hierarchy between three levels of analysis: frameworks, theories and

models. According to her, frameworks help to identify general concepts including important

elements and constraints for understanding the problem. Nevertheless, theories are a level lower

and are used to focus on the key parts of a framework and make certain assumptions to analyze the

problem. Regarding models, they fill in a selected group of parameters which result in a set of

outcomes. This chapter will introduce the research framework, which will be the start to select

applicable theories and models.

For a structured approach, a research framework is designed based on Verschuren & Doorewaard

(2010) and using the defined research objective. The framework is presented in figure 3 and

illustrates an overview of the stages of performing this research.

Figure 3: Framework for research approach

Page 22: Towards the Optimal Inventory Review Intervals

22

Moreover, the stages will be now elaborated on:

a) Select key drivers and performance measures for inventory management

In order to understand the purpose of inventory management, first theoretical drivers of inventory

should be gathered. This includes the analysis of the currently used inventory model. Also, measures

will be selected to rate the performance of inventory management practices. This is achieved by

developing and applying a research framework to conduct the necessary literature review. A top

down approach is applied for that purpose starting from general notions such as supply chain

management and narrowing down step by step in a systemic way to the IRI that is the key concept of

this research.

b) Perform case study to analyze the inventory management situation in company X

Using the developed framework in phase a), a case study is performed within the frame of company

X. The desk research or literature review conducted in the first step is combined to better

understand the inventory management and control that is applied in company X.

c) Set up approach for developing an inventory management simulation model

A design approach for the simulation model is developed using literature on inventory modeling,

simulation and general modeling methods. Theory on systems and model, for example the black-box

model, as well as theory behind the DEMO methodology is used for that purpose.

d) Design inventory simulation model and validation tests

The design approach and the system analysis of the current situation are used to develop an

inventory management simulation model. Purpose of this model is to test how the different IRIs

influence supply chain performance. Supply chain performance is defined in terms of service levels

and measuring inventories.

e) Analyze the effects of IRI on supply chain performance using the simulation model

The effects of IRI on supply chain performance are tested using the simulation model. Results are

recommendations for inventory management at company X and the added value of using a

simulation model for academic and business purposes.

f) Validate new inventory management practices and evaluate results

The results of the simulation model are validated with expert opinion, use of historical data and

comparison to the model of Macomi. Evaluation on the results is performed and generalization of

them is tackled. Further, limitations of the simulation model are identified and recommendations are

defined for future research both from a business and from a scientific perspective. In the end,

reflections on the research process are also identified.

2.7 Research methods: literature review, case study design and simulation

2.7.1 Literature review Regarding the literature review, the research objective is to identify the impact of the inventory

review interval (IRI) on supply chain performance for the case of company X. Soon it became clear

Page 23: Towards the Optimal Inventory Review Intervals

23

that there is not an easy way to start the research by focusing on the two main variables. Therefore

first a comprehensive literature review was conducted on supply chain management and inventory

management in particular. Afterwards, an additional literature review is performed that involves the

summary, collation and/or synthesis of existing research. The performance of this literature review is

implemented based on the top-down approach.

2.7.2 Embedded single case study strategy First, the core method used in this research is the conduction of a case study as the research will

consists of company X’s case. Case study research is used because it is particularly suitable for

studying a phenomenon in its real-life context, and when the phenomena are intertwined with the

situation (Yin, 2009)

The problem defined is on the one hand, a problem that can be considered by many companies.

However, in order to quantify it and perform an analysis, it has been chosen to do so from the scope

of a specific company, company X. The case study method allows the researcher to retain the holistic

and meaningful characteristics of real-life events (Yin, 2009). Moreover, it is not possible to identify

the effect of the IRI on supply chain performance on the basis of highly aggregated data when every

company is unique because in every separate case there is a different set of parameters and

variables that depend on the company’s strategies, characteristics and policies that are followed.

Further, the results will be more accurate and meaningful not only from a technical perspective for

this specific company, but also form a social perspective by contributing in solving a real-life problem

of a company with similar characteristics as company X.

For the purpose of this research, the author will use a single-case study design to confirm, challenge,

and broaden existing theory (Yin, 2009). A single case study focuses on only one case to be

thoroughly examined (Verschuren & Doorewaard, 2010). This will be the case of Company X.

Moreover, with respect to a case study research design, the researcher should identify if the single

case study is a holistic or an embedded one. A holistic case study means studying a case in its totality

and it is used when no logical subunits can be identified and a study might be conducted on a too

abstract level. On the other hand, in an embedded case study, various subcases or units or processes

are studied (Verschuren & Doorewaard, 2010). A characteristic of an embedded case study is that

extensive analysis is performed and also it might focus too much on subunits, thus, losing higher level

(holistic) aspects. Regarding to the characteristics of this spicific case study, it is considered by the

author to be an embedded single case study. This case is not examined with a holistic perspective but

in fact there are sub-units related to the IRI such as demand, forecast of the demand, different

categories of inventory replenishment policies that lead to different processes and to products with

different characteristics. Hence, extensive analysis will be performed in order to reach the research

objective and thus, this is the reason for using the research strategy of an embedded case study.

2.7.3 Simulation A discrete event simulation model was built for providing company X with the optimal EOQs of

company X’s suppliers for both the current and the new inventory replenishment policies by Macomi

with the author’s contribution. After the model is finished, the author built another version of the

model to use it in order to answer to the relevant research questions.

Simulation Design Approach

Page 24: Towards the Optimal Inventory Review Intervals

24

The conceptual design of simulation will be developed by perceiving the model as a system. A system

is by definition a combination of elements of parts that format a complex or unitary whole

(Blanchard, et al., 1998). When dealing with complicated systems, it is not easy to analyze every

system aspect. Thus, systems engineering constitutes the endeavor of adopting a goal centered and

systematic approach in order to analyze and integrate every aspect of a system. To continue, system

engineering aims to ensuring that the system requirements are met. Hence, the life-cycle of a system

starts with a requirements analysis and subsequently the system design and development have to be

elaborated in order for the system to be implemented. Moreover, regarding the structure of a

system, every system entails components, attributes, and relationships (Blanchard, et al., 1998).

Components are the parts of a system that consist of inputs, processes and outputs. Further,

attributes refer to the properties of manifestations of the components of a system. Finally, the links

between components and attributes express the relationships (Blanchard, et al., 1998).

A model is the representation of a system and its involved relationships (Blanchard, et al., 1998).

Building a model is essentially the process of developing a model for a real life system for which

information exists. However, it is not easy in many cases to experiment with a real life system. Thus,

decisions and improvements can derive from the use of models and their analysis by gathering

information regarding the system when it is not easy to experiment directly on it.

When it comes to analyze and interpret a complex system, there are various approaches from which

one can choose. One approach is the analytical approach that entails the building of a mathematical

model. Another approach of building a mathematical model is simulation. Simulation is a useful tool

that can be considered when a model of a system is built and a researcher wants to perform an

experiment to test the various scenarios that can occur during the life cycle of a real life system. This

way the behavior of a system can be represented through the use of simulation. The most usual type

of simulation is to build a computer model using simulation software in order to imitate and explain a

system’s behavior.

More specifically, simulation is very useful for supply chains for the decision makers. In general, in a

supply chain there are decisions to be made on an operational level and also on a structural level.

With the use of simulation improvements on an existing supply chain or on a new supply chain to be

implemented can derive. In the case of company X, simulation is used to derive to operational

improvements for the company, as the focus is on inventory management and control.

According to Law & Kelton (1991), in order to build a simulation model for any system there is a

sequency of steps to take. The process starts with the proper definition of the problem and also of

the system that is studied and will be modeled. Further, after the implementation of specific steps,

the process ends with the implementation of the simulation results on the real world system.The

steps are presented bellow (Law & Kelton, 1991):

1. Formulate the problem and plan study

2. Collect data and define the model

3. Check the validity of data

4. Construct the computer program

5. Make the pilot runs

6. Check the validity of the model

7. Design the experiments

8. Run the experiments

9. Analyze the output data

10. Document, present, and implement the results

Page 25: Towards the Optimal Inventory Review Intervals

25

For this research, those steps will constitute the design approach that will be used in order to build

the simulation model. This approach is also applied in order to define the conceptual design of the

simulation.

The simulation model

The simulation model’s main outputs are inventories and service levels expressed in Ley performance

indicators (KPIs). The author’s model is based on Macomi’s model and is used to conduct this

research to identify the impact of the review interval on supply chain performance. In other words,

to explore the effect of the interval on supply chain performance, it is needed to investigate how

supply chain performance is affected by different time intervals. For example, what changes in supply

chain performance metrics if instead of every month the inventory review is implemented every

week or every day? Such tests are performed using the simulation model that is built.

Fixed Parameters

Regarding the structure of the supply chain in the model, 18 locations of the suppliers are included

as well as the central warehouse in UK, a virtual customer (only one), the logistic services, the types

of products and an inventory manager that manages the inventory in two different ways : within

the ( s,S) policy and within the Min/max policy. The safety stocks are given in both cases and are

considered fixed by company X. The whole supply chain will be considered as fixed including the

aforementioned elements as well.

Varying Parameters: Product- dependent

There are two types of parameters that can vary in this model in order to make sense and generate

valuable and relevant results for this research: the product- dependent parameters and the

replenishment policy- dependent ones. Regarding the product-dependent parameters that can vary,

those are:

Fast versus slow moving products

Demand volatility

Demand predictability or forecast accuracy

In general, by demand volatility, it is meant that we could have the same average demand although

the demand could be more volatile and vary along time. In company X’s case it is intended to select a

number of products that are distinctive in terms of demand volatility, of differentiation between

slow or fast movers and demand predictability. Nevertheless, demand predictability refers to the

estimation of how accurate is the forecast and it will be scoped in the research. Forecast accuracy

relates forecasted sales quantities to actual quantities and measures the ability to forecast future

demands. (Stadtler & Kilger, 2008)

Furthermore, in multi-item inventory systems, classification of inventories can help to reduce the

complexity of managing thousands of items. In many cases of inventory management the use of two

dimensional classification systems is also usual as the first is the traditional ABC classification and the

second is based on criticality. In this research a two dimensional classification of items is used. The

first is the traditional ABC classification and the second is based on variability (XYZ classification).

Those classifications provide indications on how to distinguish products based on the

aforementioned product-dependent parameters. (Fast versus slow moving products, demand

volatility and demand predictability)

Page 26: Towards the Optimal Inventory Review Intervals

26

ABC- XYZ classification

In order to build scenarios, products should be first selected. More specifically, the products will be

selected according to the ABC classification being an indicator of slow or fast movers, and to the XYZ

classification that is an indicator of the items ‘demand predictability and expected forecast accuracy.

Regarding the ABC classification, the items are classified based on demand (consumption) rate:

Inventory control is based on a form of Pareto analysis. The inventory items are divided into three

categories (A, B, and C), according to a criterion such as revenue generation, turnover, or value.

Typically, 'A' items represent 20 percent in terms of quantity and 75 to 80 percent in terms of the

value. Regarding the XYZ classification, it is an indicator of the items ‘demand predictability and

expected forecast accuracy. Hence, the items are classified according to demand variability in a way

that X items include all items in which use is relatively constant and fluctuates only rarely. The

probability of correct predictions is very high. Y items include all items with substantial fluctuations in

demand due to seasonal reasons or because of trends in product use. Z-products are all articles with

highly irregular use. The reliability of predictions in this case is low.

Hence, to get more insight into the behavior and sensitivity of the demand pattern of each item, the

items will be clustered or grouped based on the combination of two classifications (ABC & XYZ

classifications). Thus, nine different categories (classes) of items will be generated (XA, XB, XC, YA, YB,

YC, ZA, ZB, and ZC). From this category, due to the fact that products are mostly books, the category

CZ will be out of scope, as it refers to items that are slow moving products with very high demand

variability for which company X follows the strategy “Make to order”. This is a business production

strategy that typically allows consumers to purchase products that are customized to their

specifications. The Make to order strategy only manufactures the end product once the customer

places the order. Hence, the CZ items are not stored at all and that is why they remain out of scope

concerning the selection process. Thus, there are 8 different categories of products.

Varying Parameters: Inventory replenishment policy- dependent

The inventory replenishment policy- dependent parameters that were estimated as relevant to the

existent model and to the frame of this research are the following:

The type of replenishment policy (the (s,S) policy or the Min/max policy)

Preferred type of replenishment frequency (either through intervals or quantities)

Regarding the aforementioned, the main parameter is the type of inventory replenishment policy

that will be either the (s,S) policy or the Min/max policy. Moreover, the preferred type of

replenishment frequency are characteristics of every type of inventory replenishment policy and at

this research it gets the values that correspond to the given values of the two inventory

replenishment policies.

Methodology to Build the Scenarios

The scenarios are built after coming to the decision on which of the aforementioned parameters vary

and how. This way certain products will be selected with varying characteristics. For the product

dependent parameters, the method with which they will vary is to choose a number of products that

have very different properties in terms of slow or fast movement, of demand volatility and demand

predictability. Regarding the inventory replenishment policy- dependent parameters that were

defined above, the plan is to see how those products behave according to the two inventory

replenishment policies.

Page 27: Towards the Optimal Inventory Review Intervals

27

More specifically, due to the classifications that were mentioned above, it is possible for the products

to be distinguished by categorizing them into 9 categories, from which the 8 are relevant to this

research. These are the items that belong to XA, XB, XC, YA, YB, YC, ZA, and ZB. Hence the product

selection will be performed based on such criteria. In addition, two different inventory

replenishment policies the old and the new one, will be investigated in this research. Moreover, six

IRIs are selected to be investigated. Hence, the methodology for selecting products and building the

scenarios leads to the formation of 12 scenarios.

Scenario Analysis

Scenario building is another methodology that is applied in order to use the simulation model. In

fact, building scenarios is the way that the simulation model will be used, to test those scenarios and

answer to the relevant research questions. A scenario analysis is a method in which possible future

events are thought of in order to be able to analyze possible outcomes. It is not the purpose of a

scenario analysis to present an exact picture of the future. Instead, it presents several alternative

future developments or combinations of future developments. This methodology will be used to

explain and analyze the results of the 12 scenarios that are built.

2.7.4 IDEF0, system theory, black box theory The conceptual design of the simulation is developed using IDEF0. The conceptual model is

considered as a system, and hence, a system diagram is developed in IDEF0. IDEF0 stands for

Integrated Definition for Function Modeling and is designed to model the decisions, actions, and

activities of an organization or system. Moreover based on system theory, the system is perceived as

a black box. In general, a black box model is a conceptual system that has no direct relationv with

the construction and operation of the concrete system that it models.

2.7.5 DEMO methodology DEMO methodology DEMO stands for Design & Engineering Methodology for Organizations. In

general DEMO is a methodology for designing and engineering organizations. Its main goal is to align

the design and development processes to the core processes of an organization. This methodology is

selected to open up the system that describes company X’s case.

DEMO: the theory behind the methodology

DEMO stands for Design & Engineering Methodology for Organizations. In general DEMO is a

methodology for designing and engineering organizations. Its main goal is to align the design and

development processes to the core processes of an organization. To achieve that, DEMO abstracts

away from the detailed description of each process and focuses on the generic concepts and roles

(Dietz, 2006). DEMO was first introduced by Jan Dietz in the early 90s and then evolved into a

methodology which can represent a coherent, comprehensive, consistent, concise and essential

conceptual model of the organization (Dietz, 2006).

DEMO relies on a sound theory which identifies the principals and definitions of the system and also

the entities within that system. This theory defines the world, the existing entities in it and the

behavior of these entities along with their interdependencies. According to this theory that is the

core milestone of DEMO methodology, each system is identified by a set of elements interacting with

each other and with the elements in the environment. The environment itself is composed of the

same type of elements. By referring to elements in DEMO, it is meant human beings who perform

specific tasks in the system and have specific responsibilities and are known by the role they play.

Page 28: Towards the Optimal Inventory Review Intervals

28

Thus, the core elements forming the system and the environment in DEMO are the actor roles (Dietz,

2006).

Actor roles are able to perform certain types of acts. They have the ability to interact with each other

by performing coordination acts. Every coordination act is performed by two actors, the performer

and the addressee. By performing a coordination act the performer informs the other party

regarding his intention towards a production. In coordination acts, actor roles can request, promise

to deliver, question or declare a production. Production is the result of a production act which is

performed by one actor role. Production will be delivered to the environment or another actor role

inside the system who has requested that production (Dietz, 2006).

The productions are categorized in three different layers. The layers differentiate form each other by

the level of intellect used for producing that production. The highest layer is called ontological layer

in which the production is an innovation or a decision. In the second layer, the Infological layer, the

level of intellect is reduced to only interpreting the data and producing information out of data (No

new innovative idea is created in this layer). The lowest layer is called the Datalogical layer and its

production is regarding data or documents.

Every action is done based on an agreement between two actor roles through a series of

negotiations. This process is called a transaction. Essentially, a transaction is composed of several

coordination acts revolving around performing one production act. Thus, transactions become

unique and identifiable by their productions. Since the productions are associated with layers, each

transaction can also be associated with layers. Hence, a transaction can be categorized as an

ontological, infological or a datalogical transaction. Moreover, sometimes the execution of a

production act is dependent on the production of another transaction and thus, the actors in a

transaction may have to wait for some other actors to finish their transactions before they can

proceed with their own (Dietz, 2006). A transaction between two actors can be seen in the following

figure.

Figure 4: The actor roles and their binding transaction (Janssen, 2016)

Moreover, the figure shows an actor role (A) that initiates the transaction by making a request. Actor

role B is the one that delivers the product (the black dot). The transaction is as follows: A makes a

request, B makes a promise, B provides the product, and B performs the state (for instance he states

the product to be ready, handing it over to the customer), and A accepts the offer. Thus, a

transaction consists of three phases: the order phase (request and promise), the execution phase

(the product is created) and the result phase (state and accept). The person who makes a request is

the initiator of the transaction. The person who makes a promise is the executor.

Overall, this methodology has some core concepts such as the actors, their roles and relationships

and the transactions among them. At that point, it should be noted that this is the main reason for

choosing this methodology over other similar ones to dive into the analysis of the processes in

company X. More specifically, there are other relevant methodologies that can be used, for instance

Page 29: Towards the Optimal Inventory Review Intervals

29

BPM methodologies that stand for Business Process Management. BPM however, does not have a

sound and standard theoretical basis like DEMO has. On the contrary, it is mostly founded on

practical experiences (Smart, et al., 2009). Thus, the correctness of the final results of such a

methodology cannot be validated and different experts in that methodology may not come up with

the same results after applying it to a problem. Further, another reason for choosing DEMO over

BPM is the fact that BPM methodology BPM only focuses on the functionality and the end result of

each process while it ignores the interaction of actors in an organization contrasted to DEMO.

Furthermore, the unique feature of displaying this cooperation in the form of a separate transaction

unlike flowcharts or other business process management practices is that it presents a more

complete picture. More specifically, it indicates which actor role starts the transaction, the name of

the actor role, the name of the transaction, the actor role which is tasked with providing the product

and the name of that actor role (Janssen, 2016). Hence, even if both methodologies are used to

explain organizations, in that case the approach of DEMO is selected to be applied for the

aforementioned reasons.

DEMO: The way of modeling

The models that can be built using DEMO methodology are essentially a representation of the

concepts discussed in the previous sub chapter that introduced the sound theory that DEMO is based

on. DEMO can represent an organization in four partial models: the construction model, the process

model, the action model and the state model.

The construction model (CM) specifies the identified transaction types and the associated actor roles,

as well as the information links between the actor roles and the information banks (the collective

name for production banks and coordination banks). It is considered to be the most concise model

compared to the other models and there is nothing above it. The CM consists of an interaction model

(IAM) and an interstriction model (ISM). The interaction model (IAM) specifies the actor roles in the

organization and the transactions that take place between them. The IAM shows the active

influencing relationship between actor roles. Moreover, the interaction structure of an organization

consists of the transaction types in which the identified actor roles participate as initiator or

executor. For this it provides an Actor Transaction Diagram (ATD) and a Transaction Result Table

(Dietz, 2006). Further, the interstriction model (ISM) The interstriction model (ISM) constitutes the

right side of the CM. The ISM specifies the relationship between the actor roles in the organization

and the information banks used by them. The ISM shows the passive influencing relationship

between actor roles. The interstriction structure of an organization consists of the information links

between actor roles and coordination and production banks. The ISM provides an Actor Bank

Diagram (ABD) and a Bank Contents Table (BCT). When they are merged, the ATD and ABD are called

the Organization Construction Diagram (OCD).

2.8 Outline of report The report consists of several chapters. In this first chapter, an introduction along with relevant

background information is provided. The second chapter presents the research methods that are

used for this report. The following chapters are used to answer the research questions.

Hence, in the third chapter a supply chain analysis is presented. First, the analysis starts with

performing a literature review that involves the summary, collation and/or synthesis of existing

research. The performance of this literature review is implemented based on the top-down

approach. The forth chapter includes a case study that is presented and performed to scope the

Page 30: Towards the Optimal Inventory Review Intervals

30

research in the frames of company X. Chapter 5 discusses the development of both the conceptual

model and the simulation model for inventory management and control of company X. After defining

the right modeling approach, several modeling cycles are discussed which result in testing the model

and identifying the impact on the defined performance measures.

Further, chapter 6 discusses the results from the simulation and draw conclusions. In chapter 7,

limitations, opportunities for further research and reflection on the research are elaborated on.

Figure 5 shows a visual representation of the thesis outline.

Figure 5: Structure of the thesis

2.9 Concluding remarks In this chapter, the knowledge gap and problem statement of this research were identified and

presented and subsequently. Further, the research objective is to investigate the impact that the IRI

has, on the supply chain performance of Company X. Moreover, the research will reflect its

relevance both scientifically and socially. The research questions were presented in this chapter and

thus a preliminary conceptual model of research was illustrated that sums the main research

question: “What are the effects of review intervals on the supply chain performance of company X?”

Further, the research approach was tackled. Subsequently, a literature review, an embedded case

study strategy, simulation, IDEFO, system theory, black box theory and DEMO methodology were the

core research methods introduced as appropriate to conduct this research. The chapter ends with

the presentation of this report’s outline.

Part 1: Introductio

n

Part 2:

Research problem

and Approach

Part 3: Supply Chain

Analysis

Part 4:

Case study Analysis

Part 5: Inventory Simulation

Model

Part 6:

Conclusions

Part 7:

Limitations, Recommen

dations, Reflection

Page 31: Towards the Optimal Inventory Review Intervals

31

3 SUPPLY CHAIN ANALYSIS Analysis starts with performing a literature review that involves the summary, collation and/or

synthesis of existing research. The performance of this literature review will be implemented based

on the top-down approach. The top- down approach is a strategy of information processing and

knowledge ordering used in a variety of fields. Hence, a research framework is developed to conduct

the literature review according to which the analysis starts by describing broader notions that aim to

lead to more specific ones that are relevant to the research objective of this thesis. Afterwards, a

case study is presented and performed to scope the research.

3.1 Research framework for literature review For the conduction of this research, a literature review is firstly necessary to combine and present

the information that was reviewed and to further develop the initial causal conceptual model that

has been formulated. The literature review was conducted with a top-down approach, meaning that

literature was reviewed starting from a broad perspective and then continuing with narrowing down

gradually the theories that the author came across until reaching the objectives. In this case, the

objectives represent the main variables: The independent variable is the IRI and the dependent

variable is supply chain performance. Under this scope, the following framework illustrated in figure

6, presents the building of the conceptual model.

Figure 6: Literature review conduction and further development of the causal conceptual model.

3.2 Inventory Theory Inventories are kept at almost all parties within a supply chain. One reason why this happens is

uncertainty (Waters, 2003). In general, inventory theory deals with mathematical theories of

inventory and production. This field of study constitutes a subspecialty within the fields of operations

research and operations management and above all, it concerns the design of inventory and

production systems aiming to minimize costs. Furthermore, this field includes the decision making

process of firms regarding manufacturing, warehousing, supply chains, spare parts allocation and so

on. Inventory theory covers both the field of Inventory management and inventory control.

Page 32: Towards the Optimal Inventory Review Intervals

32

Usually the terms “inventory” and “stock” are used to express the same thing. (Wild, 2002) However,

when it comes to inventory management there can be essential distinctions. Stock is in brief an

amount of goods that is being kept at a specific place, for instance in warehouse, and sometimes it

can also be referred to as inventory. On the other hand, inventory management refers primarily to

the processes and the decisions that should be made for specifying the size and placement of

stocked goods. Inventory management is necessary at different locations within a firm or within

multiple locations of a supply chain and aims in managing the production in a way that will not allow

the running out of materials or goods.

The scope of inventory management is broader than stock. Regarding its definition, inventory

management can be referred to as the “management of materials in motion and at rest”. (Coyle, et

al., 2003) The following activities all fall within the range of inventory management: control of lead

times, carrying costs of inventory, asset management, inventory forecasting, inventory valuation,

inventory visibility, future inventory price forecasting, physical inventory, available physical space for

inventory, quality management, replenishment, returns and defective goods and demand

forecasting.

According to Reid & Sanders (2007), inventory management basically serves two main goals. The first

goal is the availability of goods. It is essential for all the operations in process that the required

materials or goods are present in the right quantities, quality and at the right time in order to deliver

the required service level. The second goal is to achieve the aforementioned service level against

optimal costs. This leads to the inevitable challenge, that is, to find the suitable equilibrium between

keeping sufficient inventory with the optimal costs with respect to the desired service level that one

has to deliver. In that sense, not all items can be held in stock against every cost and therefore

choices have to be made.

3.2.1 Reasons for focusing on inventory management According to Harrington (1996), inventory often is related to the most significant costs a firm faces.

With regard to the literature, there are three reasons that explain the importance of focusing on

inventory management. Those are: costs, risks and the higher possibility to identify and to cope with

those risks.

Regarding costs, according to Goor & Weijers (1998) stocks are responsible for up to about one third

of the total working capital costs. Moreover, inventory costs represent a big part of the overall

logistics costs (Coyle, et al., 2003). As a consequence, there are important benefits that can be gained

by reducing these costs. Further, according to literature (Wild, 2002; Fawcett, et al., 2007), working

capital invested in stocks could be a useful resource if it was used differently. That is why from a

company-perspective, capital invested in stocks can be considered as a waste of money. Generally,

cost reductions are required from the market in order for firms to be competitive and as illustrated,

reducing the working capital costs by performing efficient inventory management is one way to

achieve that.

As far as the risks are concerned, according to literature (Visser & Goor, 2004; Fawcett, et al., 2007),

keeping stock is related to risk as there are events that if they occur, they could influence negatively

the business processes. For instance, inventory could catch fire, or be stolen, lost or become

obsolete. Such events could even block the production process and this subsequently would lead to

late deliveries, hence, lower service levels and customer dissatisfaction. Risks caused by keeping

stock are interrelated with costs, because in order to maintain stock secure to prevent risky events

from happening, firms invest on inventory management.

Page 33: Towards the Optimal Inventory Review Intervals

33

The third reason for focusing on inventory management is that inventory costs are costs that can be

easily identified and encountered compared to other costs that are not that easy to identify and

reduce in a supply chain (Johnson & Pyke, 2001)

3.2.2 Why keeping inventories? As mentioned, keeping inventory always comes with costs and also entails possible risks. According

to Waters (2003)

“Stocks are expensive, because of the costs of tied-up capital, warehousing, protection, deterioration,

loss, insurance, packaging, administration and so on”.

However there are some main reasons why inventories are essential and even inevitable to maintain.

First, uncertainty is an important reason explaining why most parties within a supply chain keep

inventories. Inventories can function as a buffer to cope with uncertainties (Waters, 2003). Further,

uncertainty stems from various causes. For example, there may be a mistake that leads to wrong

goods delivered or damaged goods even though the delivery is on time. In the case of the uncertainty

of having a delivery not on time, it is also necessary to keep a stock to prevent a stock-out situation.

This stock is called safety stock for the aforementioned reason. On the other hand, it is not only the

supply parties that deal with uncertainty issues. The demand side as well may constitute a source of

uncertainty as well in a supply chain. According to Wild (2002), it is hard to predict the expected

orders by the customers. Hence, in order to be able to deliver on time and reach the desired service

level, it is of crucial importance to keep a stock to meet demand while preventing disruptions due to

uncertainty. In sum, maintaining stock is a way to cope with variation and uncertainty in both the

upstream and downstream parts of a supply chain and to handle possible problems that may arise

during the supply chain operations (Waters, 2003).

Another reason for keeping inventories is time related issues. Time can also constitute a root of

uncertainty regarding inventories. Nevertheless, time lags that exist in a supply chain can be covered

by keeping stock. The term for better describe those time lags is “lead times”. An order that is placed

usually requires an amount of time in order for the goods to be delivered. During this period of time

inventory functions as a buffer to cover the required demand. Moreover, not only time lags

concerning the deliveries can lead to significant fluctuations, but also they can be magnified

downwards a supply chain provoking the so called Bullwhip effect (Lee, et al., 1997; Fawcett, et al.,

2007; Johnson & Pyke, 2001). Consequently, keeping inventory is a way for an organization to protect

itself against this effect.

Finally, in terms of cost reduction, in some cases keeping stock is cheaper that the opposite. In

addition to the former, economies of scale constitute an exemplary reason why organizations keep

inventories. In that sense, buying bigger amounts of quantity from a supplier is in general more

advantageous than buying smaller ones because of discounts that are made (Waters, 2003; Coyle, et

al., 2003). Moreover, ordering less but more frequently increases the logistics costs. Further, price

fluctuation may be an additional reason to decide to keep stock, as ordering products from a supplier

at a relatively low price is beneficial for a firm (Waters, 2003). However, this should be considered in

case of overall benefit, meaning that if the total costs of keeping additional inventory is more

beneficial t than buying at a higher price because stocking costs have to be considered as well.

3.2.3 Inventory in the supply chain There are three main types of inventory that one can come across in a supply chain. These are

namely:

Page 34: Towards the Optimal Inventory Review Intervals

34

1. work-in-progress (WIP) inventory

2. pipeline inventory

3. finished inventory

According to Silver & Peterson (1985), the need of inventory control focuses on one of the

aforementioned types of inventory and depends on the type of business to be analyzed.

3.2.4 Inventory strategies As illustrated previously, keeping inventory costs and also can be a risk or uncertainty source. On the other hand, the reasons why one should keep inventory were previously presented as well. Inventory management can even lead to additional profits or overall cost reduction due to economies of scale. Hence, the important question that still remains unanswered is what is the optimal amount of inventory that once should keep in stock? Essentially, this still remains a question that needs various specifications upfront in order to be answered and needs to be considered as a separate, identical and individual issue one should encounter. As Wild (2002) also proposed, the amount of inventory that should one keep depends on various factors and upon the activity responsible for defining the stock.

3.2.5 Business operations model In addition to the corporate strategy, an organization chooses a business operations model that is

aligned with their strategy. The business operations model is also referred to as basic structures or

strategies. The four most common structures are the following (Hoekstra & Romme, 1991; Dijk, et al.,

2007) :

1. Make-to-stock: In this strategy, products are manufactured regardless of any order placed

by a client. It is even possible that products are produced, but there is no demand at all. In

this structure the pressure is on the sales department when the demand drops and is highly

reliant on production according to forecasts (Olhager, 2003).

2. Deliver-from-stock: This structure resembles the make-to-stock structure but in this case the

assortment is much bigger and products are not being manufactured first. This structure is

usually found at a wholesaler or retailer. In additional literature (Olhager, 2003), apart from

this strategy another one is also mentioned, the engineer-to-order strategy that is

characterized by the firm engineering the product after the receipt of an order.

3. Assemble-to-order: There are manufacturing companies that only assemble. They combine

different components based on the desired configuration of the customers. Using a limited

number of components, they are able to produce various different end-products. However,

final assembly is delayed until after orders are received.

4. Make-to-order. A make to order strategy is most common for products that have to be

tailored to the consumers’ requirements. In this case, products are only produced upon

customers’ demand.

Figure 7: Different product delivery strategies relate to different order penetration points (Olhager, 2003).

Page 35: Towards the Optimal Inventory Review Intervals

35

Figure 7 illustrates the aforementioned business structures- strategies that were discussed. The

dotted lines depict the production activities that are forecast-driven, whereas the straight lines

depict customer-order-driven activities. The OPP stands for the order penetration point and refers to

the stage in the production process where customer orders are accepted by the manufacturer

(Olhager, 2003). Other authors refer to the OPP with the term Push/Pull-boundary (Simchi-Levi, et

al., 2003).

Different perspectives on stock level

Parties in a supply chain tend to acquire different perceptions when It comes to inventory

management depending on their role in the supply network. Different perspectives on that matter

render the ideal stock levels that one should keep ambiguous. Management often aims to cost

reduction to be competitive in the markets and also to enhance the firm’s profits. It is already

mentioned why keeping high levels of inventories costs significantly, not to mention the working

capital that is consumed for that and could be alternatively used.

On the other hand, a sales perspective could imply high service level to be the objective. More

specifically, in order to provide best service to the customer a sales department could follow the

strategy of always having high stock levels rather than risking having a running-out of stock situation.

Moreover, a manager of a warehouse could have another point of view on the matter. If for example

the price that he buys the raw materials from the suppliers is one of the important criteria of his

work evaluation by his superior department, then probably his strategy of purchasing would align

with buying in large quantities in order to have discounts to reduce the total purchase costs.

However, in such a case, it might happen that stocks levels are too high and their preservation at the

warehouse is translated to additional costs.

3.2.6 Introduction to inventory control Inventory control is the set of activities that coordinate purchasing, manufacturing, and distribution

to maximize the availability of raw materials for manufacturing or the availability of finished goods

for customers. (Wild, 2002) In other words, inventory control problem is in brief the problem that

every firm faces when decisions must be taken regarding how much to order in each time period in

order to meet demand for its products. Usually, such problems can be modeled using mathematical

techniques of optimal control, dynamic programming and network optimization and this field of

study constitutes a sub part of inventory theory.

There are basically three questions that seek for answers from the use of inventory control theory

(Waters, 2003):

1. What items should we keep in stock? To answer to this question warehouse managers

should evaluate the advantages and disadvantages of keeping an item in a warehouse before

they add it to the inventory as they should only consider keeping it only in case of emerging

benefits.

2. When should we place an order? This question can be answered in three ways: First,

determine a fixed interval time and then place an order of a product in the needed quantity

at each interval. In this approach, order quantity may vary from one time period to another

(a fixed time period). Second, monitor the stock level continuously. When it falls to a specific

level, place an order with a fixed size. In this approach, the time periods between two orders

may vary (a fixed-order quantity, FOQ). Third, with respect to known demand over a specific

time period, enough stock should be ordered. In this approach, both time period and order

quantity depend on demand and may vary (Farahani, et al., 2011)

3. How much should we order? Different costs play a crucial role when answering this question.

For instance, if the order size is large and the time period between ordering is long, the cost

Page 36: Towards the Optimal Inventory Review Intervals

36

of ordering and inventory shipment decreases, but inventory holding costs increase. If on the

other hand the order size is small and the time period is short, then inventory holding costs

decrease but two other costs increase. Thus, it is important that there is a trade-off between

the related costs.

Inventory aggregation level

When it comes to making inventory management decisions, there has to be an aggregate stock that

can be measured and manipulated on the level of an individual product. This level of distinction is

called a stock keeping unit (s.k.u.). An s.k.u is defined as an item of stock that is completely specified

as to function, style, size, color and location (Silver & Peterson, 1985).

Safety stock

Safety stock is stock that is kept because of uncertainties of demand, lead time, or both. The required

service level determines the quantity of safety stock kept. In a sense, a higher service level needs

more safety stock. For example, if the demand is higher or lead time is longer than expected, then

the safety stock will compensate for this variation (Johnson, et al., 1999).

Lead time

Lead time (LT) is defined as the period between placing a replenishment order and the time it is

actually received. Lead time can occur due to various reasons (Farahani, et al., 2011). For example,

order preparation takes time before sending the raw material to a supplier. Moreover, order

processing and delivery preparation are two business processes from the supplier side that also

require time. Further there is the time after the supplier will deliver the materials to be received by

the next actor in the supply chain.

Demand

Every company’s management is challenged to deal with the seemingly conflicting objectives of low costs, high production quality, and a fast response to customer demand. The levels of demand influence the choices of inventory strategies. Hence, it is necessary to include demand as a determinant for inventory related decisions. Demand can be known, deterministic but it can also be uncertain. However, it is not unpredictable due to the fact that most companies operate the process of demand planning (or forecasting of demand) in order to be able to meet the needs of their customers. Customer Demand Planning (CDP) is a business-planning process that enables manufacturers to develop demand forecasts as input to service planning processes, production planning, inventory planning and revenue planning (Krajewski, et al., 2007).

Backlog

Backlog refers to any order for a product or service that is accumulating as a result of being delayed

and not being able to be met on time. It is important to mention at this point that there can be a

backlog only if the product or service has an expected time of delivery. In case of no expected time of

delivery, a product would never be delayed, and hence, there is no backlog (Anon., 2015).

Demand forecasting

Demand variability is one of the most common sources of variability in a supply chain. Customer

order behavior is mostly uncertain, not easy to predict and much harder to influence compared to

production variability. For example, regarding functional products, they usually show a more stable

and predictable demand, longer product life cycles, lower product variety, less stock-outs, low stock-

out costs, higher volume per SKU (stock keeping unit), and low obsolescence if they are compared to

innovative products (Cachon & Fisher, 1997).

Page 37: Towards the Optimal Inventory Review Intervals

37

This differentiation gives a good overview of aspects concerning demand as demand variability could

be caused due to randomness or due to management decisions. If a classical B2B relationship is

considered, the procedure indicates that the manufacturer will receive the customer orders and will

deliver the goods without sharing further information. This means that the manufacturer needs to be

prepared to fulfill uncertain customer demand. Hence, planning activities are based on demand

forecasting. The manufacturers that operate with the Make to stock (MTS) and Make to order (MTO)

structure rely highly on demand forecasting. Moreover, when it comes to demand variability,

another cause can be management decisions. More specifically, the product portfolio is a significant

source of variability. For instance, it is common logic that the more products and the more product

variety a firm has, the higher the variability.

Forecast techniques can be classified into two groups: First the qualitative forecasting methods and

second the quantitative forecasting methods. Regarding the qualitative forecasting methods and

their use, they are selected if historical data is not available and if in order to predict future demand

relies on the knowledge of experts. Such techniques are the Delphi method that is defined as opinion

consensus of a group of experts (Dalkey & Helmer, 1963) and judgmental forecasting (Gaur, et al.,

2007).The second group of methods is usually used in B2B environments that frequently deal with

functional goods and quantitative models can be designed and implemented with the existence of

historical data. The models usually belong to one of the following two groups: The first group is

causal models and the second refers to time series models. The first category focuses on predicting

demand for a specific product as a function of related parameters. On the other hand, time series

models rely on past demand data in order to predict future demand, thus, they use past values of the

same parameters.

The importance of forecast error

Although forecasting is a key process to plan the demand, there are still cases where uncertainty

cannot be completely eliminated. In addition, the law of forecasting: “Forecasts are always wrong”

(Hopp & M.L, 2007) indicates the importance to measure the forecast error. The main goal of

forecasting, thus, is to minimize this error. The forecast error is briefly the difference between the

predicted and the real value. In literature, various types of forecast error techniques are found. . For

instance the MAPE stands for mean absolute percent error is a forecast error measurement

technique. The most relevant ones for inventory management are presented in Table 1 (Everette &

Gardner, 1990; Hyndman & Koehler, 2006)

Table 1: Types of forecast error calculations

3.2.7 Inventory classification Inventory in any organization can run in thousands of part numbers or classifications and millions of

part numbers in quantity. Therefore, inventory is required to be classified with some logic in order to

be managed efficiently and correctly. There exist various important inventory categorization

techniques that are used by warehouse managers in order to have their inventories classified. The

Page 38: Towards the Optimal Inventory Review Intervals

38

most common ones to be used are the ABC classification and XYZ classification, and when combined

the ABC-XYZ classification is formed.

ABC classification

ABC classification is a practical approach applied for multi-item inventory problems. (Ballou, 2004)

(Elsayed & Boucher, 1994) Different item classifications need different replenishment and inventory

control policies. Different types of items also need different ways of replenishment. Some authors

(Sylver , et al., 1998)propose the ABC classification, based on the perception that items with a large

turnover (A-items) need to be controlled differently compared with items with a low turnover (C-

items). Usually the ABC classification is based on annual expenditure. Moreover in this classification

approach, the items are classified based on demand (consumption) rate: Inventory control is based

on a form of Pareto analysis. The inventory items are divided into three categories (A, B, and C),

according to a criterion such as revenue generation, turnover, or value. Typically, 'A' items represent

20 percent in terms of quantity and 75 to 80 percent in terms of the value. Example of the ABC

classification, with respect to the Pareto law, could be the following:

A items: 20% of the items accounts for 70% of the annual consumption value of the items.

B items: 30% of the items accounts for 25% of the annual consumption value of the items.

C items: 50% of the items accounts for 5% of the annual consumption value of the items.

These are the slow moving products.

In general the A products are considered as fast movers. Fast movers usually are more vulnerable to

a probable stock out situation and thus they order more often. On the contrary, products that belong

to the third category, the C products, are considered to be the slow movers and most of the times

represent the stock that companies must carry that doesn't turn over so quickly.

XYZ classification

An ABC-analysis is not sufficient for optimized stocking as the accuracy of its predictions and

therefore the optimization of the ordering process are affected by fluctuations in product demand

and product price. Any observation of the frequency of product use over a period of time will show

that the quantity consumed is relatively constant for some products, while the demand for other

products may vary and other products are used very irregularly. To handle these fluctuations in the

optimization process, the probabilities of product use patterns predicted must be considered. As in

the case of the ABC analysis, these probabilities can be grouped by a process commonly known as an

XYZ-analysis. (Clevert, et al., 2007) Hence the items are classified according to demand variability in a

way that:

X items include all items in which use is relatively constant and fluctuates only rarely. The

probability of correct predictions is very high. (high forecast accuracy expected)

Y items include all items with substantial fluctuations in demand due to seasonal reasons or

because of trends in product use. The probability of correct predictions is medium.

Z-products finally are all articles with highly irregular use. The reliability of predictions in this

case is low. (low forecast accuracy expected)

This classification method is also based on Pareto’s Law which states that “in most situations 85% of

the total effect is due to 15% of the cause sources”. This is also the basis of the 80/20 rule.

Page 39: Towards the Optimal Inventory Review Intervals

39

3.2.8 Inventory control in certain conditions

The EOQ model

The simplest model that exists in the literature and has been widely used as it is or with variations to

depict how inventory is controlled in certain conditions is the economic order quantity (EOQ). The

classic EOQ introduced by Ford W. Harris in 1915 and is a simple model that illustrates the trade-offs

between ordering and storage costs (Farahani, et al., 2011). The model has the following

assumptions (Farahani, et al., 2011):

There is a known, continuous, and constant demand.

Costs are known and constant.

Shortages are not permitted.

The lead time between placing and receiving orders is zero, and replenishment time can be

ignored.

Whereas:

EOQ = optimal order quantity

D = annual demand quantity

K = fixed cost per order, setup cost (not per unit, typically cost of ordering and shipping and

handling. This is not the cost of goods)

h = annual holding cost per unit, also known as carrying cost or storage cost (capital cost,

warehouse space, refrigeration, insurance, etc. usually not related to the unit production

cost)

The Min/Max replenishment policy

The Min/Max replenishment policy usually uses the EOQ model. The Min/Max policy is an inventory

replenishment method that has been implemented in a variety of inventory management software

so far due its simplicity. The Min value depicts a stock level that triggers an order and the Max value

depicts the new targeted stock level that follows the order. The difference (Max-Min) is frequently

interpreted as the EOQ. Figure 8 presented below, illustrates an inventory operated by the Min/Max

policy.

Page 40: Towards the Optimal Inventory Review Intervals

40

Figure 8: Inventory operated with the Min/Max replenishment policy. (Vermorel, 2014)

3.2.9 Inventory control in uncertain conditions: stochastic inventory models The basic assumption of the EOQ model mentioned above is determinant demand and lead time. An

uncertain parameter is one that does not have an exact quantity but its probability distribution is

known (Waters, 2003).

The uncertainty in demand or lead-time depicts almost every real-life inventory management

situation. For example, the retailer wants enough supply to satisfy customer demands, but ordering

too much increases holding costs and bears risk. An order too small increases the risk of lost sales

and unsatisfied customers. Similarly, an operations manager sets a master production schedule

considering the imprecise nature of forecasts of future demand and the uncertain lead time of the

manufacturing process. Such cases are common and usually they cannot be dealt by using

deterministic approach due to uncertainty. Hence, a decision maker faced with uncertainty does not

act in the same way as the one who operates with perfect knowledge of the future. That is why

stochastic inventory models are used when uncertainty is present. The main difference to the

deterministic models that exist is that demand here is stochastic. Three different inventory

replenishment policies will be discussed: the (s, Q) replenishment policy, the (s,S) policy and the (R, S)

policy (Jensen & J.F., 2003).

The (s, Q) replenishment policy

The (s, Q) replenishment policy is alternatively called the reorder point, order quantity system. Figure

9 shows the inventory pattern determined by the (s, Q) inventory policy. The model assumes that the

inventory level is observed at all times. This is called continuous review. When the level declines to

some specified reorder point, s, an order is placed for quantity Q. The order arrives to replenish the

inventory after a lead time, L.

Page 41: Towards the Optimal Inventory Review Intervals

41

Figure 9: Inventory operated with the (s, Q) replenishment policy (Jensen & J.F., 2003)

The values of s and Q are the two decisions required to implement the replenishment policy. The

lead time is assumed known and constant. The only uncertainty is associated with demand. In Figure

9 it is shown that the decrease in inventory level between replenishments is a straight line, but in

reality the inventory decreases in a stepwise and uneven fashion due to the discrete and random

nature of the demand process (Jensen & J.F., 2003).

The (s, S) replenishment policy

The (s, S) policy is characterized by a fixed re-order point for a variable product quantity. With an

interval of M, the inventory position is monitored in order to make a replenishment decision. When

the inventory position is below s, a variable amount of Q is ordered in such a way that, after a lead

time T the inventory position is raised to a value between s and s + Q (Janssen, et al., 1998). Figure 10

illustrates an inventory operated by this policy.

Figure 10: Inventory operated with the (s, S) replenishment policy

The (R, S) replenishment policy

A different way to manage a stochastic inventory system is to follow the (R, S) replenishment policy.

This is also called a periodic review policy in the sense that the inventory level is only observed at

time intervals of R. If the inventory is at level y, a quantity S – y is ordered to bring the inventory

Page 42: Towards the Optimal Inventory Review Intervals

42

position to S. S is called the order level. After a lead time interval L, the replenishment order is

delivered. The analysis of this policy is much like that for the (s, Q) policy. For the (s, Q) policy, the

reorder point s is set to protect against the possibility of shortage during the lead time L. For the (R,

S) policy, the order level S is set to protect against a shortage in the time interval R. In the event of a

particular order at time t, the lowest inventory that is affected by that order occurs at time t + R. The

quantity S must be large enough to keep the probability of a shortage in that time interval small. The

advantage of this policy is that it does not require continuous review. (Jensen & J.F., 2003) However,

the (R, S) policy is much more affected by variability than the (s, Q) policy because of the longer

interval.

3.2.10 The independent variable: the inventory review interval The IRI refers to the frequency regarding how often to review the inventory in order to determine

when to place the next order. In other words, a review period is the length of the interval between

two consecutive inventory reviews. It is observed in both literature and real life cases that this

frequency varies from replenishment policy to replenishment policy and from company to company.

With respect to the inventory time interval, there are two kinds of inventory reviews: continuous

review and periodic review. A periodic replenishment policy indicates periodic review intervals. A

continuous replenishment policy in fact lacks of predetermined IRIs due to the real time inventory

review that is implemented continuously. However, a continuous inventory replenishment policy can

sometimes be considered as a periodic one, with a very small IRI, for example one day.

3.2.11 The dependent variable: Supply chain performance The goal of a supply chain is in general to deliver the required performance depending on the

business requirements of any organization. More specifically, in case of inventory management, the

goal is usually twofold, how to minimize the costs and at the same time achieve the required service

level according to the organization’s policies and strategies and to the clients’ needs. Further, supply

chain performance can have many definitions and each and every one of them should be tailored

according to the requirements of the case to be analyzed. In order to define supply chain

performance in every separate case, one has to consider what he wants to measure. Measurement is

important because it affects every behavior that impacts supply chain performance. Hence,

performance measurement provides the means by which an organization can assess whether its

supply chain has improved or degraded. There is a variety of supply chain performance metrics in

literature and many companies use them or adapt them based to their own needs in order to

measure the performance of their supply chain. Different perspectives of Supply Chain Performance

Measures are cost and non-cost perspective, strategic, tactical or operational perspective, business

process perspective and financial perspective. In the literature, the most common metrics that a

researcher will find are the following: key performance indicators (KPIs), return on investment (ROI),

net present value (NPV), balanced scorecard (BSC) and critical success factors (CSFs). While it is

subjective to the decision maker which metric to use, it can also be stated that KPIs are a key metric

compared to the other measurement techniques.

3.2.12 The key performance indicators (KPIs) Key Performance Indicators (KPIs) are quantifiable measurements that are decided beforehand from

an organization in order to provide the reflection of its critical success factors. KPIs are metrics which

are usually defined and measured over a period of time or during a specific time interval. KPIs were

first introduced in 1961 by D. Ronald Daniel and Jack F. Rockart. A KPI generally has to follow the

SMART principle: A KPI needs to be specific, measurable, attainable and relevant, and time sensitive.

(Shahin & M.A, 2007).

Page 43: Towards the Optimal Inventory Review Intervals

43

Regarding inventory management and control, there are plenty KPIs that are widely used in in order

to measure performance. Examples of those are:

Inventory turnover: Inventory Turn tells you how many times inventory has been sold and

replaced in a given period – calculated as Sales divided by average inventory value.

Stock outs: Stock outs indicate where a demand cannot be met due to the absence of the

required inventory.

Service Level or product availability: This KPI refers to the ability of a firm to fill a customer’s

order from available inventory. The following are among the several ways to measure

product availability: The product fill rate is defined as the fraction of demand that is satisfied

from stock. The order fill rate is defined as the fraction of orders that are filled from stock.

The cycle service level is the fraction of replenishment cycles that end with no shortage.

Hence, there are various KPIs that are used when inventory performance needs to be measured

whether focusing on the economics or performance of the inventory.

3.3 Concluding remarks To sum up, a literature review is performed with a top down approach. Hence, starting from general

notions such as inventory management the author narrows down the literature in a systemic way

until the IRI notion is reached. The author presents the IRI as a key variable for this research along

with the theories that are tied to the term. Hence, the reasons why one should focus on inventory

management are explained along with the reasons of keeping inventories. Moreover, selecting an

inventory strategy is tackled. Further, the available operations models are reviewed, such as the

Make-to–stock and the Deliver –to-order and also different perspectives on stock level are examined.

Inventory theory entails as well inventory control and therefore relevant notions are reviewed such

as aggregation level, safety stock, lead time, demand and its forecast, and backlogs. Additionally,

classification of inventories is presented and analyzed. Hence, ABC and XYZ categorization are

reviewed. Moreover, regarding supply chain performance measurement, the role of KPIs is also

presented. In essence, the firs sub question of this research is already answered in this chapter as the

theories that are relevant regarding the IRI and the reasons of focusing on the IRI were identified and

presented in this chapter.

Page 44: Towards the Optimal Inventory Review Intervals

44

Page 45: Towards the Optimal Inventory Review Intervals

45

4 CASE STUDY ANALYSIS This chapter starts with the selection of case study strategy and the justification of this choice.

Subsequently, the case study is performed. Hence, the current and future replenishment policies of

company X are presented. Further, detailed descriptions of the logic for the current and the future

replenishment policies are performed in both cases of having a periodic or a continuous review

expressed in discrete steps accompanied by respective graphs that depict those steps. Moreover, the

data that is used for the case study is described and the linkage between the literature review and

the case study with the conceptual design of the simulation takes place in the end of the chapter.

Finally some concluding remarks are presented.

4.1 Selection of case study strategy: embedded single case study This research is approached as a single embedded case study and the reasons of this identification

are further elaborated in this chapter. The following figure presents the different types and

structures of case studies ‘design including the selected one for this research.

Figure 11: The types of case study design adapted from Yin (2009).

First, the core method used in this research is the conduction of a case study as the research will

consists of company X’s case. Case study research is used because it is particularly suitable for

studying a phenomenon in its real-life context, and when the phenomena are intertwined with the

situation (Yin, 2009). The problem defined is on the one hand, a problem that can be considered by

many companies. However, in order to quantify it and perform an analysis, it has been chosen to do

so from the scope of a specific company, company X. The case study method allows the researcher to

retain the holistic and meaningful characteristics of real-life events (Yin, 2009). Moreover, it is not

possible to identify the effect of the IRI on supply chain performance on the basis of highly

aggregated data when every company is unique because in every separate case there is a different

set of parameters and variables that depend on the company’s strategies, characteristics and policies

Page 46: Towards the Optimal Inventory Review Intervals

46

that are followed. Further, the results will be more accurate and meaningful not only from a

technical perspective for this specific company, but also form a social perspective by contributing in

solving a real-life problem of a company with similar characteristics as company X.

For the purpose of this research, the author will use a single-case study design to confirm, challenge,

and broaden existing theory (Yin, 2009). A single case study focuses on only one case to be

thoroughly examined (Verschuren & Doorewaard, 2010). This will be the case of Company X.

Moreover, with respect to a case study research design, the researcher should identify if the single

case study is a holistic or an embedded one. A holistic case study means studying a case in its totality

and it is used when no logical subunits can be identified and a study might be conducted on a too

abstract level. On the other hand, in an embedded case study, various subcase or units or processes

are studied (Verschuren & Doorewaard, 2010). A characteristic of an embedded case study is that

extensive analysis is performed and also it might focus too much on subunits, thus, losing higher level

(holistic) aspects. Regarding to the characteristics of this specific case study, it is considered by the

author to be an embedded single case study.

This case is not examined with a holistic perspective but in fact there are sub-units related to the IRI

such as demand, forecast of the demand, and different categories of inventory replenishment

policies that lead to different processes and to products with different characteristics. Hence,

extensive analysis will be performed in order to reach the research objective and thus, this is the

reason for using the research strategy of an embedded case study. The three sub-units of this case

study are: the products, the two inventory replenishment policies concerning company X and the IRIs

that are going to be researched. First, every product is different in terms of demand, forecasted

demand, ABC-XYZ categorization, lead time of supplier or supplier ID. Hence, the first unit of research

that allows the combination of all these factors is the product unit of analysis. Further, the two

replenishment policies are not seen as multiple case studies but as a sub-unit of the same analysis.

The two policies are driven by various factors such as, demand, forecast, inventory management data

etc., a fact that led the author to seek for relevant data during data collection. Additionally, the IRI is

driven by many of the aforementioned factors as well. Thus, the data were acquired and analyzed

with regard to the three sub-units of this case study.

To conclude, it is shown that this research is identified as a single embedded case study with three

sub-units of analysis, as illustrated in figure 12.

Figure 12: Presentation of the case study design for this research with the three sub-units of analysis.

Page 47: Towards the Optimal Inventory Review Intervals

47

4.2 Case study performance In order to perform the case study it is crucial that the project of company X is thoroughly discussed.

This will be performed by combining all the elements of the case study with the literature review that

has been performed in the 1st step of the research framework.

The company selected: company X

Company X is a multinational publishing and education company. It is one of the largest education

companies and book publishers in the world. Furthermore, the company has a large portfolio of

products.

SKU and ISBN

Decisions on inventory management have to be made on the level of an individual product. This is

called a stock keeping unit (SKU) which is defined as an item of stock that is completely specified as

to function, style, size, color and location (Silver & Peterson, 1985). The SKU is the unique number

and it can be perceived as a code assigned to a supplier's billable entities. Usually an SKU numbers

apart from the ISBN code because it is used for inventory control purposes rather than identifying

what an item's identity is.

Since this is a publishing company most products of company X is books. Hence, the distinction

among different products occurs by referring to each product by its ISBN number. ISBN is the number

the publisher assigns to an individual book title. It stands for International Standard Book Number.

The barcode corresponds to the ISBN number, so that when a book is scanned, the scanner reads the

number and identifies what the book title is. Without the ISBN number there would be no barcode.

Nevertheless, in this research when referring to product identification the ISBN numbers will be

used.

The Supply chain

Company X supply chain consists of a central warehouse in UK, 16 Litho-suppliers worldwide and 2

digital suppliers based in UK. Figure 13 maps the aforementioned situation from the suppliers to the

central warehouse in Rugby.

Page 48: Towards the Optimal Inventory Review Intervals

48

Figure 13: The 18 suppliers and the Central Warehouse of company X mapped.

From the central warehouse in Rugby, the books are distributed to 6 distribution points in UK as

illustrated in figure 14:

Figure 14:The Central Warehouse of company X and the 6 distribution points

4.2.1 Current and future replenishment policy of company X Company X’s current replenishment policy is similar to the (s,S) policy. Currently the inventory is

reviewed in a similar to a continuous way. At this inventory policy, when the inventory level on-hand

falls below a minimum, s, the site will generate a request for a replenishment order that will restore

the on-hand inventory to a target, or maximum, number, S. When using this policy, the Reorder

Point field is the minimum, or trigger level. The Reorder/Order up to quantity field is the maximum,

Page 49: Towards the Optimal Inventory Review Intervals

49

or the number to which the inventory level is restored. In the case of company X, in brief there is a

fixed frequency of ordering with order sizes based on expected demand between order moments.

However, according to consultants supplying their services to company X, company X is in a transition

phase of changing its current replenishment policy. The future policy to be implemented has the

form of a Min/Max policy. Min/Max policy is a special case of (s,S) policies (Goetschalckx, 2011)

Hence, the policy is not a completely different one. However, the review interval of the new policy

will be one day. Hence, it will be considered to be a continuous policy. As mentioned before, a

continuous policy means that inventory is reviewed continuously, thus, it can be considered as

periodic one with a very small IRI of one day for instance. In brief, the new policy will focus on having

a fixed ordering quantity, the EOQ, where order delivery moments are based on expected moment of

reaching the safety stocks.

Stakeholders and presentation of the project

Accenture is a management consulting, technology services and outsourcing company helping clients

become high-performance businesses and governments. Company X is currently one of the many

clients of Accenture. Accenture has been asked from company X to propose the optimal Economic

Order Quantities (EOQs) for the 18 suppliers. EOQ is a calculation that determines the most cost

effective quantity to order or produce by finding the point at which the combination of order cost

and carrying cost is the least. There is a specific calculation that has historically been known as EOQ,

but any calculation that uses similar logic would also be considered EOQ.

Moreover, the EOQs will be used within a new inventory replenishment policy that company X will

have. As mentioned before, now company X has an inventory replenishment policy similar to the

(s,S) policy. The new policy to be implemented is the Min/Max replenishment policy that is visualized

in Figure 15. At the Min/Max policy the EOQ can be simply calculated as the difference between the

Max and the Min. For the shake of facilitation, in this thesis the author will refer to “current policy”,

meaning the policy that is now implemented in company X and to “new policy”, meaning the policy

that company X plans to implement.

Figure 15: Min/Max policy

At this project conducted by Accenture, Macomi provides the simulation capabilities. Macomi is a

business consulting firm based in the Netherlands and specialized in simulation based decision

support and predictive analytics. Macomi is building a simulation model using their simulator

software, S3N. In general, S3N (Strategic Supply Network Simulator) lets you design and optimize a

supply chain. S3N can support the designing of a new supply chain or the assessment of the risks of

an existing one. Macomi is building the model based on data from company X to define the optimal

Page 50: Towards the Optimal Inventory Review Intervals

50

EOQs of the 18 suppliers. The model needs to have two behaviors: one according to the current

replenishment policy of company X (s,S policy) and one according to the new policy that is planned to

be implemented (the Min/Max policy).

Although during literature review and company X’s presentation in previous chapters have

introduced in theory the relevant inventory replenishment policies, it is important to understand

what happens in practice. For that reason the logic behind the current inventory replenishment

policy of company X and also the new one to be implemented is presented in the following

subchapters. Each subchapter illustrates each of the two inventory replenishment policies in two

cases with regard to the IRI: First, the logic of the current replenishment policy is presented in case of

having a periodic review and in a second case that is considered continuous. It should be noted here

that in most parts of the analysis in this thesis, the continuous inventory review can be considered as

a periodic review with a very small time interval, for example one day. Secondly, in the next

subchapter the logic of the new inventory replenishment policy is illustrated in two cases, again with

regard to the IRI. The first case illustrates the logic of this policy with a periodic inventory review and

the second presents the same policy being continuous.

4.2.2 Detailed description of the logic for current replenishment policy (the (s,S)

policy)

Current policy with periodic review

The logic of the current policy in the case of periodic review is depicted in the following steps for a

random product. Further, to understand the steps a figure has been designed to illustrate the

sequence of the steps.

Figure 16 illustrates the sequence of the steps in a diagram of economic stock versus time. The

definitions that are depicted in both the figure and the aforementioned steps are the following:

ES: ES stands for economic stock. Economic stock is the physical inventory plus

replenishments ordered but not yet received less items sold but not yet delivered. It

indicates the amount exposed to the risk of fall in prices. In this case, it expresses the stock

that company X has not promised yet to a customer.

SS: SS stands for safety stock and as discussed in previous chapters is the stock that is kept

because of uncertainties of demand, lead time, or both.

LT: LT stands for lead time

OQ: OQ stands for order quantity, meaning the quantity that will be ordered.

Interval: in this case the interval stands for the time between orders meaning the time

between days with consecutive orders for a product.

The steps of the logic are:

1. Calculate the economic stock (ES) at present.

2. Calculate the next delivery date (NDD). The NDD as depicted in the following figure can be

calculated by NDD= T1 + LT

3. Check: If the expected economic stock (ES) is below the safety stock (SS), then order now.

Else, do not order.

4. Determine the next delivery date (NDD).

5. Calculate the expected economic stock (ES).

6. Order the required quantity. (order quantity = SS – ES)

Page 51: Towards the Optimal Inventory Review Intervals

51

Figure 16: Logic behind the current replenishment policy with periodic inventory review for a random ISBN.

Current policy, continuous:

Likewise, the logic of this case is presented along with the logical steps in figure 17. The figure depicts

the sequence of steps. The MTBO stands for the time between days with consecutive orders for a

product.

Figure 17: Logic behind the current replenishment policy with continuous inventory review for a random ISBN.

The steps of the logic are:

1. Calculate the economic stock (ES) at present.

2. Calculate the next delivery date (NDD).

3. Calculate the expected economic stock (ES) at delivery date.

4. If the expected economic stock (ES) is below the safety stock (SS), then order.

5. Calculate the desired next delivery date (NDD) based on the MTBO.

6. Calculate expected stock at delivery date.

7. Order up to safety stock (SS).

Page 52: Towards the Optimal Inventory Review Intervals

52

4.2.3 Detailed description of the logic for new replenishment policy (Min/Max policy)

New policy with periodic review

The logic of the new policy in the case of periodic review is depicted in the following steps for a

random product accompanied by Figure 18. This case is presented but it is not company X’s direct

interest at the moment to focus on the periodic review mode of this policy.

This case entails the calculation of the economic order quantity (EOQ). EOQ is a fixed quantity.

Figure 18: Logic behind the new replenishment policy with periodic inventory review for a random ISBN.

New policy with continuous review (daily review)

This case of the Min/Max inventory review policy with continuous inventory review presented here is

the new policy that company X wants to implement. The logic behind the implementation of this

policy is depicted in the following graph along with the accompanied steps that show the sequence

of occurrences.

The steps of the logic are:

1. Calculate the economic stock (ES) at present.

2. Calculate the next delivery date (NDD) based on lead time (LT)

3. Calculate the expected economic stock (ES) at delivery date.

4. Check: If the expected economic stock (ES) is below or equal to the safety stock (SS), then

order the economic order quantity (EOQ). EOQ is a fixed quantity.

The steps are visualized in Figure 19 that follows.

Page 53: Towards the Optimal Inventory Review Intervals

53

Figure 19: Logic behind the new replenishment policy with continuous inventory review for a random ISBN

4.2.4 Data description Quantitative data were provided by Accenture in order to perform this research. The validation of

the data has been performed by Accenture because based on those data the simulation model was

built.

The data were Excel spreadsheets with different inventory information. The data were reviewed and

were categorized afterwards to the following categories:

1. List of products by ISBN

2. Forecast data and forecast accuracy data

3. Sales data

4. Inventory management data

5. Supplier mapping data

6. Lead times for suppliers data

4.3 Linking the literature review and the case study with the conceptual

design of simulation The approach that is followed in this research states that the literature review along with the case

study analysis combined together, help in conceptualizing the design of the simulation. This is

depicted in the followed figure. The literature review has been performed and the case study has

been thus implemented and analyzed in the case study chapter. With respect to the research

approach, the combination of the two aforementioned methodologies, the literature review along

with its analysis and the case study, results in the conceptualization of the simulation design.

Moreover, with applying an empirical research strategy as a case study, real world data acquired

from company X are integrated into a simulation model structure. Hence, combining case study

analysis and simulation modeling enables researchers to conduct case study driven experimental

theory-building and theory testing in a virtual world (Kapmeier, 2006). This is depicted in the

following figure.

Page 54: Towards the Optimal Inventory Review Intervals

54

.Figure 20: Linkage between the desk research framework (literature review), the case study performance and the conceptual design of simulation

As it is presented in the figure above, the conceptual design of the simulation is considered to be a

system. The elaboration on this system will be conducted in the following chapter.

4.4 Concluding remarks In this chapter, the case study performance was presented and analyzed. First, the selection among

the various types of the case study strategy was performed and the embedded single case study

strategy was selected as the appropriate type of strategy for this research. Further, the case study

was performed: The Company selected for this case study was illustrated in detail along with the

presentation of its supply chain. Moreover, detailed descriptions of the logic behind the current

replenishment policy (the (s,S) policy) and the new replenishment policy (Min/Max policy) took

place. Subsequently, the relevant data for conducting the case study were described. The chapter

ended with linking the literature review conducted in chapter 3 and the case study conduction with

the conceptual design of simulation that is presented in the next chapter.

Page 55: Towards the Optimal Inventory Review Intervals

55

5 INVENTORY SIMULATION MODEL In this part the development of the simulation model is described. The model first is conceptualized

by defining the model objectives, the design and how to describe it. Subsequently, the conceptual

design of the simulation is being specified and all its parts are analyzed using the appropriate

research methods. Afterwards, the verification and validation of the model follows. Next, the

scenarios to be tested with the simulation model are introduced. Hereafter, the simulation

experiments are designed and performed and in the end the results are presented.

5.1 Model conceptualization In order to conceptualize the model, first the model objectives and its boundaries need to be clearly

defined. Subsequently, with respect to the case study analysis that has scoped the research within

the frame of Company X, the conceptual model is formulated.

5.1.1 Model objectives In the first part of this thesis the problem has been defined, along with the research objective of this

thesis with the intension to reach to solutions by answering the research questions that have been

formulated. Moreover, the problem has been scoped in the case study chapter within the case study

analysis that places the boundaries on the specifications of company X. Hence, the objectives of this

model is to be able to give answers regarding how the IRIs affect supply chain performance in terms

of finished inventory and service level with respect to the current and the new inventory

replenishment policy.

5.1.2 Conceptual design of simulation The conceptual design of simulation will be developed by perceiving the model as a system. The

conceptual design is visualized as a system diagram. The objectives of such a system according to

Thissen et al (2008) are to:

1. define the system and borders and to identify what falls within the system and what outside,

2. define the structure and relations within the system,

3. define the output of the system as ‘outcomes of interest’ (criteria) and

4. define the relevant contextual factors (external factors)

The conceptual design of the simulation is presented in Figure 21 and will be elaborated further on.

The model is considered as a system, and hence, a system diagram is developed in IDEF0. IDEF0

stands for Integrated Definition for Function Modeling and is designed to model the decisions,

actions, and activities of an organization or system.

Figure 21: The conceptual design of the simulation presented as a system diagram.

Page 56: Towards the Optimal Inventory Review Intervals

56

5.1.3 Description of the conceptual design: system as a black box In the previous subchapter, the conceptual design of the simulation was developed and presented in

a figure as a system diagram. In this chapter the system diagram will be explained in detail. This will

be done step by step by applying the procedure explained by Thissen et al (2008). First the system is

defined along with its borders, what fall in and what out. In this case, the system represents the

model to be developped and it is percieved as a black box: In general, a black box model is a

conceptual system that has no direct relation with the construction and operation of the concrete

system that it models. An example of a black box is a control or management model of enterprises

(Dietz, 2006). Basically, in a black box only the interactions between the composition and the

environment are taken into account, but in an abstract way: they are represented as aggregated

values of input and output variables. Hence, one may make and use a black box model of a system

without knowing its construction and operation a black box (Dietz, 2006). A representation of a black

box can be seen in figure 22.

Figure 22: Representation of the black-box model (Dietz, 2006)

In this case, it is not yet clear or understood how the system is developped and what are the discrete

business processes that occur. Further, it is not yet clear how the inputs and the controls influence

the output of the model. Thus, the model at this phase is presented as a black box.

5.2 Specification of the conceptual model In this part, an insight will be given to the conceptual model described in the previous subchapter.

The aim of the specification step is hence, to feed all the model parameters with data, formulas and

all the necessary variables to acquire usable output. Hence, all the inputs, the external factors, the

control variables and the mechanisms are defined in detail as well as the outputs that are required to

perform this research. Furthermore, the system is perceived as a black box in the conceptualization

phase but in this specification phase it is “opened” using the DEMO methodology that is described

later in the chapter to acquire a better understanding on the actors, their actions and all the relevant

processes that happen and constitute the system.

5.2.1 The control variables : IRI and inventory replenishment policy The contorl variables are the variables that will be manipulated and varied in the model. Since the

purpose of this research is to investigate the impact of the IRI on supply chain performace, the

control variaboles that are identified in the case of company x are the inventory replenishment policy

and the IRI. The first control variable will be either the old inventory replenishment policy (the s,S

policy) or the new inventory replenishment policy that will be implemented on company X (the

Min/Max policy). The manipulation of the IRI being the seccond control variable, cannot be defined

yet because the model is a black box and it is not yet easy to propose a methodological variation of

this variable.

5.2.2 The mechanisms: the choice of the simulation archetype In order to define the simulation model, first the system should be defined and analyzed. At this

point, a literature research is conducted to investigate the status quo of possibilities for simulation

Page 57: Towards the Optimal Inventory Review Intervals

57

modelling. This research is essential as it has been shown that simulation is a satisfying tool for such

a dynamic and complex system that could generate various possible scenarios and simulation is a

cheap way to test such scenarios without having to bear the expenses that would derive from all the

different scenarios if they were performed in the real world system. Hence, the research derives to

the main types of simulation that one can choose from to test a supply chain model. According to

this information, a type of simulation is selected for this research and subsequently a discussion is

performed to check if the conditions of the selected type of simulation are met. These are described

below:

Discrete event simulation

Discrete event simulation is a tool selected in many cases in order to simulate supply chains. This

type of simulation is used when one is interested in simulating individual events in time. In this case,

the discrete events include randomness, hence, uncertainty as well. Moreover, the analysis of

discrete events that are stochastic and occur in a dynamic system over time is a characteristic of

discrete event simulation.

Spreadsheet simulation

Spreadsheet simulation is another type of simulation. It is considered to be as an easily accessible computer simulation method and that fact makes spreadsheet simulation more credible for managers and decision makers (Kleijnen, 2005). In addition, there are various types of spreadsheet simulation methods that are used for supply chain problems but not all deal with a dynamic environment. Also some of them have been criticized for too much simplicity and therefore unrealistic outputs (Kleijnen, 2005).

System dynamics

Another type of simulation is system dynamics. This is a type of simulation that is primarily used for modeling complex systems over time and there exist various computer software choices for systems dynamics simulation. An example of using system dynamics application is the investigation of the so called “bullwhip effect” (Kleijnen, 2005). Using this type of simulation in this case one can examine the dynamic behavior among real and target inventory level. In general, a system dynamic study is suitable to understanding the nonlinear behavior of complex systems over time using stocks and flows as well as internal feedback loops and time delays (Wolstenholme, 2003).

Business simulation games

Finally the popularity of business simulation games has been gradually increased over the years due to the development of Internet technologies and the augmentation of Internet’s users that led to the development of the games industry and particularly the business simulation. Hence, this type of simulation can be widely used from professional for a company’s purpose to students and academics for research purposes. Moreover, the human behavior is included in this type of simulation, hence, providing the users with the advantage of an enhanced possibility of expected results that reflect adequately reality. In order to acquire an orientation for a right choice of simulation among the simulation methods that were described, a table is presented that shows the possible behavior, the system behavior for this case and an explanation on this behavior.

Possible behavior System’s Behavior Explanation of choice

Static vs Dynamic Dynamic The evolution of inventory over time is

seen as a dynamic process.

Page 58: Towards the Optimal Inventory Review Intervals

58

Continuous vs Discrete Discrete The model changes in discrete time steps:

The need for replenishment is reviewed

in discrete time intervals, the material

flow, the customer orders and other

processes are discrete events.

Deterministic vs Stochastic Stochastic The average daily demand is stochastic as

there is no clear relationship between

time and its value.

Table 2: Presentation of possible behaviors of a system, the behavior of the system in this case and the explanation of its behavior.

IDEF0 can be used to analyze the functions (processes) the system performs and to record the

mechanisms by which these are done.The mechanism selected for this system hence, represents the

means by which the processes will be implemented through simulation. From the proposed

simulation types that are previously discussed and taking into account the system’s behavior

explained in the table presented above, the author arrived to the selection of discrete event

simulation as the most suitable simulation method to use in this case. Hence, a discrete simulation

software will be selected in order to build the computer simulation model.

Selection of the simulation environment: S3N simulation environment

Prior to the construction of the simulation model, the suitable program environment should be

identified and defined. Hence, the software first will be selected and subsequently the construction

of the model will be analyzed.

In order to select the appropriate software for building the simulation model, the author considered

the available software programs that exist to solve supply chain problems using simulation. Hence,

considering again the system that has been described and analyzed, a further definition on the

mechanisms of the system should be made as the mechanism is used as means to analyze the

system’s performance. From the literature research performed regarding the possibilities for

simulation modeling, the findings were used as indicators for explaining the reasons why the model

should be constructed using discrete event simulation. Therefore, the first step is here is the

selection of the appropriate software. Among the available software that exists, S3N was chosen to

be used for the construction of the computer program.

S3N is a scalable simulation environment for modeling supply networks and analyzing their dynamic

behavior. S3N focusses on the communication and coordination layer of a supply chain, as well as the

planning and physical layer. In other words, it is a cloud-based analytics platform used to model,

simulate, optimize and visualize end-to-end supply chains. It also supports strategic alignment of

tactical and operational supply chain management practices and provides a comprehensive view of

the supply chain including revenue, productivity, costs, cash flow and risks. It can also represent

coordination and incentive mechanisms between supply chain partners and enables the distinction

between planning and execution layers for obtaining better transparency. Due to the fact that one

can build stochastic models using S3N it should be noted that the platform captures variability and

uncertainty. Further, it is data-driven on the input and output. In short, it enables high fidelity

modeling of the supply network, what–if analysis of a variety of scenarios; simulation based

optimization and advanced business analytics (Anon., 2015).

S3N has some distinguishing features when compared to similar software (Anon., 2015). For instance,

S3N is built around actors with specific roles which intelligently manage their operational processes

Page 59: Towards the Optimal Inventory Review Intervals

59

and engage in transactions with each other. This allows for realistic simulation of different

coordination structures. Moreover, it enables customizability as both the physical supply chain and

the managerial intelligence of the actors can be fully customized.

S3N is structured in five separate layers: the physical, data, planning, coordination, and behavioral

layer (Anon., 2015). The first three layers are common to traditional supply chain modeling and

analysis software: they contain the elements needed to simulate the operational level of a supply

chain. The added and unique value of S3N can be found in the coordination and behavioral layers

that contain tactical planning and intelligence of actors. More specifically, the behavioral layer

contains the roles actors have and the intelligent behavior associated with it. It describes the

decisions an actor will make in specific situations, such as an external order request, the completion

of an internal production order or an unexpected disruption. In the coordination layer, the tactical

management is executed. In this layer, actors evaluate the outcomes of their operations and adjust

their planning. They engage transactions with other actors to coordinate their operations. Regarding

the first layers, In the planning layer the operational management is executed, such as production

and transportation scheduling, inventory and ordering management, forecasting etc. Furthermore,

the data layer contains all the information processes. Here information systems calculate, store and

communicate data. Finally, the physical layer contains the operational processes, such as production,

transportation, storage, handling, repackaging etc.

Moreover, S3N outputs every physical, communication and payment operation in a supply chain

(Anon., 2015). It provides standardized dashboards and enables the creation of a large amount of

KPIs that can be visualized in external business intelligence software. The KPIs are subdivided into

different groups: Service levels, Costs, Utilization, Inventories and Cycle times. The most relevant

Service levels KPIs are: % Fill rate and % Delivered full and on time. In additions, the most relevant

Inventories’ KPIs are: Inventory turnover rates, Average inventory carrying costs and Days of

inventory outstanding.

5.2.3 Inputs: data collection and data analysis Data has already been collected at the case study chapter. Hence, all the essential data that will be

needed have already been gathered by Accenture in pursue of the case study performance and also

the design of the system that will be used to build the simulation model. Furthermore, concerning

the validity of the data, the data provided to the author have been validated by Accenture

consultants that were engaged in the project of company X. This form of validation is known as

expert validity as it entails the review of the object to be validated by experts. This validation

occurred prior to the author’s data collection, hence, the data were considered sufficiently validated

and therefore, the author proceeded in the analysis of them for the sake of the case study

performance.

The inputs for the model that represent the external factors are the sales data, the productlist of the

books per ISBN, the forecast data for the same products along with their forecast accuracy figures,

the inventory management data as elaborated in the data collection chapter, the suppliers mapping

data and the lead times of the suppliers. Regarding the demand, it is also an input calculated in days

(daily demand) for each ISBN and it is based on the daily forecast input. There are two ways that the

forecast data can be inputted: Either by month or by specific date. With respect to the data that

Macomi acquired, the forecast was inputted by month, hence, the demand is created based on the

forecast and repetitive. Moreover, it should be clarified at this point that the costs (transportation

costs, poduct lanes related costs). A more comprehensive description of the specification of the data

and its analysis follows as it is crucial to define the inputs of the model to acquire usable outputs:

Page 60: Towards the Optimal Inventory Review Intervals

60

List of products by ISBN

The list of products per ISBN was provided to conduct this thesis. The initial list entailed not only

books but also CD’s. The total amount of products that was selected for this research after

performing data analysis was 748 products. All these products are books. The initial list of products

provided was much higher but after analyzing the data, it was concluded that for 748 products there

is all the necessary information to perform this research.

Forecast data and forecast accuracy data

Forecast data was provided by Accenture along with forecast accuracy figures for all products. The

forecast data included the forecasted demand in monthly values of demanded quantity per ISBN for

the period of 2015-2016.

The forecast accuracy figures were formulated based on the ABC-XYZ categorization of products.

Hence for six product categories (AX, AY, AZ, BX, BY, BZ) with regard to the ABC-XYZ classification, the

minimum MAPE was calculated. The following figure shows how the MAPE is usually calculated

(Everette & Gardner, 1990).

Figure 23: MAPE calculation (Everette & Gardner, 1990)

The MAPE was calculated as a percentage and as quantity of ISBNs per product category for the six

aforementioned categories.

Sales data

Sales data was provided for the period of 2012-2015 entailing the sales of products per ISBN in

monthly values of SKUs sold.

Inventory management data analysis

Inventory management data files were provided to the author that entailed useful information for all

products. All information provided was per ISBN. Moreover the current stock situation was included

for 2015, meaning the average stock, the current safety stock, the current service level, total cost, as

well as a report of the stock site with the monthly values of stock for each ISBN for the year 2015.

Further, information that was included was the product segmentation according to ABC-XYZ

classification for every ISBN and the supply strategy. Moreover the frequency of replenishment and

the supply strategy are also important data for this research.

Supply strategy

Regarding the supply strategy, most products were following the make to stock (MTS) strategy and

the relative process for those is that after having been printed by the suppliers, they are delivered in

the central warehouse in Rugby to be stocked. There are however products of company X that

according to the data, follow the make to order strategy (MTO) and the process differs in the sense

that suppliers print the books once the customer places the order. This strategy in general creates

additional wait time for the consumer to receive the products but at the same time allows for more

flexible customization compared to purchasing from retailers' shelves. In the sense that this research

is focused on inventory management and control, this thesis is only interested in products that are

stocked in the warehouse. Hence, the products that are selected for the analysis belong only to the

MTS supply strategy.

Page 61: Towards the Optimal Inventory Review Intervals

61

At this point, it should be clarified that company X does not print the books but only orders them,

and when the products are delivered to company X, they are stored accordingly in the warehouse

until their shipment to the customers. Hence, it is more correct to say that the supply strategy is

more likely an “order to stock” than a “make to stock” strategy. However, for reasons of simplicity

when the author refers to the company’s strategy, the MTO abbreviation will be used but in reality it

will express the “order to stock”. Following the same logic, the products that are left out of this

research ‘scope, are the ones that are supplied according to the MTO strategy. Nevertheless, in

reality, in the case of company X the MTO strategy is specifically an “order to order” strategy.

ABC-XYZ classification

The ABC-XYZ classification is performed at the inventory data by means of categorizing the products

into 8 categories: AX, AY, AZ, BX, BY, BZ, CX, and CY. The ABC classification is an indicator of slow or

fast movers and the XYZ classification is an indicator of the items ‘demand predictability and

expected forecast accuracy. The CZ category is not included due to the fact that it refers to slow

moving products (C) with a very high demand variability (Z) for which company X follows the strategy

make to order (MTO), thus, no reason to include them in the inventory keeping data.

Replenishment frequency

In addition, the replenishment frequency of each ISBN in months is calculated based on the

categorization of the product according to the ABC-XYZ segmentation. More specifically, the

frequency of the products is calculated in Figure 24:

If C AND Z frequency= 0 months (no replenishment) Print to order

If (A or B) AND (X or Y) frequency= 12 months (replenished once a year) Print to Stock

If C AND X frequency= 6 months (replenished twice a year) Print to Stock

If C AND Y frequency= 0 or 6 months (replenished 0 or 2 times a year) Print to Stock

If (A or B) AND Z frequency= 0 or 4 months (replenished 0 or 3 times a year Print to Stock

Figure 24: Presentation of the calculation of the frequency of replenishment of the produts

The table presents the frequency of replenishment for every possible combination of the ABC and

XYZ classification. Nevertheless, although according to theory the C products are often characterized

as slow movers, the replenishment frequency rule of company X suggests for C products to be

sometimes more often replenished than some A or B products. For instance, a CX product is

replenished twice a year but an AX product respectively only once a year. After reviewing the stock

situation for A products and C products, it can be assumed that the reason for this occurrence is the

fact that slow movers have lower stocks, hence, replenishing more often makes sense in order to

prevent an out of stock situation. It should be noted that this classification comes from the

company’s data. Hence, this is the way that company X is currently performing the ABC-XYZ

classification on its products.

Calculations based on forecast

Moreover, based on the forecast data the inventory management data provide the estimated cycle

stock, the minimum and maximum stock value, the target stock and the total costs for each ISBN.

Cycle stock

Cycle stock is the term that expresses the part of inventory available or planned to be available for

the normal demand during a given period excluding excess stock and safety stock. Cycle stock is

Page 62: Towards the Optimal Inventory Review Intervals

62

calculated in the provided data as: [(Average Monthly Forecasted demand) * (Frequency of

replenishment)].

Min and Max inventory values

The minimum inventory is considered to be the safety stock and the maximum inventory is a

calculation of the safety stock plus the cycle stock.

Target stock level

Target stock level refers to the level of inventory that is needed to satisfy all demand for a product

over a specific period of time. Target stock is estimated by calculating the average of the minimum

inventory value and the maximum inventory value for each ISBN.

Total costs

The total costs per ISBN are calculated by multiplying the price per piece of an ISBN with the target

stock level of this specific ISBN.

Supplier mapping data As mentioned before, there are in total 18 suppliers, 16 Litho-suppliers and two digital ones.

Moreover, according to the acquired data, there is one supplier per product assumed. The supplier

mapping indicates the origin of the suppliers and according to that there are 3 UK suppliers form

which the two are the Litho-suppliers, one from Canada, one from Spain, Italy, Germany, Slovakia,

Slovenia, one from Malaysia, one from Hong-Kong, one from Singapore, one from New Zealand, one

from U.A.E, one from South Africa and one from India.

Lead times for suppliers data

For each of the 18 suppliers the data regarding their lead times were also provided. The lead time for

each supplier is calculated in days and is the result of summing the print time in days and the

transportation time including customs in days. Table 3 shows the total lead time in days for the

suppliers.

Origin Total LT (Days)

UK PRODUCTION_Digital 10

UK PRODUCTION_Digital 10

UK PRODUCTION_Litho 22

GERMANY 26

SPAIN 26

SLOVAKIA 26

ITALY 26

SLOVENIA 26

AUSTRALIA 56

NEW ZEALAND 56

SOUTH AFRICA 56

CANADA 56

CHINA 66

MALAYSIA 66

SINGAPORE 66

HONG KONG 66

U.A.E. 77

INDIA 77

Page 63: Towards the Optimal Inventory Review Intervals

63

Table 3: Total lead times of the suppliers in days

Total lead times of the suppliers in days

5.2.4 Outputs: KPIs expressing finished inventory and service levels The outputs that will derive form the model are expected to depict supply chain performance. As it

has been explained in previous chapters, for the purpose of this research, the term supply chain

performance regards the measurement of finished inventory and service level for all selected

products per ISBN expressed by KPIs. As mentioned before, the KPIs in S3N are subdivided into

different groups: Service levels, Costs, Utilization, Inventories and Cycle times. For the purpose of this

research, two KPIs were found to be relevant and are selected as outputs of the simulation model.

More specifically, the two KPIs that are measured are:

1. % delivered on time vs requested: This KPI expressed the service level provided by company

X as it is essentially the calculation “Sales/ Initial demand”, meaning how much was sold

divided by how much was initially requested for ordering.

2. Inventory final product: This KPI measures the quantity of the products that are in stock by

outputting snapshots of the inventory as time goes by and is expressed in terms of product

quantities per ISBN.

5.2.5 Specification and analysis of the system: Opening the black box As it is described in the previous chapter, the system is defined as a black box because it is not yet

clear or understood how the system is developed and what are the discrete business processes that

occur.

Further, it is not yet clear how the inputs and the controls influence the output of the model. Hence,

the primary objective at this point is to select a suitable methodology in order to map the business

processes that are relevant to the inventory management and inventory control of company X. Thus,

a methodology called DEMO methodology will be used to open up the black box/the system and to

shed light on all the relevant business processes that occur and on how those are formed via the

input and through the controls in order to lead to the supply chain measurement outputs.

DEMO methodology is selected as an appropriate choice for opening up the system that is perceived

as a black box. First, the theory of DEMO should be explained in detail because it is important for the

reader to enable him getting a better understanding of this methodology. After having explained the

theoretical background of the DEMO methodology and the reason why it is a suitable methodology

to use in this case in chapter 2, the application of the methodology takes place.

5.2.5.1 Application of the DEMO methodology: building the construction model

Actors and actor roles

First step to apply the methodology is to define all the actors that are included in the model. In total

there are 20 actors: Those are the 18 different suppliers, company X and the customer. For reasons

of simplicity, due to the fact that the focus is not on the customer side, one virtual customer will be

considered to be the customer of company X, in a sense that the products are delivered from the one

central warehouse in UK to this virtual customer. Table 4 presents the actors in this case.

Actors

1st UK production Digital Supplier

2nd UK production Digital Supplier

Page 64: Towards the Optimal Inventory Review Intervals

64

Table 4: All the identified actors

According to DEMO methodology, assigning actor roles to people is a matter of implementing the

enterprise: one completely abstracts from it on the ontological level of understanding the enterprise.

An elementary actor role is an atomic amount of authority and responsibility. It produces exactly one

transaction, and can be a customer of zero, one, or more transactions. An actor role can also be

instead of elementary, composite. The composite actor role comprises a number of elementary actor

roles plus their interaction and interstriction relationships. It should be noted that by convention one

should draw environmental actor roles as composite actor roles, even if it happens that an actor role

is elementary (Dietz, 2006). The reason for doing this is that generally it is not known whether an

environmental actor role is elementary or composite. Moreover, it is also not important to know: the

interest lies in the kernel of the organization. The start of modeling begins by drawing this kernel as

one composite actor role. The resulting CM is usually referred to as the global CM of an organization.

The following table shows all the identified actor roles, both the composite and the elementary ones.

The elementary roles start with B-A (number of role) whereas the composite ones with B-CA

(number of role).

In this case the composite actors identified are three: the suppler role, the company X role and the

consumer role. All these three roles can be opened up to many elementary roles. As mentioned, the

environmental actors are always considered as composite. Company X is considered to be the kernel.

UK Production Litho Supplier

Supplier from Germany

Supplier from Spain

Supplier from Slovakia

Supplier from Italy

Supplier from Australia

Supplier from New Zealand

Supplier from Slovenia

Supplier from South Africa

Supplier from Canada

Supplier from China

Supplier from Singapore

Supplier from Malaysia

Supplier from Hong Kong

Supplier from U.A.E

Supplier from India

Company X

Virtual customer

Page 65: Towards the Optimal Inventory Review Intervals

65

Actor roles: Composite and Elementary

B-CA01 Virtual customer as consumer

B-A02 Company X as supplier

B-CA03 Virtual customer as demand provider

B-A04 Company X as forecaster

B-A05 Company X as inventory manager

B-A06 Company X as customer

B-CA7, …, B-CA24 1st UK production Digital Supplier, …., Supplier from India as supplier

B-CA25, …, B-CA43 1st UK production Digital Supplier, …., Supplier from India as producer

B-CA44, …, B-CA62 1st UK production Digital Supplier, …., Supplier from India as transporter

Table 5: Combined composite and elementary actors table

In the actor roles table, Table 5, it can be seen that the elementary roles all refer to company X roles.

The environmental roles are the composite roles that refer to all the roles of virtual customers and of

the 18 suppliers as well. Further the B- that all roles’ names start with refers to the fact that the actor

roles according to theory are B-actor roles. This refers to the three distinct aspect organizations (the

B-, I-, and D-organizations). According to the organizational theorem, an enterprise is a layered

nesting of three homogeneous aspect systems: the B-organization (from Business), the I-organization

(from Intellect) and the D-organization (from Documents). The relationships among them are that

the D-organization supports the I-organization, and the I-organization supports the B-organization.

The integration is established through the cohesive unity of the human being (Dietz, 2006).

Another remark regarding the actor roles presented in the previous table is that since the suppliers

are 18, as has been discussed, they are presented for the sake of simplicity as groups instead of

entities as one can observe when looking the last three rows of Table 5 that refer to the suppliers.

A more cumulative perspective is presented by simplifying the roles of the case and dividing all roles

to three composite roles: The supplier role, the company X kernel role and the Virtual customer role.

Table 6 shows how this division is implemented.

Cumulative actor roles

Virtual customer role B-CA01, B-CA03

Company X kernel B-A02, B-A04, B-A05, B-A06

Supplier role B-CA7, …, B-CA24, B-CA25, …, B-CA43, B-CA44, …, B-CA62

Table 6: Table of Cumulative actor roles

Transactions

Table 7 is the transaction-result table that shows all the transactions that can take place in the

system along with their results, given the actors that were previously presented.

Transaction Type Result Type

Page 66: Towards the Optimal Inventory Review Intervals

66

T01 Buy backlog R01 The backlog is bought

T02 Buy products R02 The products are bought

T03 Consume products R03 The products are consumed

T04 Define the demand R04 The demand is defined

T05 Forecast the demand R05 The demand is forecasted

T06 Manage inventories R06 The inventories are managed

T07 Produce R07 The products are produced

T08 Replenish products R08 The products are replenished

T09 Transport R09 The products/orders are transported

Table 7: Transaction result table (TRT)

As it has been previously stated, a transaction is triggered by an initiator and afterwards the executor

responds. Specifically, the initiator makes a request, the executor makes a promise, provides the

product, and performs the state (for instance he states the product to be ready, handing it over to

the customer), and finally the initiator accepts the offer. For getting a deeper insight in the

transactions presented in the previous table, a detailed description follows for each transaction.

Transaction “Buy backlog”

Backlog refers to any order of products that is accumulating as a result of being delayed and not

being able to be met on time. It is assumed that the products are always sold because of the backlog

existence; hence, there are no lost sales. This transaction is initiated by the virtual customer that in

this case has the role of consumer and is executed by company X that here has the role of supplier. In

this transaction the virtual customer being a consumer, he will look first which backlog he has and

then he will try to buy what is in the backlog. He first needs to buy the backlog and then proceed in

buying the product. This follows a FIFO process (first in first out). Once the virtual customer performs

the request, then company X, being the supplier, checks whether he has the amount of products

requested. Although it is possible that the products are in the backlog, however, another possibility is

that the products are still in the backlog but he doesn’t have them yet in the warehouse. Hence, in

the second case the supplier will decline the request. This case could occur if there is a big quantity of

backlog and the warehouse is out of stock, because in such an extreme case the request from the

customer would be “do you have the products?” but the supplier would have to decline the request.

In any other case, if the products exist, the executor will provide the products so the customer can

consume them.

Transaction “Buy products”

This transaction is initiated by the virtual customer being the consumer and is executed by company

X that has the role of the supplier. Essentially, every day the consumer will check the daily demand of

the products, how much for each product should be bought and then the transcaction buy products

is triggered. The virtual customer being the consumer will request for the quantity of producys he

needs, then company X being the supplier checks if he has the product and quantity requested. If he

has it, he will provide it; otherwise he will add it to the backlog. At the execution phase, the supplier

provides the product and at the state phase, if the customer asked the product and he got it, he will

consume the product.

Page 67: Towards the Optimal Inventory Review Intervals

67

This transaction refers to one of the last steps of the supply chain, where the service levels are can be

measured because in essence, the type of check “What was I asked for and what I was able to

deliver” is performed by company X. Moreover, there are no lost sales , hence when an order is

placed from the virtual customer being the consumer , if company X being the supplier declines the

request, then, it will always be delivered later because it will be added to the backlog.

It should be noted that the sequence of events of the process are as follows: First, the virtual

customer being the consumer, starts buying the backlog. Then the virtual customer buys the daily

demand of that day. So for instance, if company X being the supplier does not have products at all,

every day the demand of the day will be added to the backlog and it will be increased every day.

Hence, first the check on the backlog occurs, then the backlog is bought and after that the

transaction buy product is triggered.

Transaction “Consume products”

This transaction is initiated and executed by the same actor role, the consumer. It refers to the

process of consuming the products that are bought. The virtual customer initiates the transaction

being the consumer and executes as well the transaction. Since the concept of transactions is the

core characteristic of the DEMO methodology, some processes such as this specific one, need to be

described from a transaction perspective even if they constitutes a solo action. In these cases, the

different approach of DEMO application is that the initiator and the executor of the transaction is the

same person that has been assigned to a specific actor role. Hence, in this case, the actor virtual

customer when having the role of consumer, he performs all phases of the transaction (request,

promise, state and accept) by himself simply by requesting from himself to consume the products,

then to promise and state that he will and then to accept that he did that.

Transactions “Define demand” and “Forecast Demand”

These two transactions are technical transaction not relevant to the model itself but still they are

important transactions that help in preparing and using the demand data. More specifically, the

transaction define demand is performed to create the initial demand figures. This transaction is both

initiated and executed by the virtual customer having the role of a demand provider. In reality this is

not a real role with a meaning but it is needed to assign a role to apply the methodology. The

transaction Forecast demand is also a transaction initiated and executed by the same actor role.

More specifically the actor here is company X with the role of Forecaster.

Transaction “Manage inventories”

This transaction in one of the most important ones as it involves most relevant processes that have

been discussed in this document. The transaction is initiated and executed by the same actor role.

This specific transaction can have two versions. Since it lies on managing the inventories of the

warehouse of company X and since there are two inventory replenishment policies that are discussed

in this report, there should be two versions of this transaction, one for the current policy and one for

the Min/Max policy. Hence, there is the same transaction initiated and executed by the same actor,

that is the inventory manager, but he has different roles depending to the version that is examined.

In the scenario of the current (s,S) replenishment policy the actor gets the role of “inventory

manager for current (s,S) policy”, whereas in the second scenario of the Min/Max policy the actor

gets a different role named as “inventory manager for new Min/Max policy”. In essence, in both

cases in those transactions the inventory manager follows the logic that has already been described

in the chapter “Detailed description of the logic for current replenishment policy (the (s,S) policy)”

Page 68: Towards the Optimal Inventory Review Intervals

68

and in the chapter “Detailed description of the logic for new replenishment policy (Min/Max policy)”

respectively.

Transaction “Replenish products"

This transaction is performed 18 times. It is always initiated by company X having the role of the

customer and is executed by every different supplier out of the 18 suppliers, each of which gets the

role of “supplier”. Hence, the 18 different actors get the same role to execute this transaction. In this

transaction when company X as a customer requests for products then the ordering costs should be

calculated and here as well the product prices are considered based on the ordered quantities. In

the execution phase, the supplier actor role will produce the product and then transport it. Hence,

this transaction should trigger the additional transactions “Produce” and Transport”.

Transactions “Produce” and “Transport”

These two transactions are also considered as technical because they are useful for preparing and

using the data for production of goods and for their transportation to the warehouse of company X.

The transaction “Produce” is initiated and executed by the same actor role, the supplier. Since in the

case there are 18 suppliers, this transaction is performed 18 times for all the actors having the role of

producers. The same applies for the second transaction, the “Transport” transaction. It is both

initiated and executed by the same actor role being the supplier role. Hence it is also performed 18

times and each time a different supplier out of the 18 in total gets the role of the transporter for

being the initiator and the executor of the transaction.

Actor Transaction Diagrams

The Transaction Result Table (TRT) presented above is used to build up an Actor Transaction Diagram

(ATD). In this diagram, the transaction types with their initiators and executors are drawn. Two Actor

Transaction diagrams are drawn. The first one is based on a more cumulative perspective and is

named global ATD. In this global ATD, the transactions that company X (the kernel) executes for its

environment and that it initiates in its environment are depicted. More specifically, the actor roles of

the organization-company X are included in the kernel and are presented as a composite main actor

role named as “company X kernel”. Moreover, it has already been stated that the environmental

actor roles are considered by convention as composite roles. Hence, a very composite approach is

first implemented to draw the global ATD by considering two general composite environmental actor

roles: the supplier role and the customer role. The global ATD is presented in figure 25.

Page 69: Towards the Optimal Inventory Review Intervals

69

Figure 25: The global ATD modeled

For a better understanding of the system, a detailed ATD is built to identify the several actor roles

and responsibilities that can be defined within the organization and outside of it. Hence, the kernel

of company X is opened up and the relevant actors are presented as elementary actors that ca

execute only one transaction. The same thing is applied to the environmental actors as well, as both

the supplier role and the virtual customer’s role get various elementary roles. Because by convention

the environmental roles cannot be presented as elementary even if they are, the various supplier

roles and the various virtual customer roles will be depicted again as composite roles in the detailed

ATD, even if in reality they are elementary due to the fact that they are supposed to execute only the

transaction that is described. Figure 26 presents the detailed ATD of the company.

Page 70: Towards the Optimal Inventory Review Intervals

70

Figure 26: The detailed ATD modeled

5.3 Verification and Validation of the model Before using the simulation model to run the tests, it is necessary to perform assessments are for the

verification and validation of the model. The first assessment verifies that the conceptual model is

coded in the right way, whereas the second assessment assesses if the model behavior resembles

reality sufficiently.

5.3.1 Verification At this point, two pilot scenarios were performed, one for one of the 746 products and a second run

for all the 746 products in order to check the parameters and to make sure that they produce

reasonable results. They were approved as the behavior of the code was verified. Moreover, the pilot

runs gave an indication of how to determine the initialization period of the simulation and the

number of replications to be made for each scenario that runs. The aforementioned are discussed in

detail in the part of the execution of the experiments in the model set up paragraph. At this point, it

should be noted that a run is considered the same thing as a replication. For the sake of this analysis,

the author refers to one or more runs of a scenario or alternatively to one or more replications

meaning the same thing in both cases.

Pilot scenarios were essential in order to verify the model. In general, the verification process is

essential to ensure that the conceptual model is translated using the right coding into a

computerized model. In other words, verification answers the question "Have we built the model

right?" There are various ways to verify a simulation model: For instance, general good programming

Page 71: Towards the Optimal Inventory Review Intervals

71

practices, checking intermediate simulation outputs through tracing and statistical testing per

module, statistical testing of final simulation outputs against analytical results or animation can be

found in the literature as ways to verify simulation models (Kleijnen, 1995).

For the verification of this simulation model, two ways were used: a structured walkthrough of the

model was performed and then VBA debugging tool was used for verifying the simulation model.

More specifically, a structured walk through was performed in order to ensure that the conceptual

model and the input of the model were properly implemented on the computer (Hillston, 2003). The

focus was on the right use of formulas and parameters that were desired to vary and the

resemblance with the conceptual model. Moreover, the checking of all inputs was important to

ensure that the inputs of the simulation mode depicted the ones of the conceptual model. The

walkthrough showed that the model construction was accurate. Further, S3N is a simulation

environment whose construction was based on the fundamentals of Enterprise Ontology theory.

Enterprise Ontology is defined by Dietz as the implementation independent essence of an enterprise

(Dietz, 2006). The Enterprise Ontology has a strong theoretical foundation and further builds upon

the results from theories from philosophy, sociology and language, such as Habermas’ theory of

Communicative Action and the Language-Action Perspective. The methodology on the other hand

that was selected in this thesis in order to analyze the system and develop the conceptual simulation

model that at first is perceived as a black box is the DEMO methodology. The DEMO methodology

that was used to “open up” the black box, applies the Enterprise Ontology theory. According to Dietz

(2006), the strong theoretical foundation ensures that DEMO models can be claimed to be formally

correct.

The second way with which the simulation model was verified was the use of VBA debugger tool

during the model development face and when the model was finished. VBA was used in order to

ensure that the simulation language is properly used.

5.3.2 Validation While verification answers the question "Have we built the model right?” validation answers the

question "Have we built the right model?” More specifically, validation is the process needed to

demonstrate that the model is a reasonable representation of the actual system, meaning that it

reproduces system behavior with enough fidelity to satisfy analysis objectives (Hillston, 2003). There

are various ways that one can validate a simulation model.

For this specific model three types of validation were performed: Validation by comparison to other

models, expert validation and historical data validation. The validation techniques that were applied

to the model are described thereafter.

Validation by comparison to other models

Comparison to other models is one of the ways of validating a model. This technique indicates that

various results, for instance outputs of the simulation model being validated, are compared to results

of other models that have already been validated (Sargent, 2007).

In this case, the intention was to compare the model that Macomi built in order to calculate the EOQ

values of company X’s products to the author’s model in order to validate the behavior of the

author’s model. Macomi’s simulation model has already been validated: it has already been

acknowledged and agreed by company X that Macomi’s model is verified, validated and ready to be

used. Hence, primarily the author’s model should be set in a way that the outputs can be compared

to the outputs of Macomi’s model for validation purposes. The situation regarding Macomi’s model is

Page 72: Towards the Optimal Inventory Review Intervals

72

that it has been built based on two versions: the current inventory replenishment policy and also the

respective version for the new inventory replenishment policy that will be implemented in company

X. The version of the model that has already been validated is the version of the current inventory

replenishment policy because the model simulates the “as is” situation at company X. This model is

validated by experts and also by historical data validation. The simulation output was compared to

the historical data of the current situation: Real data for the year 2014, forecast data for the year

2015 for the (s,S) inventory replenishment policy whereas the IRI is one month. The version of the

new inventory policy has not yet been implemented in company X. Moreover, the only validation

that can derive from Macomi’s model regarding the version of the Min/Max inventory replenishment

policy is the confirmation that this policy will be implemented in company X and an expert validation

to done by company X and Accenture acknowledging that the behavior of the model seems

reasonable despite not being able to validate the model towards reality. Hence, it is indicated that

the author’s model will be validated by being compared to the version of Macomi’s model that

simulates company X’s situation with the current inventory review policy.

Essentially by comparing the two models, a validation on outcomes is aimed. Both models will be

simulated for one year in order to compare the behavior of inventories. The demand is expected to

be moderately different: More specifically, there is variability up to a logical extent. The reason for

that is that the starting point is different, hence, the demand that is been generated every month

slightly differs from the forecast because it is being randomized. It makes sense therefore to state

that if one makes multiple runs, every time the demand will be slightly different. Nevertheless, this

can be accepted because demand is randomized and further, the importance is focused on how the

inventories behave.

For this reason, the simulation model was set accordingly: the scenario was set to the current

replenishment inventory policy; the IRI was set to 21 days. The model perceives everything in

working days. Thus, one month is the amount of days without the weekends. The output from both

models that were compared was a KPI from S3N called Inventory final product. This KPI is expressed

in terms of product quantities per ISBN. Two scenarios were executed, one using the Macomi model

for the current inventory replenishment policy and one using the author’s model. The results showed

that the values are very similar. Hence, the author’s model is validated by the comparison with the

model of Macomi.

Expert validation

Expert validation is a form of face validity according to which experts that have knowledge about the

system, are questioned whether the model and /or its behavior are reasonable. Further an expert

can be asked about the logic of the conceptual model, whether he finds it correct and regarding the

relationships between inputs and outputs, whether they seem reasonable.

For this assessment, interviews are conducted within Macomi, with employees in different fields of

expertise to verify the correct behavior of the model. Each expert was asked to look at the dynamics

of the model by checking the specific parts of it and the excel outputs (tables and graphs). All

acknowledged that the model is valid.

Historical data validation

Historical data validation is a form of validation that requires that historical data that are information

regarding the real world system exist and therefore, can be used to determine if the model behaves

according to the real system. By comparing the simulation outputs to the historical data of the real

system, one can validate a simulation model.

Page 73: Towards the Optimal Inventory Review Intervals

73

The validation is performed based on real historical data from 2014 and forecast data for 2015 for

the current inventory circumstances in company X. Hence, the model of the author that is used for

validation purposes is set again on selecting the IRI to be 21 days (one month) for the current

inventory replenishment policy that company X uses. Furthermore, due to the fact that the product

list contains 748 products and it would be difficult to analyze manually the results of such a high

quantity of products, an analytical approach was selected in order to validate the model for all the

products. The initial idea was to select valuable parameters and compare for all products the output

from S3N to the respective parameters from the historical data, instead of just looking to every single

figure of the 748 products. Those parameters hence represent the validation indicators for this

model.

The validation indicators selected for this purpose are various: First, inventory related data are

compared such as the average inventory figures, the Min and the Max inventory values that were

reached and the ordering frequency of products in the real world and in the simulation world.

Secondly, a comparison of all the provided data sets (actual sales and forecasted demand data for

both the real and the simulation world) is performed by first discussing the relationships between

the different data sets that were acquired and then by performing the comparisons to result in the

model validation. All calculations that were required for analyzing the historical data were done in

Excel spreadsheets. Further, the simulation results were saved respectively in the database; hence,

their visualization is based on the calculations that are on the database. Also, the outputs are

exported in Excel spreadsheets in order to be easily analyzed and compared to the historical data.

Inventory data comparison for historical data validation

At this point, it was decided to perform comparisons between different sets of inventory data among

the system results and the simulation results. The comparisons that were conducted are the

following:

Comparison between the average inventory figures: By average inventories the author

means the calculation for every ISBN of the average of the 12 monthly inventory quantities

that are kept in stock for one year.

The Min and Max inventories: For every ISBN as it is already mentioned, the historical data

and also the simulation outputted KPI results provide monthly values of the products’

quantities in stock. Hence, the Min and Max inventories refer to the two of the 12 monthly

values that correspond to the minimum and the maximum values out of the 12 values for

every product representing the inventory quantity. Hence, the intention is to compare the

Min and the Max inventory value for every product in the real world to the respective values

that are outputted from the simulation. For better understanding of the compared results,

the second Min and Max values are also calculated. To find the second Min of every ISBN,

one simply looks up in which month of the year the second smaller inventory quantity is

noted and this corresponds to the second Min for this specific ISBN. The same rule applies to

finding the second Max value for every ISBN.

Comparison between the annual replenishment frequencies of products: By annual

replenishment frequency the following is meant: How often is a product ordered during one

year. The data that derive from company X provide this information. Further, for this purpose

the KPI “Delivered” of S3N was used and after calculations in Excel the frequencies of

ordering for all products in one year were obtained to perform the comparison.

The following paragraphs describe in more detail the aforementioned comparisons.

Page 74: Towards the Optimal Inventory Review Intervals

74

Comparing the average inventory figures

In order to compare the average inventory figures, first the average inventories of the real data were

calculated in Excel. Further, the KPI from S3N called “inventory final product” was exported for all

products and some calculations resulted in the average inventory values from the simulation.

More specifically, the inventory status data for all the products were for the year 2014 and for each

product there were 12 monthly values of inventory status provided. Hence for each product the

average inventory was calculated by summing up the 12 values and then dividing the amount by 12.

For the respective simulation values of average inventories, first the average inventory values were

exported in Excel for every ISBN. Then, the data were processed by summing for every ISBN the

amounts of inventory for every month so that every product has 12 monthly values of inventory

status. Then, a second calculation was performed for every product to estimate the average

inventory for one year by summing up the 12 monthly values per product and dividing the result by

12. As a result, for every ISBN the author could compare two values: one value that expressed the

average inventory according to the real data from the year 2014 and a second one accruing from the

calculations based on the simulation results.

This way the results appear in Excel in 3 columns: the first one containing the ISBN numbers, the

second one being the average inventories of the real data and one being the average inventories of

the simulation data. For a systematic comparison of the results, the percentage difference was

calculated for the two values. This indicator was selected as it expresses the difference between two

values when both values mean the same kind of thing, for instance when one value is not obviously

older or better than the other. The cumulative results of the comparison can be seen in Figure 27.

Figure 27: The distribution of percentage difference of average inventories

The graph presented shows the distribution of percentage differences between the real data and the

simulation values of average inventories. The results are satisfactory as it can be seen that for 515

out of 748 products, the percentage difference between the historical and the simulation average

inventory values fluctuates between 1-10 %. Overall, it can be estimated based on the illustrated

graph that for 683 out of 748 products, alternatively for the 91,3% of the products, the percentage

difference fluctuates between 0-30% and only for the rest and only for 8,7% of the products present

a percentage difference bigger than 30%. As a consequence, it can be said that a first step towards

the validation of the model has been implemented. Further, another conclusion that can be drawn

when inspecting these results is the fact that the variability in S3N is less than the variability in reality

for the average inventories.

Comparing the Min and Max inventory values

515

113 55

19 17 6 3 1 0 3 16

0-10% 10-20% 20-30% 30-40% 40-50% 50-60% 60-70% 70-80% 80-90% 90-100% >100%

Products

Percentage Difference ranges

Distribution of Percentage Difference of average inventories

Page 75: Towards the Optimal Inventory Review Intervals

75

The second validation indicators that were selected are the Min and the Max inventories values that

were reached for every ISBN during one year. For obtaining these values from the historical data for

the year 2014, the Min and the Max value of inventory status was found for every ISBN from the 12

monthly inventory status values that are recorded for every product. Moreover, the KPI from S3N

called “inventory final product” that was exported for calculating average inventories was also used

in order to find the Min and Max inventory values of the simulation results. For the comparison of

the Min and the Max values regarding the historical data and the simulation results data, after

calculating per ISBN the Min inventory value for reality and for simulation first and then the Max

value respectively, the percentage change between the two values for every ISBN was estimated. The

formula used to calculate the percentage change was:

(Absolute (Min of Inventory in S3N – Min of inventory in historical data)) / Min of Inventory in S3N.

In the two following graphs presented in Figures 28 and 29, the distribution of the percentage

changes for the Min inventory values and for the Max inventory values are presented.

Figure 28: The distribution of percentage change for Min inventory values

Figure 29: The distribution of percentage change for Max inventory values

It can be noticed that the results are not that satisfying. More specifically, it can be seen that

regarding the Min inventory values, only 36 out of 748 products have a percentage change between

0 and 10%. Moreover, 133 of the products have a percentage change that can be acceptable ranging

from 0 to 30%. It should also be noted that for 80 products the percentage change cannot be

Page 76: Towards the Optimal Inventory Review Intervals

76

calculated because of one of the two values being 0. Hence, it can be stated that if the errors due to

zero values are excluded, only 19, 97% of the products have a percentage change ranging from 0 to

30%, whereas the rest 80, 03 % of the products’ percentage changes fluctuate between 30% and 100

%, or are even higher than 100%.

As far as the Max inventory values are concerned, similarly in the second graph the distribution of

the percentage change for the Max inventory values are presented in order to draw meaningful

conclusions when comparing the real data to the respective simulation data. Particularly, it is shown

that the results are better compared to the first graphs’ results but still they are not satisfactory

enough. From an error inspection perspective, it is rational not to have impossible calculations as

zero values for Max inventory status would not make any sense. Further, it is estimated based on the

second figure that those 347 out of 748 products, or in other words 46, 51% of the products show a

percentage change that ranges between 0 and 30%. ON the other hand the rest 53, 49% of the

products’ percentage changes are bigger than 30% and for 31 products even higher than 100%.

To summarize, it is evident that the results from this partial analysis are not satisfactory and cannot

be used for validation purposes. Such significant differences between the Min and Max inventory

values could be translated to various statements. For instance, one reason that could explain the

differences between real data and the simulation data is that in reality it is possible that company X

deviates in practice from the inventory replenishment policy and hence, the real data differ from the

simulation data. Alternatively, it could be stated that the suppliers could constitute the problem in

the chain in a sense that if the suppliers do not deliver perfectly then the simulation results could be

different than the historical data due to suppliers’ delays. Another explanation for such an

occurrence could be the fact that the simulation model assumes no losses. Consequently, the

simulation results will present a perfect handled situation of no lost inventory whereas reality

contradicts this no losses scenario. Sometimes inventory gets lost. An example on such a case would

be to misplace inventory in wrong places. In that case, the inventory review wills record this deficit

and this amount will be ordered although it already exists due to the fact that it is placed elsewhere

by mistake. Then, if someone finds this misplaced inventory and moves it to the place that it should

be stored, all of the sudden a next inventory review would show excess inventory values. This could

lead to a much higher Max value of a monthly inventory status to be recorded after an inventory

review. However, the simulation would not include such an occurrence, thus, resulting in a possibly

significant percentage change between the two Max inventory values.

For this reason, the same process will be repeated for the Min and Max inventory values but this

time the second minimum and second maximum values will be selected in order to test if there is a

significant difference in the percentage changes between the real data and the simulation results.

Comparing the second Min and Max inventory values

The comparisons are performed again in Excel and the two graphs below present the distribution of

the percentage changes for the second Min inventory values and for the second Max inventory

values. In the same graphs, the first calculations on the Min and Max inventory values have also been

added in order to get a better insight on the differences between the percentage changes and to

understand if the results are reasonable enough.

Page 77: Towards the Optimal Inventory Review Intervals

77

Figure 30: The distribution of percentage change for the 1st Min and 2d Min inventory values

Figure 31: The distribution of percentage change for the 1st Max and 2d Max inventory values

Figures 30 and 31 are twofold. First they show the distribution of the percentage changes between

reality and simulation for the second minimum and maximum monthly inventory values. Second, the

initial unsatisfactory Min and Max inventory values are added in order to be compared. It is evident

by looking at the charts that the differences are not that significant if the second values of Min and

Max inventories are considered. In those charts, based on the values of the graph that represent the

second Min and second Max inventory values, it can be seen that only 194 products have a

percentage change regarding the Min value that increases up to 30%. Further, if the calculations that

are expressed as errors due to the zero values are excluded, then those products represent the 28,

49% of the product list. On the other hand, for 487 products this percentage change fluctuates

between 30 and 100%. Hence, it can be stated that this is an indication of a systematic error that

does not allow validation by using those values. Based on the fact that the initial simulation model

has already been validated and verified and the author’s version of the model has only been verified,

it can be assumed that the inventory data that were provided to Macomi are not reliable and thus,

they cannot be further used for validation purposes.

This can be easily explained by the more cumulative graphs that are presented below.

Page 78: Towards the Optimal Inventory Review Intervals

78

Figure 32: The cumulative distribution of percentage change for the 1st Min and 2d Min inventory values

Figure 33: The cumulative distribution of percentage change for the 1st Min and 2d Max inventory values

In Figures 32 and 33 presented above, one can draw the conclusion that since the percentage

changes between the Min and Max inventory values from historical data and Min and Max inventory

values from simulation are so high, the inventory data that have been used are considered as

unreliable. The Min and Max differences are much higher systematically so it can be stated that the

calculations are invalid systematically for all products. Hence, it can be concluded that even though

the average inventories’ comparison was satisfactory enough, however based on the Min and Max

inventory values comparison, the inventory data are wrong and unreliable.

Comparing the annual ordering frequency of products

After not being able to validate the model based on inventory figures, the author proceeded in

parametrically comparing real world and simulation world by comparing the annual ordering

frequency in simulation and in reality for every ISBN. The annual ordering frequency expresses how

many times per year is a product being ordered. In general, the best way to perform a validation is

on actual behavior which in this case is depicted by the volumes of inventories. Nevertheless, since

this is not possible anymore due to the fact that inventory data were proven to be unreliable, the

second way that the author thought of to validate the model is validation on parameters. Hence, the

ordering frequency of products was chosen for comparing the real and the simulation world.

Regarding the real world, the ordering frequencies for all ISBNs are provided form company X as a

time interval and thus, the annual ordering frequencies are calculated for all products. For obtaining

same parameter from S3N, a KPI is exported from S3N that contains the Sales and ordering costs.

Page 79: Towards the Optimal Inventory Review Intervals

79

From these data, the number of different dates is counted for each product. This number equals to

the number of annual orders of each product.

Figure 34 presents this exact comparison of ordering frequencies for both the real world and the

simulation world.

Figure 34: The comparison of annual ordering frequencies

It can be seen that the results deviate slightly. For example, there are 286 products that are ordered

once a year according to historical data and 284 products that have the same ordering frequency

according to S3N.On the contrary, 423 products seem to be ordered 3 times a year according to

historical data, but the same ordering frequency corresponds to only 190 products from the S3N

perspective. To get clearer insight regarding the results a distribution is drawn that depicts the

differences between the real and the simulation values. The distribution is shown in Figure 35.

Figure 35: The distribution of ordering frequencies’ differences

It can be observed from the figure that the dataset resembles a normal distribution, as most of the

observations seem to fall within of the mean. According to the figure, for 431 of the 748 products the

ordering frequencies of the historical data match the ordering frequencies that are exported from

S3N. Moreover, it can be seen that for 259 products the differences in frequencies are concentrated

around -1 and +1. Further only for one product, the difference between the ordering frequency in

S3N and in historical data was -3. Thus, it is estimated by the author that for most observations the

Page 80: Towards the Optimal Inventory Review Intervals

80

results are satisfactory especially when considering the fact that with simulation the results are a

little bit different for each run. For example, for a different number of replications or for a different

replication itself, it is reasonable to expect slight deviations regarding the simulation results.

Conclusions on validation on inventory data

It was shown that the comparison between average inventories for all products generated

reasonable results. However, when the Min and Max monthly value of inventory reached within a

year for every product was considered, the results were not satisfactory. Thus the author proceeded

on comparing the second highest and second lowest inventory values for all products. This second

Min and Max comparison of inventories between real and simulation world was also unsatisfactory

due to the significant differences. The conclusion at that point was that the inventory data used were

considered to be unreliable. Thus, they were not further used for validation purposes. To further

validate the model based on inventory figures, since the Min and Max inventory comparison was

proved to be invalid, the author proceeded on a parametric validation by comparing the ordering

frequency between the real and the simulation world, in other words, the amount of times that

every product is being ordered during a year in reality and in the simulation world. This process was

satisfactory enough as the results of comparison showed that for most products the ordering

frequency in reality either matched or was approximate to the ordering frequency of the simulation.

For clarification purposes, it should be mentioned that the initial model of Macomi, as well as the

model of the author used the “unreliable” inventory values only to set the initial inventories at the

simulation model. Considering that inventories are essentially the result of the simulation, expressed

by KPIs of S3N, this is the only point where the validity of the model could be influenced. However,

the programmers at Macomi changed the inventory values and instead, they gave some random

points between the values in the Min/Max policy. Moreover, before running the model a year of

warm up time is set, thus, this warm up time normalizes the inventories. For instance, if for a product

the inventory was initialized with a very high value, then due to the warm up time the value after one

year will normally be decreased and over time the effect of these initial inventory values is

diminished. In an extreme case where the values would be extremely high, the warm up time of one

year could be considered insufficient. However, it has been checked that the differences in the values

for this model are not that high, hence, there is not such a risk in this case. As a result, those wrong

inventory data influence neither the functionality not the validity of the model. To conclude, despite

this problem that was encountered, the model was validated on average inventories and on ordering

frequencies.

Sales and Forecast Data comparison for historical data validation

At that point we need to know what are the data provided and further if they can be validated, now

that it is known that the inventory historical data for the year 2014 are considered unreliable.

First, Figure 36 shows the variety of data provided or created and their relations. The data are two

KPIs from S3N, meaning the “deliveries” and the “initial demand” whereas the actual sales and the

forecast represent reality, as they constitute the historical data sets.

Page 81: Towards the Optimal Inventory Review Intervals

81

Figure 36: Variety of existent data and the comparisons that can be made.

The deliveries are a KPI that expresses the amount per product that are delivered from company X to

the customer. The initial demand is the amount per product that was initially requested from

company X to be delivered to the customer. The actual sales are the sales in reality, that company X

recorded according to the sales in 2014 towards the customer. Further, the forecast has been

performed by Accenture and depicts the forecasted orders for the next year, 2015. At this point the

relationships depicted in the figure among the various datasets are discussed thereafter.

1. Deliveries-Initial Demand comparison:

The first one is between the two KPIs from S3N, the Deliveries and the Initial Demand. A comparison

between those data shows how the service levels are in the model. Figure 37 shows the distribution

of percentage change between deliveries and initial demand.

Figure 37: The distribution of percentage change between deliveries and initial demand

The comparison is performed by calculating the percentage change among the two different data

sets from S3N and the results are satisfactory as 732 products that constitute the 98,26% of the

whole product list show a percentage change at most up to 30%.

2. Initial Demand- Forecast comparison:

This relationship expresses essentially how well the simulation model uses the forecasted demand. It

has already been stated that the forecast is an input in the simulation model and that demand is

generated according to the forecast. Hence, it can be understood that if the forecast data is correctly

732

13

Percentage change 0-30% Percentage change >30%

Distribution of Percentage Change between Deliveries and Initial Demand

Number of products

Page 82: Towards the Optimal Inventory Review Intervals

82

inputted, then the KPI initial demand shows reasonable results. Macomi has already performed this

step as the forecast data has already been inputted in Macomi’s validated model and the results of

the KPIs were verified. For the sake of the analysis however, comparison between those different

data sets is performed at this point. For this comparison the Absolute percentage error was

calculated per ISBN between the annual initial demand and the annual forecasted demand for all

products. The cumulative results are illustrated in the following figure that presents the distribution

of the absolute percentage error calculations for all the products.

Figure 38:The distribution of absolute percentage errors for all products when comparing initial demand to forecast

As it can be observed from this figure, for the 84, 42% of the products reasonable 629 out of 748

products show an absolute percentage error that fluctuates between 0 and 30% whereas regarding

the rest of the products, 15, 4% of them present an absolute percentage error that ranges from 30 to

100% and only one product has an absolute percentage error higher than 100%. Hence, the results

are satisfactory and the relationship of the data is considered reasonable.

3. Actual Sales – Forecast comparison:

Actual Sales and Forecasted demand historical data is a comparison that Accenture has already

performed and validated. However, since the file that the author used resented wrong inventory

figures that do not match reality as the Min and Max comparison shown, at this point a comparison

between the forecast data that were provided to the author and the actual sales for the products will

be performed.

For this purpose, the author calculated in Excel the annual forecasted demand by the monthly

forecast values per product and the annual actual sales respectively. After obtaining the two annual

values per product, the author calculated the Absolute Error Percentage for every ISBN and in the

end the Mean Absolute Error was calculated. The mean absolute error percentage is calculated

according to the mathematical formula:

The following figure shows the distribution of the absolute error percentages for all the products.

629

115

1

Absolute Percentage Error 0-30%

Absolute Percentage Error 30-100%

Absolute Percentage Error >100%

Distribution of Absolute Percentage Errors for all products

Number of products

Page 83: Towards the Optimal Inventory Review Intervals

83

Figure 39: The distribution of absolute error percentages when comparing actual sales to forecast.

It can be seen that 505 out of 748products have an absolute error that ranges between 0 and 30%.

This accounts for a percentage of 67, 69% of the products. On the other hand, the rest 241 products

have an absolute error percentage that is higher than 30%. The results are quite satisfactory.

However, the MAE is estimated to be 49, 47%. However, this can be partially explained if we see that

for 72 products the absolute error is higher than 100% and some products show extreme absolute

error values such as 1538% or 1600% for instance. Hence, those products can cause an important

discrepancy in the calculation of the MAE. To sum up, according to the results, it can be stated that

the comparison has been performed sufficiently.

4. Initial Demand – Actual Sales comparison:

This relationship between the actual sales and the initial demand is the most crucial one and it

actually expresses the validation process because on the one hand, there are the historical data

represented by the annual sales for all products for the year 2014 that represent the real world and

on the other hand there is the Initial Demand KPI that is exported from S3N that represents the

simulation world. Hence, the Absolute percentage error is calculated for all products regarding their

annual actual sales for the year 2014 and their annual initial demand according to the S3n KPI export.

Figure 40 that follows shows the distribution of the absolute percentage errors according to their

ranges for all products.

252

165

88

59 46

12 17 10 16 9

72

0-10% 10-20% 20-30% 30-40% 40-50% 50-60% 60-70% 70-80% 80-90% 90-100% >100%

Distribution of Absolute Error Percentages

products

Page 84: Towards the Optimal Inventory Review Intervals

84

Figure 40: The distribution of absolute error percentages when comparing initial demand to actual sales

It can be seen by the figure that for most products the absolute percentage errors are not that high.

More specifically, it can be seen that for 181 products the percentage error is up to 10% whereas for

187 products it fluctuates between 10-20 percent. It can be noticed that high percentage errors

correspond to low amounts of products. For instance only 17 products have percentage errors that

range between 80-90% and further, only 6 products show an absolute percentage error between 90-

100%. The next figure shows the same calculation from a more cumulative perspective.

Figure 41: The cumulative distribution of absolute percentage errors when comparing initial demand to actual sales

From the last graph that is presented here, it can be stated that 65, 37% of products have a very

satisfactory percentage error on actual sales and simulated initial demand, whereas for 31, 95% of

the products the percentage error rises by ranging between 30 and 100%. The last fact is still

acceptable as from the previous graph it was observed that higher percentage errors correspond to

groups with lower product quantity. Hence, from the perspective of comparing the real world (actual

sales for 2014) to the simulation world (KPI “initial demand” exported from S3N) the model of the

author is valid.

Conclusions on validation on Sales and Forecasts data

181 187

119

70 61

37 29

18 17 6

20

0-10% 10-20% 20-30% 30-40% 40-50% 50-60% 60-70% 70-80% 80-90% 90-100% >100%

Absolute Error Percnetage ranges

Distribution of Absolute Error Percentages

products

65,37%

31,95%

2,68%

Absolute Percentage Error0-30%

Absolute Percentage Error30%-100%

Absolute Percentage Error>100%

Distribution of Absolute Percentage Errors

Percentage of Products

Page 85: Towards the Optimal Inventory Review Intervals

85

After checking the relationships between different data sets that were acquired from company X and

that were generated from S3N, it can be concluded all relationships checked were reasonable. The

validation regards the last relationship that was investigated regarding the comparison between the

initial demands of products, a KPI representing the simulation world and actual sales data

representing the real world. For the majority of products, nearly for 70% of the products, the results

of this comparison were regarded satisfactory and thus, the model is validated.

5.4 Simulation and Results This chapter continues with the approach defined in the previous chapter but focuses on the

execution of simulation experiments and the results. First, the experiments are designed after which

the experiments are executed using the simulation model. Finally the results of the experiments are

presented.

5.4.1 Design of experiments The simulation model has been built, tested, verified and validated and at this point the experiments

that need to be performed are designed in order to acquire a better insight on how the IRI can

influence specific KPIs that express the supply chain performance of company X for both the current

inventory policy and the new inventory policy to be implemented.

Selecting IRIs: 6 IRIs

While building the simulation model, for both inventory replenishment policies, an actor attribute

was also built called IRI for inputting the IRI in working days. Thus, adding the IRI as a parameter

allows making scenarios for different time intervals for both inventory replenishment policies. The

IRIs were chosen in a systemic manner, specifically following a geometric sequence that every time

doubles the previous value, taking into account that the simulation model estimates are in working

days: Starting from 1 day, being the first value, doubling to 2 days, then 5 days expressing one week

without the weekend, then double to 10 days that represents half month, then 21 days being

approximately one month and 42 days representing two months. These intervals are shown in the

following table:

Inventory review interval selection (in working days)

1 day

2 days

5 days (1 week)

10 days (1/2 month)

21 days (1 month)

42 days (2 months) Table 8: IRI selection

In S3N as it has already mentioned, an actor attribute was created for this purpose and for the sake

of this research this attribute works as a parameter and is changed six times according to the

aforementioned values.

Selecting inventory replenishment policies: 2 inventory replenishment policies

As described above, the IRI will change between six different time intervals expressed in working

days in order to understand how its variance can affect supply chain performance. Moreover, apart

from the IRI, another decision was made according to which the simulation for all six different IRIs

will be performed for both inventory replenishment policies discussed: the current one and new one.

In S3N, this can happen by changing in the yellow pages, a function of the simulation platform, within

the same scenario, the initiating actor role and the executing actor role of a transaction. More

Page 86: Towards the Optimal Inventory Review Intervals

86

specifically, regarding the transaction called “Manage inventories” that is both initiated and executed

by the actor “company X” the simulation runs will be first performed by setting the initiating actor

role to: “company X as inventory manager-current policy initiating” and the executing actor role to

“company X as inventory manager-current policy executing”. After the runs that regard the current

inventory replenishment policy, these settings are changed in order to perform the required

replications for the new inventory replenishment policy, the Min/Max policy. This is done by

changing in the same transaction the initiating actor role to: “company X as inventory manager-

Min/Max policy initiating” and the executing actor role to “company X as inventory manager-

Min/Max policy executing”.

Selecting products: 57 products

Selecting product from the product list of 748 products is a way to vary other meaningful variables.

Hence, by selecting a number of products with specific characteristics that vary, a new product list is

constructed being an input of the simulations instead of the initial one that contained all the 748

products.

Varying Lead times: supplier selection

First, a decision on lead times was made. The author decided to vary the lead time by choosing three

different suppliers: The UK litho-production supplier with a lead time of 22 working days that

correspond to one month. Further, the China supplier and the Malaysia supplier were selected and

both have a lead time of 66 working days that correspond to 3 months. According to the supplier

selection, 25 out of 57 products come from UK, 15 come from Malaysia and the rest come from

China.

Table 9: Varying lead time by selecting suppliers

Varying lead time by selecting suppliers

Varying slow and fast movers- ABC classification

As stated in this chapter, the products are selected according to various criteria. One of the criteria is

the identity of products, meaning being either slow or fast movers. This happens as it has already

been discussed according to the ABC classification. Hence, products from all 3 categories are

selected. 20 fast movers are selected belonging to category A, 23 products belonging to category B

and 14 slow movers belonging to category C.

Page 87: Towards the Optimal Inventory Review Intervals

87

Varying the demand variability- XYZ classification

Another criterion for selecting products with different characteristics is their variability of demand.

This distinction between products can be performed with the XYZ classification. The classification

allows selecting products with different demand variability level. Thus, 23 products of the selected

products have a low variability of demand belonging to category X, 18 products belong to category Y

and 16 show a high variability of demand belonging to category Z.

Clustering and checking for abnormalities

To select a reasonable amount of products the aforementioned criteria were taken into

consideration simultaneously. Hence, 57 products were selected in total. Each product originates

from one of the three aforementioned suppliers, and is classified according to the ABC and XYZ

classification.

Moreover, after the selection pf products, their demand patterns and forecasts were checked for

abnormalities, for instance zero values of demand that could cause miscalculations or not reasonable

simulation results. All products passed the check.

Overview of experiments: 12 Scenarios

By the analysis above, it is concluded that 12 scenarios will be tested as 12 experiments were

designed. Figure 42 explains how the conclusion was drawn on designing 12 experiments.

Figure 42 : Estimation of the number of scenarios.

Moreover, Table 10 presents the overview of all experiments to be tested.

Scenario 1 inventory review every 1 day Current policy ((s,S) policy)

Scenario 2 inventory review every 2 days Current policy ((s,S) policy)

Scenario 3 inventory review every 1 week Current policy ((s,S) policy)

Scenario 4 inventory review every 0,5 month Current policy ((s,S) policy)

Scenario 5 inventory review every 1 month Current policy ((s,S) policy)

Scenario 6 inventory review every 2 months Current policy ((s,S) policy)

Scenario 7 inventory review every 1 day New policy (Min/Max policy)

Scenario 8 inventory review every 2 days New policy (Min/Max policy)

Scenario 9 inventory review every 1 week New policy (Min/Max policy)

Scenario 10 inventory review every 0,5 month New policy (Min/Max policy)

Page 88: Towards the Optimal Inventory Review Intervals

88

Scenario 11 inventory review every 1 month New policy (Min/Max policy)

Scenario12 inventory review every 2 months New policy (Min/Max policy)

Table 10: Overview of the Scenarios to be tested

5.4.2 Execution of experiments In this part the execution of the 12 experiments takes place. First, the model is set according t the

requirements of the tests and second, the simulations are performed.

Model Set up In order to execute the experimental design on the verified and validated model, this section first is

essential to discuss the simulation set up. This refers to defining the warm up time, the length of the

simulation and the number of replications for each scenario that has been designed.

As it is already mentioned, there is a warm up time for every scenario that equals to one year. For all

scenarios, the simulation period is from 1 January 2013 to 1 January 2015. Hence, with a warm up

time of 12 months, the results depict what occurs during January 2014 and January 2015. Thus, the

length of the simulation is two years: from 2013 to 2015.

Regarding how many replications are needed in order to gain an outcome that is reliable enough, too

few simulation runs can cause unreliable results but too many runs cause waste of time. Due to the

initial tests of the model where the model was tested to estimate a sufficient number of replications,

it was decided that each scenario will run for 50 replications. Moreover, in order to be certain that

the number of replications is correct, one scenario was performed first and used to decide on how

many runs are needed in order to get a convergent average and at the same time how long it takes.

The scenario that was run for this reason was Scenario 5 that regards the current policy with an IRI of

one month for a random product. For this purpose, a random product was selected from the product

list for this random scenario. Then, this scenario was executed for 50 replications. The KPI of finished

inventories was exported for all the 50 replications. Regarding the plot above, for every replication

out of the 50 replications in total, the average inventory value for the random product is calculated.

Figure 43: Average inventory values for the random product

600700800900

10001100120013001400150016001700180019002000

0 10 20 30 40 50 60

Ave

rage

In

ven

tory

val

ue

Number of replications from 1 to 50

Annual inventory value for 50 replications for one random product for one random scenario

Page 89: Towards the Optimal Inventory Review Intervals

89

In figure 43 the x axis is the number of replications counting from the 1st replication to the 50th

replication performed. The y axis shows the annual inventory value of the product. It can be

observed that the average inventory value of this product varies very little among the 50 replications

and it the average of 50 replications seems sufficient. However, this is not a scientific proof of

defining the sufficient number of replications.

Moreover, It has been determined that 50 replications of the simulation model results in an

acceptable 95% confidence interval: After performing a set of replications, the average value for a

KPI can be determined as well as the accompanying standard deviation within a 95% confidence

interval (Verbraeck, 2010). The purpose of a confidence interval is to estimate an unknown

population parameter with an indication of how accurate the estimate is and of how confident we

are the result is correct. The confidence interval μ around the average outcome is determined using

the following equation:

.

In this equation, x is the expectation of the outcome and h is the half width value of the confidence

interval. The lower the value for h, the higher the confidence of the outcome is. When 𝑥𝑗 is the result

of the j-th replication, the values for x and 𝑠2 for x and x are given by the following equations:

The half width h provides the range in which the outcomes vary from . H can be determined using

the following equations:

In order to determine the confidence interval for the created simulation model, μ is being

determined using the aforementioned formulas along with excel formulas. Initially, 50 replications

are being performed. With high certainty (95%) the result range can be determined in the following

table.

Statistics

Mean 1849,377

Standard deviation 14,642

Sample size 50

Alpha 0,05

Confidence interval 4,058

Upper limit 1853,435

Lower limit 1845,319

There is a 95% chance that the true mean is between 1845,319 and 1853,435

Tabel 11: Calculating the confidence interval

Based on the results of the table, it can be concluded that there is a 95% chance that the true mean

is between 1845,319 and 1853,435. Hence, based on the aforementioned, the amount of 50

replications is sufficient to reach an acceptable range of the 95% confidence interval.

Running the scenarios

Page 90: Towards the Optimal Inventory Review Intervals

90

After the model was set, the simulations took place. Each scenario ran for approximately 45 to 55

minutes. The product list, as mentioned, contains 57 products, thus, the simulation for each run can

be considered as quite fast. After having executed all the experiments, the author proceeded in

exporting the relevant KPIs for obtaining the results and for being able to analyze them afterwards.

The KPIs that are relevant for the sake of this analysis are two: The first one is the final inventory

which in S3N can be found in the “Inventory turnover” KPI category under the name “Inventory final

product”. This KPI depicts for all 57 ISBNs their final inventory status during one year. The second KPI

that is relevant for the analysis belongs to the “Service Level” KPI category and is named “%

Delivered on time versus Requested” and expresses the rate of actual sales that were delivered on

time to the initial demand that was requested from the customer.

5.4.3 Results

5.4.3.1 A macro approach to the simulation results

Presentation of results

Because each experiment consists of 50 replications, much data is generated from one experiment.

The results that are reviewed are the two KPIs that are exported from the simulation software and

their further analysis in Excel. For this purpose, as it has been already mentioned the two KPIs called

“Final inventories” from the Inventory Turnover KPI group and the KPI called “% Delivered on time

versus Requested” from the Service Level KPI group are exported in Excel for all 12 scenarios that

were tested. In particular, due to the 50 replications for every scenario, the result of the KPI values

exported for every ISBN are the average values of those 50 replications in both the case of the

inventory KPI and the service level KPI.

Using Excel, the data from every experiment is accumulated in a matrix with all the relevant details

that the author wants to focus on. Hence, one matrix is generated for each KPI in Excel hose columns

are: ISBN, lead time, supplier, ABC categorization, XYZ categorization, and then 12 columns (for the

12 scenarios) containing the KPI exported data from the model. The cumulative results can be seen in

appendix A. Afterwards, a thorough analysis is conducted by first considering the results that regard

the KPIs for all 12 scenarios and second, by making meaningful plots regarding the KPIs, taking into

account the factors that are examined in this research, such as the different lead times and the ABC-

XYZ classifications.

Figure 44 presents the way that all results are outputted from the simulation model:

Page 91: Towards the Optimal Inventory Review Intervals

91

Figure 44: Presentation of the way that the results are outputted from the model in Excel in order to be analyzed.

It is evident that the various scenarios can be compared and analyzed in a more efficient way in Excel

if they are organized in a way that product clusters can be easily formulated in order to observe the

simulation behavior in every scenario. The reason that the author wanted to approach the results of

the simulation this way is that in essence every scenario includes the same amount of products that

have different characteristics such as the lead time, their supplier and the different category

according to the ABC and the XYZ categorization. Hence, formulating the outputted results in Excel

the way that has been described, facilitates the analysis of the results and the conclusions drawing

process.

Generic averages from the simulation runs for the KPI “% Delivered on time versus

Requested” :

In order to visualize the model behavior, usable plots have been created to show the cumulative

results for both KPIs. Hence, for the KPI “% Delivered on time versus Requested” the following plot

in Figure 45 shows the average values of the service level KPI “% Delivered on time versus

Requested” for every one of the 12 scenarios.

Page 92: Towards the Optimal Inventory Review Intervals

92

Figure 45: The average values of the service level KPI “% Delivered on time versus Requested” for all scenarios

This plot is only for demonstrating all the average values but it is not practical to use it in order to

distill valuable observations. The reason for that is that scenarios 1 to 6 regard the current

replenishment policy (the (s,S) policy), whereas scenarios 7 to 12 regard the new policy (the

Min/Max policy). Thus, the following plot has been created and is presented below:

Figure 46: Average values of the service level KPI “% Delivered on time versus Requested” for both policies.

The plot summarizes the simulation behavior of the KPI “% Delivered on time versus Requested”

regarding both policies. Moreover, the numbers in the x-axis represents the number of scenarios as

they have been specified in chapter 4. Hence number 1 represents scenario 1 for the current policy

and scenario 7 for the Min/Max policy.

It can be observed that in the case of the current inventory policy, the KPI is not fluctuating much as

it presents small increases and decreases for the current policy. On the other hand, regarding the

Min/Max policy, the service level expressed in terms of the KPI “% Delivered on time versus

Requested” seems to diminish steadily until the 11th scenario and then at the 12th scenario the

decrease is more intense.

Page 93: Towards the Optimal Inventory Review Intervals

93

Generic averages from the simulation runs for the KPI “Inventory final product” :

The respective plot for the inventory KPI “Inventory final product” is presented in the following

figure. In this figure one can see the average values of the finished inventories for the scenarios 1 to

6 that regards the current policy and the same KPI for the scenarios 7 to 12 that refer to the new

Min/Max policy:

Figure 47: Average values of the service level KPI “Inventory final product” for both policies

It can be observed by looking at the presented figure that the simulation behavior regarding the

current policy seems somehow less consistent than the behavior regarding the Min/Max policy.

Moreover the decrease is more observable at the Min/Max policy than at the current policy. It

should also been noted that the scales of the x –axis are very different, since the current policy starts

with very high inventories compared to the Min/Max policy. It is observed that the two different

policies affect the results regarding the inventories: at the current policy the inventories are very

high and at the Min/Max policy the inventories are substantially lower in scale.

Conclusions on generic averages from the simulation run for the two KPIs:

It seems necessary at this point to highlight that there are some first conclusions regarding the two

KPIs that have been presented above. First, the replenishment policy seems to define the simulation

behavior substantially. Thus, the first six scenarios (Scenarios 1 to 6) that refer to the current policy

need to be separately reviewed and the same applies to the rest six scenarios (7 to 12) that refer to

the Min/Max policy. Secondly, it is observed that the unit of measurement for comparing the

scenarios in the case of the service level KPI facilitates this process as the KPI is expressed as a %

percentage. Hence, in all cases the KPI results are comparable. However, in the case of the second

KPI that regards finished inventory, the KPI is expressed in units of product quantities. Thus, the

comparison is not that simple. Moreover, it is noticed while reviewing the simulation results that the

inventory values differ in scale substantially between the scenarios that regard the current policy and

the scenarios that refer to the Min/Max policy. This means that the inputs regarding the inventories

for the two different policy cases are very different. It can be seen that there are products with a

stock of more than 1880 units on average in the current policy. However, the Min/Max policy works

Page 94: Towards the Optimal Inventory Review Intervals

94

with fewer inventories in general as it is observed that for the same products the highest value of

kept stock does not exceed the 550 units on average.

Thus, it can be stated that Scenarios 1 to 6 differ substantially to Scenarios 7 to 12 due to inventory

scales that were specified according to the requirements of each replenishment policy that has been

selected. Therefore, it seems interesting to seek different relations between these scenarios by

obtaining a more micro-approach on the analysis of the results.

5.4.3.2 A micro approach to the simulation results

To make more valuable observations on the simulation outputs, the results have been visualized with

acquiring different perspectives. The way that was selected to present the results presented in

Appendix A facilitates the process of obtaining a micro approach on them. The reason is that this

way, valuable conclusions can be drawn when isolating specific clusters of results depending on

which influencing factors the author chooses to investigate each time. Thus, for both KPIs, first for

the “% Delivered on time versus Requested” and then for the “Inventory final product” KPI, the

results are shown primarily from the ABC and subsequently from the XYZ perspective. Then, a

presentation of the results regarding the different Lead times follows.

“% Delivered on time versus Requested” KPI results regarding the ABC classification

The KPI will be visualized by showing the first six scenarios that regard the current policy separately

than the scenarios 7 to 12 that refer to the Min/Max policy. Following this pattern, the graphs in

Figure 48 and 49, present the ABC categorization in the case of the two policies.

Figure 48: Average values of the service level KPI “% Delivered on time versus Requested” for the current policy according to the ABC categorization.

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

1 2 3 4 5 6

“% Delivered on time versus Requested” KPI averages for ABC product groups for current policy

A products

B products

C products

Page 95: Towards the Optimal Inventory Review Intervals

95

Figure 49: Average values of the service level KPI “% Delivered on time versus Requested” for the Min/Max policy according to the ABC categorization

The figures that are presented above show the average values of the service level KPI “% Delivered

on time versus Requested” first for the 6 first scenarios that refer to the current policy and then for

the rest 6 scenarios (scenarios 7 to 12) that regard the Min/Max policy, according to the ABC

categorization. The main conclusion here is that in the current policy the IRIs do not seem to

influence the KPI substantially as they show a steady evolution when observing the different

scenarios of different IRIs. However, in the case of the Min/Max policy, it seems that the IRIs

influence in a more evident way the service level KPI. This is evident for all product clusters.

Moreover, there is not any highly observable difference between the rate that the service levels

decrease as we switch to higher IRIs between the A, B, and C product clusters as the rate of decrease

seems steady.

In Appendix B a more extensive analysis is implemented to compare the “% Delivered on time versus

Requested” KPI results regarding the ABC classification according to the different product clusters.

More specifically, to shed more light on the different product clusters, the author decided to

visualize the separate product clusters in order to be able to compare the A products in the current

policy to the A products in the Min/Max policy, the B products in the current policy to the B products

in the Min/Max policy, and the C products in the current policy to the C products in the Min/Max

policy.

“% Delivered on time versus Requested” KPI results regarding the XYZ classification

Again as well the KPI will be visualized by showing the first six scenarios that regard the current policy

separately than the scenarios 7 to 12 that refer to the Min/Max policy. Following this pattern, the

graphs below present the XYZ categorization in the case of the two policies in figires 55 and 56.

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

1 2 3 4 5 6

“% Delivered on time versus Requested” KPI averages for ABC product groups for Min/Max policy

A products

B products

C products

Page 96: Towards the Optimal Inventory Review Intervals

96

Figure 50: Average values of the service level KPI “% Delivered on time versus Requested” for the current policy according to the XYZ categorization.

Figure 51: Average values of the service level KPI “% Delivered on time versus Requested” for the Min/Max policy according to the XYZ categorization.

The figures that are presented above show the average values of the service level KPI “% Delivered

on time versus Requested” first for the 6 first scenarios that refer to the current policy and then for

the rest 6 scenarios (scenarios 7 to 12) that regard the Min/Max policy, according to the XYZ

categorization. The main conclusion here is that in the current policy the IRIs do not seem to

influence the KPI substantially as they show a steady evolution when observing the different

scenarios of different IRIs. However, in the case of the Min/Max policy, it seems that the IRIs

influence in a more evident way the service level KPI. This is evident for all product clusters.

Furthermore, there is not any highly observable difference between the rate that the service levels

decrease as we switch to higher IRIs between the X, Y, and Z product clusters as the rate of decrease

seems steady.

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1 2 3 4 5 6

“% Delivered on time versus Requested” KPI averages for XYZ product groups for current policy

X products

Y products

Z products

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1 2 3 4 5 6

“% Delivered on time versus Requested” KPI averages for XYZ product groups for Min/Max policy

X products

Y products

Z products

Page 97: Towards the Optimal Inventory Review Intervals

97

In Appendix C, a more extensive analysis is implemented to compare the “% Delivered on time versus

Requested” KPI results regarding the XYZ classification according to the different product clusters.

Moreover, separate product clusters are visualized in plots separately in order to be comparable: the

X products in the current policy are compared to the X products in the Min/Max policy, the Y

products in the current policy to the Y products in the Min/Max policy, and the Z products in the

current policy to the Z products in the Min/Max policy.

“% Delivered on time versus Requested” KPI results regarding the different Lead times

At this point the aim it to observe and compare the “% Delivered on time versus Requested” KPI

results regarding the two different lead times that have been examined for both the replenishment

policies. As it has been stated, the author decided to vary the lead time by choosing three different

suppliers: The UK litho-production supplier with a lead time of 22 working days that correspond to

one month. Further, the China supplier and the Malaysia supplier were selected and both have a lead

time of 66 working days that correspond to 3 months.

Figures 52 and 53 visualize the behavior of the service level KPI “% Delivered on time versus

Requested” at the current policy first and then at the Min/Max policy for the two different lead time

product groups.

Figure 52: Average values of the service level KPI “% Delivered on time versus Requested” for the two lead time groups for

the current policy.

Page 98: Towards the Optimal Inventory Review Intervals

98

Figure 53: Average values of the service level KPI “% Delivered on time versus Requested” for the two lead time groups for the Min/Max policy.

In the figures presented above, the x-axis represents the service level KPI. The y-axis refers again to

the different review interval selections as they have been specified in the scenarios that have run.

Since the first figure is for the current policy, the y-coordinates correspond to the first six scenarios.

The y-coordinates of the second figure that is for the Min/Max policy correspond to the scenarios 7

to 12.

The main conclusion on this comparison is that varying the lead times does not seem to affect in an

abnormal way the service level within one policy. However, if we compare how the same lead time

groups behave for the two different policies, it can be observed that for the current policy, the

decrease of the service level is not that significant. It can be seen that the IRI does not affect much

the service level in the current policy for both lead time product groups. On the other hand, while

viewing at the Min/Max case, the decrease of the service level in both lead time product groups is

more distinct. Furthermore, the rate of decrease seems to increase while we move to higher

inventory intervals.

Moreover, it is evident that for the case of LT= 22 days, the KPI values fall down more quickly than in

the case of LT= 66 days. This can be explained by the fact that when the lead tine is 22 days, you are

going to order more often. When the lead time is 66 days, however, you have bigger orders hence,

one can order less often, so he will have more inventories on average. This will make the chance of

running out of stock smaller.

In Appendix D a more elaborate analysis is implemented to compare the “Inventory final product”

KPI results regarding the two different lead times that have been examined for both the

replenishment policies.

The same procedure that was presented so far for the presenting and explaining the behavior of the

service level KPI “% Delivered on time versus Requested” is followed to explore the behavior of the

inventory KPI. The procedure starts with presenting first the “Inventory final product” results from

the ABC categorization perspective and subsequently from the XYZ categorization perspective. Then,

a presentation of the results regarding the different Lead times follows.

0,6

0,65

0,7

0,75

0,8

0,85

0,9

1 2 3 4 5 6

“% Delivered on time versus Requested” KPI averages for two lead time groups for Min/Max policy

LT 66 days

LT 22 days

Page 99: Towards the Optimal Inventory Review Intervals

99

“Inventory final product” KPI results regarding the ABC classification:

The KPI will be visualized by showing the first six scenarios that regard the current policy separately

than the scenarios 7 to 12 that refer to the Min/Max policy. Following this pattern, the graphs below

in Figures 54 and 55 present the ABC categorization in the case of the two policies.

Figure 54: “Inventory final product” KPI averages for ABC product groups for the current policy

Figure 55: “Inventory final product” KPI averages for ABC product groups for the Min/Max policy

Moreover, it is noticed that the inventory values differ in scale substantially between the scenarios

that regard the current policy and the scenarios that refer to the Min/Max policy. As it has been

previously stated, the Min/Max policy works with fewer inventories compared to the current policy.

This is obvious if the x-axis is checked. The maximum averages of inventory units in the current policy

Page 100: Towards the Optimal Inventory Review Intervals

100

even exceed in the 4th scenario for the A products the value of 4000 for example, while for the 10th

scenario in the second figure (y-coordinate=4) for the same A products at the Min/Max policy, this

average corresponds to less than 1300 units of products.

Further, a lower rate of decrease is observed at the current policy compared to the rate of the

inventory decrease at the Min/Max policy. In particular, at the Min/Max policy it is distinct enough to

notice that the inventory KPI decreases in a faster rate as we switch to higher IRIs.

In Appendix E a more extensive analysis is implemented to compare the “Inventory final product” KPI

results regarding the ABC classification according to the different product clusters. More specifically,

to shed more light on the different product clusters, the author decided to visualize the separate

product clusters in order to be able to compare the A products in the current policy to the A products

in the Min/Max policy, the B products in the current policy to the B products in the Min/Max policy,

and the C products in the current policy to the C products in the Min/Max policy.

“Inventory final product” KPI results regarding the XYZ classification:

Again as well the KPI will be visualized by showing the first six scenarios that regard the current policy

separately than the scenarios 7 to 12 that refer to the Min/Max policy. Following this pattern, the

figures 56 and 57, present the XYZ categorization in the case of the two policies.

Figure 56: “Inventory final product” KPI averages for XYZ group for the current policy

Page 101: Towards the Optimal Inventory Review Intervals

101

Figure 57: “Inventory final product” KPI averages for XYZ product groups for Min/Max policy

The main conclusion regarding this comparison is that at the current policy the IRIs do not seem to

influence the KPI substantially as they show a steady evolution when observing the different

scenarios of different IRIs. However, in the case of the Min/Max policy, it seems that the IRIs

influence in a more evident way the service level KPI. This is evident for all product clusters. Further,

there is not a highly observable difference between the rate that the average inventory decreases as

we switch to higher IRIs between the X, Y, and Z product clusters as the rate of decrease seems

steady.

In Appendix C, a more extensive analysis is implemented to compare the “Inventory final product”

KPI results regarding the XYZ classification according to the different product clusters. Moreover,

separate product clusters are visualized in plots separately in order to be comparable: the X products

in the current policy are compared to the X products in the Min/Max policy, the Y products in the

current policy to the Y products in the Min/Max policy, and the Z products in the current policy to the

Z products in the Min/Max policy.

“Inventory final product” KPI results regarding the different Lead times:

At this point the aim it to observe and compare the “Inventory final product” KPI results regarding

the two different lead times that have been examined for both the replenishment policies. Figures 58

and 62 visualize the behavior of the service level KPI “Inventory final product” at the current policy

first and then at the Min/Max policy for the two different lead time product groups.

Page 102: Towards the Optimal Inventory Review Intervals

102

Figure 58: "Inventory final product" KPI averages for two lead time groups for current policy

Figure 59: "Inventory final product" KPI averages for two lead time groups for Min/Max policy

In the figures presented above, the x-axis represents the inventory KPI. The y-axis refers again to the

different review interval selections as they have been specified in the scenarios that have run. Since

the first figure is for the current policy, the y-coordinates correspond to the first six scenarios. The y-

coordinates of the second figure that is for the Min/Max policy correspond to the scenarios 7 to 12.

The main conclusion on this comparison is that varying the lead times does not seem to affect in an

abnormal way the inventories within one policy. However, if we compare how the same lead time

groups behave for the two different policies, it can be observed that for the current policy, the

decrease of the inventory is not that significant. It can be seen that the IRI does not affect much the

inventory in the current policy for both lead time product groups. On the other hand, while viewing

the Min/Max case, the decrease of the inventory in both lead time product groups is more distinct.

Furthermore, the rate of decrease seems to increase while we move to higher IRIs.

Page 103: Towards the Optimal Inventory Review Intervals

103

5.5 Concluding remarks In this chapter, the simulation model was conceptualized as a black box. Then it was described in

detail by presenting all the system’s components the conceptual design of the simulation thus, was

specified and the system was analyzed: the black box was opened using the DEMO methodology that

unraveled the relevant business processes that occur and analyzed how those are formed via the

input and through the controls in order to lead to the supply chain measurement outputs. Next, the

simulation model was verified and validated in order to be used. Afterwards, the experiments were

designed and then executed. In the end, the results of the simulation were presented. A macro and a

micro approach then, were adapted for presenting the results of the simulation. An overview of the

results that are presented in this chapter is depicted in the next table.

Macro approach

Generic averages from the simulation runs for

the KPI “% Delivered on time versus Requested”

Remark: The replenishment policy seems to

define the simulation behavior substantially.

Generic averages from the simulation runs for

the KPI “Inventory final product”

Remark: Scenarios 1 to 6 differ substantially to

Scenarios 7 to 12 not only because of the

different replenishment policy but also due to

their inventory scales.

Micro approach

“% Delivered on time versus Requested” KPI

results regarding the ABC classification.

Comparison in this chapter. Further analysis in

Appendix B.

“% Delivered on time versus Requested” KPI

results regarding the XYZ classification.

Comparison in this chapter. Further analysis in

Appendix C.

“% Delivered on time versus Requested” KPI

results regarding the different Lead times.

Comparison in this chapter. Further analysis in

Appendix D.

“Inventory final product” KPI results regarding

the ABC classification.

Comparison in this chapter. Further analysis in

Appendix E.

“Inventory final product” KPI results regarding

the XYZ classification.

Comparison in this chapter. Further analysis in

Appendix F.

“Inventory final product” KPI results regarding

the different Lead times.

Comparison in this chapter. Further analysis in

Appendix G.

Table 11: Overview of the results presented

Page 104: Towards the Optimal Inventory Review Intervals

104

Page 105: Towards the Optimal Inventory Review Intervals

105

6 CONCLUSIONS In this chapter, we return to the main research objective: To investigate the impact that the IRI has,

on the supply chain performance of Company X.” A process is conducted that entails the evaluation

through answering the research question as formulated in the 1st chapter and the results discussion.

6.1 Answering the research questions The main research question presented in the first chapter is answered here in two parts. First, by

providing a structured overview of the answers to the sub research questions. Then, based on the

knowledge embedded in these sub research questions, the main research question is answered in

the second part. The main research question was defined as:

“What are the effects of review intervals on the supply chain performance of company X?”

6.1.1 Review of the sub research questions The set of the sub research questions that together tackle the main research question are answered

in this section.

Sub question 1: What theories are relevant regarding the IRI and why it is important to focus

on the IRI?

This preliminary question is answered in the Supply Chain analysis chapter, chapter 3, which includes

a thorough analysis combined with a literature review. The literature review is performed with a top

down approach. Hence, starting from general notions such as inventory management the author

narrows down the literature in a systemic way until the IRI notion is reached. The author presents

the IRI as a key variable for this research along with the theories that are tied to the term.

Hence, as depicted in chapter 3 where the supply chain analysis takes place, the reasons why one

should focus on inventory management are explained along with the reasons of keeping inventories.

Moreover, selecting an inventory strategy is tackled as an important issue when it comes to deciding

for example the appropriate amount of stock one should keep. Further, such decisions are associated

with the available operations models that exist, such as the Make-to –stock and the Deliver –to-

order. Different perspectives on stock level are examined. Inventory theory entails as well inventory

control. Relevant notions hence that were reviewed are inventory aggregation level, safety stock,

lead time, demand and its forecast, and backlogs. Additionally, one of the processes of inventory

management and control that was tackled is the classification of inventories. Hence, ABC and XYZ

categorization were reviewed.

Moreover, inventory control theory was distinguished in both certain and uncertain conditions and

explained via respective inventory replenishment policies. Certain conditions are expressed by

models such as the EOQ model. Uncertain conditions mean stochastic behavior and relevant models

are the (s,Q), the (s,S) or the (R,s) policies. Subsequently, the importance of the IRI is tackled

explained and the knowledge gap that triggered this research is identified and presented as the

objective of this research. The reason why it is important, hence, to focus on the IRI, is that it is

identified that there is not a clear picture regarding the effects of the IRI on supply chain

performance.

Page 106: Towards the Optimal Inventory Review Intervals

106

Sub question 2: Which inventory replenishment policies and IRI ranges are relevant to

company X’s case?

Sub question 2 is answered in Chapters 4 and 5. After defining the theory that is relevant for the IRIs

and their related notions that are crucial for this research, in chapter 4 the general notions are

applied to an embedded case study in order to define specific inventory replenishment policies and

IRI ranges for this case study. Company X is selected for the case study performance with three sub-

units of analysis: products, the inventory replenishment policy and the IRI. Subsequently, these are

specified in terms of the situation in company X: First, the supply chain of the company is presented,

and its current inventory replenishment policy ((s,S) policy) and the new inventory replenishment

policy to be implemented (Min/Max policy) are discussed. A detailed description of the logic for the

current replenishment policy is performed. In addition, a detailed description of the logic behind the

new replenishment policy is also presented. The additional details are specified as inputs in the

specification part of the simulation construction process. This can be seen in chapter 5.

More specifically, in chapter 4 the inventory replenishment policies that regard the case study

context are presented and analyzed. Further, the IRI ranges are specified in chapter 5, while

designing the experiments. Hence, sub question 2 is half answered in chapter 4 while conducting the

case study and half answered in chapter 5 while making the simulation model. The answer given in

the aforementioned chapters is that the currently applied replenishment policy, that's is the (s,S)

policy and the new policy to be implemented in company X, meaning the Min/Max policy, are the

two relevant policies to focus on in this thesis. Additionally, the ranges of the IRIs that are

researched are the following, calculated in working days as presented in chapter 5:

Inventory review interval selection (in working days)

1 day

2 days

5 days (1 week)

10 days (1/2 month)

21 days (1 month)

42 days (2 months) Table 12 : Inventory review interval selection

Sub question 3: How can supply chain performance be defined and measured, in the case of

company X?

In order to answer sub-question 3, the supply chain performance measurement was defined and

researched during the literature review primarily to gain a broader knowledge on how supply chain

measurement is conducted and what it represents. For the specific case of company X, the

specification was reached after conducting the case study in chapter 4. In chapter 5 that presents the

simulation model construction part, the performance measures are defined as outputs of the

system/ model as seen in the system analysis and are identified and are represented as KPIs

outputted from the simulation model. These KPIs are:

1. % delivered on time vs requested: This KPI expressed the service level provided by company

X as it is essentially the calculation “Sales/ Initial demand”, meaning how much was sold

divided by how much was initially requested for ordering.

Page 107: Towards the Optimal Inventory Review Intervals

107

2. Inventory final product: This KPI measures the quantity of the products that are in stock by

outputting snapshots of the inventory as time goes by and is expressed in terms of product

quantities per ISBN.

Sub question 4: How can one test the effect of IRIs on supply chain performance of company

X?

This sub-question entails answering the former two sub questions because in essence it expresses

the question of how can one test the effect of IRIs on supply chain performance but under the scope

of the case study and the specifications of the conceptualized model in chapter 4. The sub- questions

2 and 3 shed light into specifying the IRIs and the KPIs that are relevant in the case. Sub-question 4 is

asking how these two notions are linked and how can one find the impact of the IRI on those KPIs in

the case of company X. Hence, this question is answered in chapter 5 where first the conceptual

model of the simulation is been developed. The conceptual model is a system that is perceived as a

black box.

Figure 60: The conceptual design of the simulation presented as a system diagram.

The system’s components are analyzed using IDEF0 and the specifications of the system are

presented. The inputs, outputs controls and mechanisms are defined and specified. Moreover, the

system-black box is opened up using the DEMO methodology to present the processes that take

place expressed in terms of actors, actions and transactions. The conceptual model is then

implemented in a simulation model to actually experiment on how different IRIs affect the selected

KPIs under the two inventory replenishment policies that are tackled in this thesis.

Sub question 5: How do supply chain performance metrics behave under different IRIs for

company X?

As described, the simulation model is being developed from the implementation of the conceptual

model that has been defined in chapter 5. More specifically, the simulation model is constructed in

the S3N surface according to the specifications that are thoroughly described in chapter 5 and

subsequently the results are outputted in Excel. The results are data that regard the values of the “%

delivered on time vs requested” and the “inventory final product” KPIs. In chapter 5, after the

presentations of designing and executing the experiments, the results of the simulation are

Page 108: Towards the Optimal Inventory Review Intervals

108

presented and analyzed, after visualizing most of the data with plots drawn in Excel. Moreover, the

outputted KPI values are presented in Appendix A. This process can be seen in the following figure:

Figure 61: Presentation of the way that the results are outputted from the model in Excel in order to be analyzed.

Furthermore, to make more valuable observations on the simulation outputs, the results have been

visualized with acquiring different perspectives. The way that was selected to present the results

presented in Appendix A facilitates the process of obtaining a micro approach on them. The reason is

that this way, valuable conclusions can be drawn when isolating specific clusters of results depending

on which influencing factors the author chooses to investigate each time. Thus, for both KPIs, first for

the “% Delivered on time versus Requested” and then for the “Inventory final product” KPI, the

results are shown primarily from the ABC and subsequently from the XYZ perspective. Then, a

presentation of the results regarding the different Lead times is conducted. The answer to this sub

question therefore, stems from the aforementioned results.

Specifically, the results are presented in the last subchapter of chapter 5 and are consistent. For both

the service level KPI and the inventory KPI, the KPI values decline as we move from smaller IRIs to

bigger: Both the finished inventories and the service levels decrease as we move from smaller to

bigger IRIs. Moreover, there is a significant difference between the first six scenarios that regard the

current policy and the scenarios 7 to 12 that regard the Min/Max policy. The scenarios that follow

Page 109: Towards the Optimal Inventory Review Intervals

109

the current policy show a smaller decrease and in some plots in the appendixes that show the

extensive results of the simulation, present higher variability than those of the Min/Max policy.

If we take into account the differentiation of the scenarios according to the two different

replenishment policies, it can be noticed that for the case of the current policy, there is some normal

variability that is also visualized in the plots presented in the appendixes with the simulation results

for the current policy. This constitutes a conclusion: There is a coincidental occurrence at the current

policy that is visualized as variability, for which we do not know the reason. For the Min/Max policy it

can be seen that the decline of the KPI values is more consistent. However, on the current policy an

unexpected behavior is noticed, which clearly is dependent on the IRIs, because the KPI values for

the -5 days- IRIs are always very low and then the subsequent KPI values for the 10 days IRI is always

quite high. Apparently, it can be concluded that there is some serendipity there, as it is clearly caused

by the IRIs but it appears to have a chaotic behavior if one looks at the performance of the KPIs

suddenly increasing on the 5 days IRI and then decreasing again on the 10 days IRI. It is evident that

the number of days that constitute the IRI plays a role and influences the KPI behavior, but without

being able to further explain it. However, this pattern of behavior is stable and consistent for all

scenarios and this can be seen in the appendixes presenting the simulation results.

Further, the reason why the decrease in the inventory KPI values is more intense in the Min/Max

policy is because the current policy works with larger inventory scales compared to the ones of the

Min/Max policy. Hence, this difference is not because of the different replenishment policy in

general, but because the inventories are of significantly higher scale in the current policy when

compared to the inventory scale in the Min/Max policy.

6.1.2 Review of the main research question This section evaluates the main research question:

“What are the effects of review intervals on the supply chain performance of company X?”

Overall, both KPI values decline as we move from smaller to bigger IRIS: Both the finished inventories

and the service levels decrease as we move from smaller to bigger IRIs. However, if one looks at the

scale, the decrease is not that strong. However, looking at both the two different KPI behaviors, they

are consistent in their decrease.

Another point is that the KPI values show a more consistent decrease when it comes to the Min/Max

policy for both KPIs. Regardless the serendipity that is noticed in the current policy’s behavior, still, it

is noticed that both the service level KPI and the inventory KPI show a higher decrease in Min/Max

policy. This is mainly explained by the fact that the two replenishment policies start with different

inventory scales. The current policy as it has been stated before has a lot of inventory, whereas the

Min/Max policy works with fewer inventories. Thus, this behavior can be easily explained: If one has

too much of inventory (current policy), he can just sit down for 42 days (the highest IRI chosen by the

author) and do nothing and then, after 42 days it may happen that there is still inventory left.

However, if the inventory is less, it becomes more important to review it because it is going to run

out soon if one does not control it. Hence, the differentiation according to the replenishment policy

among the 12 scenarios is very crucial because of these significant differences in inventories. A

conclusion here is that if you have a lot of inventory in the policy, then it doesn’t matter what IRIs

you keep, because you have a lot of inventory already. But as soon as you start managing better your

inventory (for example having less inventory in the policy as it happens in the Min/Max policy), the

IRIs become more efficient: The point is that the inventory scales should be not too much as in the

current policy where one cannot see what will happen with so much inventory. On the contrary,

instead of having all this stock standing in the warehouse, the inventory should be just enough not to

Page 110: Towards the Optimal Inventory Review Intervals

110

run out and to be delivered on time to the clients. If one achieves that, then it can be stated that

shorter IRIs do influence the supply chain performance: shorter IRIs mean higher KPI inventory values

and also higher service level values. This is evident because one has less inventory, thus he will not

“close his eyes” for 42 days. This interval is too big to not take any action. He will at least guess ahead

how many books he has to sell and thus, order. In such a case, one could see the influence of the IRIs

in a way that the performance of the KPIs is decreasing quickly as we move from lower to higher IRIs.

To conclude, the question “What are the effects of review intervals on the supply chain performance

of company X?” has been answered and the conclusion can be formulated as:

It depends on the inventories of the replenishment policies:

if the inventory replenishment policy works with a lot of inventory, then it doesn’t really

matter what the IRIs are because there is a lot of inventory already and one can “close his

eyes” for even one month and still the KPIs do not drop substantially. It is shown that when

you have a lot of inventory, it starts to matter for IRIs higher than one month.

if the inventory replenishment policy works with less inventory, as it is shown by the

Min/Max policy, as soon as the inventory scales are managed better and one does not hold

in stock too much inventory, the IRIs become more efficient: shorter IRIs then, do increase

the supply chain performance of company X.

6.2 Generalization of results In the previous chapter the conclusions were drawn regarding the impact of the IRIs on the supply

chain performance of company X. Here, the generalization of the results for different IRIs, other KPIs

and other supply chains outside company X is discussed.

For other IRIs

From the perspective of the IRIs for company X, this variable was operationalized according to the

scope of company X specifications. Thus, six different IRIs were reviewed for this purpose. Hence, this

simulation model could be used to also test the impact of the IRIs on supply chain performance for

the same company or for another company but for different IRIs. This research was scoped by

reviewing the impact of the IRIs starting from an IRI of one day to the IRI=2months. The 2- months IRI

was the biggest one that was considered in this research. Further insights could accrue from

investigating even bigger IRIs, such as 3 months and 4 months.

For other supply chain performance metrics (KPIs)

Moreover, the model could be used to investigate the supply chain performance in different

operationalization terms. For company X, supply chain performance is expressed in KPI terms.

Specifically, the outputted KPIs “% Delivered on time versus Requested” and “Inventory final

product”, are the ones to express the supply chain performance measurement in the case of

company X, as conducted in the S3N simulation platform. Using the same software, a different

operationalization could be performed to supply chain performance to express additional KPIs such

as costs for company X.

For other supply chains

The same model could be used not only to test different IRIs and export additional KPIs regarding the

supply chain performance of company X. It can also be used to test what happens in other

companies, outside company X, that function under the same replenishment policies. Hence, as long

as the company functions either with a s,S replenishment policy or a Min/Max policy, the model can

be generalized to show results regarding the impact of different IRIs, even IRIs that haven’t been

Page 111: Towards the Optimal Inventory Review Intervals

111

tested here, on inventory and service level KPIs. If other replenishment policies are used, the

dynamics will differ from the results shown for company X. Hence for different policies, the

simulation model needs to be adapted and tested again. For the same replenishment policies,

different supply chains could also be tested if there is the same kind of data available to investigate

how different IRIs influence relevant service level and inventory KPIs in S3N.

To conclude, the reasoning is that although the exact characteristics will differ, the resulting

dynamics of the model and the results will remain the same. Hence, the same model can be applied

to another company that would take as inputs different sales data, different product lists ,forecast

data, inventory management data, lead times of suppliers and other supplier mapping data. This

means that the conclusion on the dynamics of inventory performance using different input data due

to referring to a different company is applicable for other supply chains outside company X as well.

The dynamics will be the same as long as the inventory replenishment policy is one of the two

policies that are reviewed in this research context.

Page 112: Towards the Optimal Inventory Review Intervals

112

Page 113: Towards the Optimal Inventory Review Intervals

113

7 LIMITATIONS, RECOMMENDATIONS AND REFLECTION

During the execution of some steps of this research, not always the optimal approach was chosen

due to scoping of the research and also due to e.g. lack of certain data. Therefore, some limitations

along with further recommendations are been formulated for both the scientific and the business

world. After this, a reflection on the research project is presented.

7.1 Limitations Some limitations were observed while conducting this research:

The decision of investigating six different IRIS fluctuating from 1 day to 2 months was a

decision based on the author’s estimation that interesting results will accrue if the IRIs were

selected in a way that they grow exponentially. Hence, this research’s scope is limited until

what happens for IRIs of 1 day to 2 months. Hence, the limitation here is that we don't know

how bigger IRIs would influence the supply chain performance of company X. Moreover, the

main conclusion was that the IRIs do not have that much of influence on the KPIs: When the

IRIs are short (IRIs< 21 days), the influence is not that important. If the inventory is not

reviewed for one month or two, (21 days or 42 days) then the KPIs start to decrease faster.

Hence, it is still unknown what happens if the IRI is 80 days. On the other hand, bigger IRIs do

not seem realistic enough, because for IRIs over which the demand is higher than the target

inventories we know for sure stock outs will occur.

Another limitation was encountered while validating the model, when it was realized that

some data were missing because not all the inventory management data were reliable. The

validation process showed that and it was explained how the problem was encountered and

how the validation process changed in order to validate the model with the other data that

were available, without using the unreliable data.

An additional limitation lies on the fact that this model does not take into account the

various costs, the way that is modeled so far and depending on the inputs that it can have

now.

Moreover, the fact that in the current policy there is too much inventory makes it difficult to

have strong results regarding the impact of the IRI on supply chain performance. As it has

been mentioned, as long as you are in a month the IRIs do not really matter: it is not that

important how often the review of the inventory will take place. If the IRIs get longer in

relation to what is happening, then the KPIs start to drop quickly. For example, it may be too

long to close your eyes and not review the inventory for 2 months. However, in case of too

much inventory, it is shown that not many differences can be observed. Hence, a limitation is

that too much inventory on the current policy does not allow seeing much of the influence of

the IRIs on supply chain metrics.

Another limitation lies on the difference between the inventories scales in the two inventory

replenishment policies. This in practice makes more difficult to compare the inventory KPIs

because of the significant difference in the inventory scales. This is not a problem for the

service level KPI as it is expressed in percentage. As it has already been stated, the inventory

levels in the current policy are significantly higher than those in the Min/Max policy. Hence,

we cannot know what would happen if the current policy was used with just less inventory

instead of switching to the Min/Max policy.

Page 114: Towards the Optimal Inventory Review Intervals

114

7.2 Recommendations for future research In recommendations for future research, a clear distinction is being made between

recommendations for future research from a business perspective, for company X or for other

businesses and also from an academic perspective.

7.2.1 Future research from a business perspective The results of the simulation show that as long as the IRI is less than a month, the KPIs do not

decrease significantly. It is after an IRI of one month where service levels and inventory KPIs show a

quicker decrease. From a business perspective, this is an important implication since Accenture will

implement the “real time” Min/Max policy. The results of this research show that for company X,

that is ready to abandon the current policy and implement the Min/Max policy with the help of

Accenture, it does not matter when the inventory is being reviewed, as long as we are in a period of

one month. If the time interval get bigger than a month, then it matters more as the IRIs seem to

influence supply chain performance if they are higher than one month.

Hence, from the perspective of company X, a recommendation would be to estimate if the Min/Max

policy that is proposed to be “a real time” policy is indeed beneficial: Based on the results of this

research, it could be stated that the Min/Max policy works efficiently with the low levels of inventory

that are kept, but still, the review interval of the inventory does not seem to influence the service

level and the inventory KPI. Thus, paying for the real time policy implementation, from the IRIs

perspective does not seem beneficial enough since during one month period it is not needed to have

a real time observation on what is happening to the inventories. The results showed that the

influence of the IRIs is not important if the IRIs are less than one month.

Moreover, an additional recommendation would be to manage better the inventories towards the

optimal ones using the current policy. Since the “real time” proposition is not that important

according to these results, from the IRIs perspective, company X could fix the inventory management

processes to first try the implementation of the current (s,S) policy with lower inventories and see

how the service levels and the inventory KPIs behave. Thus, the recommendation would be to try

the implementation of the current policy but with fewer inventory levels.

7.2.2 Future research from an academic perspective From an academic perspective, a knowledge gap was tackled as there was not much literature found

regarding the impact of the IRIs on supply chain performance. Hence, using the case study that

scoped the research on company X’s boundaries, a conceptual model was developed and then was

implemented into a simulation model. Therefore, from an academic perspective one of the

deliverables of this research is a functioning simulation model. As it has been already discussed,

future research could be conducted to expand the model for other supply chains that use the same

inventory replenishment policies, even if it regards different kind of companies with different

product characteristics. The model is made in a way that the inputs could be altered and has the

same research conducted for another company. The dynamics of the model results would be the

same as long as the (s,S) and the Min/Max policy are followed by other companies.

Furthermore, one could investigate what happens for other replenishment systems as well. Using the

same model it is possible to change some inputs or parts by writing some pieces of code in C# or

adapting existent ones. This way, one could conduct experiments using a different inventory

replenishment policy.

Moreover, the same model could be used for further research with few alterations: It would be

interesting to investigate how higher IRIs influences the supply chain performance. From the

Page 115: Towards the Optimal Inventory Review Intervals

115

simulation results it is shown that IRIs that is less than one month do not influence that much the

supply chain metrics. This research investigated IRIs from one day to 2 months. Hence, it would be

interesting to see how 80 days IRI would influence the supply chain metric for the two policies that

were reviewed.

Additionally, from an academic perspective, it would be interesting to investigate the current policy

with fewer inventories, hence to optimize the inventory management of the current policy and

change the scales of inventories in this policy. It would be interesting to see how the KPIs behave in

this case and how different their behavior would be from the real time “Min/Max” policy.

To conclude, this research was scoped for a specific company; company X, in order to answer to the

research question “What are the effects of review intervals on the supply chain performance of

company X?” From a scientific perspective, it was not clear how supply chain performance could be

influenced by different IRIs. Hence, the author selected a case in order to narrow down the scope of

the research. However, after the limitations that were also presented and discussed in the previous

subchapter, someone in the future could perform another case study to test this model for another

company with different supply chain. Nevertheless, one could go further and perform a more

fundamental study and not take just one case of company, but perform a more controlled

experiment in which he is going to turn all the variables that have been thoroughly discussed during

the implementation of the experiments.

7.3 Reflection Simulation projects can consume much more time than typically presumed. In fact it has been

observed that many simulation projects take at least twice as long as originally estimated (Benneyan,

1994). That also holds for this research that was initiated as a six month MoT graduation project. In

the end, the total span of this project turned out to be a little less than a year. In this section the

author reflects on the causes of this delay and the lessons learnt throughout the long process.

Moreover, the author realized how complex it is to validate a model. Few interesting challenges

occurred during the model validation. The author learned that simulation in a real world context

imposes larger challenges than in an educational context. The availability of a large amount of data

does not necessarily make it possible to quantify all parameters or does not imply that data are

always reliable. In the end, the plan for the validation process was reconsidered and the validation

was performed without using the inventory management data that were found to be unreliable.

Furthermore, this was the first time for me to execute a project of this size and that planning the

obligatory meetings with the graduation committee caused some delay due to agenda issues. After

all, in the process of writing my thesis I understood how important time management is and how one

can avoid delays when managing his time efficiently.

Looking back, I am satisfied with the choice I made to perform my thesis project within a company

instead of at the university. The internship at Accenture and Macomi taught me and gave me a good

insight in how a company and an office works, both content wise and socially. Getting the

responsibility to deliver results and also getting the freedom to do this in the way and time I

preferred, contributed to a great and informative internship.

Page 116: Towards the Optimal Inventory Review Intervals

116

Page 117: Towards the Optimal Inventory Review Intervals

117

8 APPENDICES

A: Simulation results The two KPIs called “Final inventories” from the Inventory Turnover KPI group and the KPI called “%

Delivered on time versus Requested” from the Service Level KPI group are exported in Excel for all 12

scenarios that were tested. In particular, due to the 50 replications for every scenario, the result of

the KPI values exported for every ISBN are the average values of those 50 replications in both the

case of the inventory KPI and the service level KPI. Using Excel, the data from every experiment is

accumulated in a matrix with all the relevant details that the author wants to focus on. Hence, one

matrix is generated for each KPI in Excel whose columns are: ISBN, Lead time, Supplier, ABC

categorization, XYZ categorization, and then 12 columns (for the 12 scenarios) containing the KPI

exported data from the model.

The following table shows the number of runs in S3N that corresponds to the respective scenario

number according to the specification that has been done in chapter 4.

Scenario 1 439

Scenario 2 440

Scenario 3 441

Scenario 4 442

Scenario 5 436

Scenario 6 443

Scenario 7 444

Scenario 8 445

Scenario 9 446

Scenario 10 447

Scenario 11 469

Scenario 12 470

Table 13: Number of runs in S3N

Results for the KPI “Final inventories”

The cumulative results for the KPI “Final inventories” can be seen in the following table:

Page 118: Towards the Optimal Inventory Review Intervals

118

Table 14: KPI “Final inventories”

Results for the KPI “% Delivered on time versus Requested”

The cumulative results for the KPI “% Delivered on time versus Requested” can be seen in the

following table:

ISBN LT Supplier ABC XYZ 439 440 441 442 436 443 444 445 446 447 469 470

9780433004806 66 CHINA B X 1858,745 1862,418 1757,082 1854,565 1849,94 1844,419 55,35804 54,38745 56,50627 54,11569 49,57725 40,14078

9780433026587 66 CHINA C X 382,1815 380,7969 356,4092 380,7273 375,2138 373,2115 32,33423 32,00577 31,66885 31,02923 25,52462 23,59115

9780433026822 66 CHINA C Y 589,9538 591,9362 554,552 588,0419 586,4715 581,4385 50,68462 50,92385 50,04346 48,33808 43,15 37,18115

9780433028840 66 CHINA B X 912,6431 912,1338 854,7688 908,8315 899,7458 887,0662 80,99346 82,68462 80,85923 79,21 76,53577 65,14231

9780433034452 66 CHINA C X 218,1192 217,515 204,3614 216,8585 214,2408 212,3785 27,28346 27,275 26,35808 27,51154 22,92192 20,03808

9780435026578 22 UK A Z 1129,82 1160,345 1129,853 1152,707 1092,003 1102,745 434,1669 445,8135 419,6027 425,2146 368,8215 179,9362

9780435026950 66 MALAYSIA A X 3835,969 3735,718 3784,923 3770,569 3530,618 3317,322 1520,128 1527,448 1485,777 1482,46 1292,858 1112,611

9780435030810 66 MALAYSIA C Y 695,8865 696,2619 649,7114 694,9962 693,1812 690,4131 23,48308 23,09154 23,12962 23,58846 21,15038 18,20077

9780435031329 22 UK A Z 5024,81 5028,59 4980,087 5044,8 4971,56 4896,554 513,6569 507,8069 494,6746 473,7554 451,3054 374,4273

9780435032050 22 UK B Z 830,9988 830,7069 825,3918 836,1323 807,9842 789,7231 118,7681 124,1592 120,0369 111,3285 96,99192 60,04577

9780435041151 22 UK C X 34,14385 34,23269 34,4386 33,935 32,54923 30,025 12,28808 12,05654 11,86346 12,39038 10,69115 6,649615

9780435046873 22 UK A X 1849,816 1953,519 1928,087 1897,877 1810,427 1717,68 596,1896 583,8258 575,5162 548,2519 530,8662 448,8308

9780435075705 66 MALAYSIA B X 1062,038 1062,121 1000,71 1060,761 1053,184 1048,777 49,53154 50,30192 48,63385 49,18231 43,82038 38,49731

9780435075750 66 MALAYSIA B Y 954,24 956,5512 895,5312 950,4627 949,0092 943,8904 40,54846 39,19346 39,72538 38,00231 35,92808 29,49038

9780435075781 66 MALAYSIA B Z 1394,49 1393,887 1299,028 1395,605 1389,611 1387,711 38,515 37,96846 40,43538 37,28692 34,33192 29,50808

9780435075804 66 MALAYSIA B Z 908,9319 910,3112 847,8306 910,3008 904,6315 905,0085 31,705 30,58423 30,04654 29,76346 27,02692 22,86577

9780435076337 66 MALAYSIA B Y 1002,556 1002,506 944,8322 1003,028 997,4035 990,2242 59,87654 61,44077 60,18538 56,65577 53,22269 45,21538

9780435118105 22 UK B Y 228,6681 230,2154 227,3628 226,5073 225,1823 221,5412 28,46 28,57846 27,98462 27,25077 24,96769 21,82192

9780435125936 22 UK B X 1472,845 1480,072 1474,022 1477,365 1469,849 1462,309 85,56423 83,33538 84,57615 80,91423 76,74 62,13846

9780435208622 66 MALAYSIA A Y 1004,36 1016,858 986,7032 1013,149 980,5319 979,9185 246,4712 242,2558 244,1673 237,1396 213,6312 191,6608

9780435208653 66 MALAYSIA A Z 1061,69 1055,306 1001,929 1077,212 1023,176 1009,723 301,3096 296,0958 297,2762 293,4562 244,1908 207,8838

9780435224875 66 CHINA A X 1615,895 1603,995 1632,842 1597,69 1531,531 1420,533 681,4742 672,7442 666,5638 673,8204 596,3812 508,1742

9780435308056 66 CHINA A X 998,2065 1022,732 1052,492 1057,745 951,1512 993,3642 883,4073 883,8158 895,9523 855,4296 801,5146 661,8638

9780435312305 22 UK B X 515,8238 524,3373 545,441 538,9346 518,5723 503,5504 88,75615 88,05923 85,00192 85,30269 76,44577 53,95577

9780435357535 22 UK A X 804,6058 796,7727 823,7594 785,4573 763,0231 748,8435 178,8592 181,6865 179,85 167,6365 154,8885 125,6812

9780435389215 22 UK A X 3254,626 3234,462 3271,831 3204,244 3145,068 3067,916 643,0662 633,5396 637,5877 616,1785 584,6965 497,7377

9780435468606 22 UK A Z 675,81 688,5119 681,7552 675,7715 660,4227 656,4969 245,1277 249,7119 250,6858 248,2927 231,8219 207,3558

9780435501105 22 UK B Z 387,985 390,5458 387,7358 389,8727 376,2931 363,1342 101,7146 98,35577 96,38923 92,81692 83,69423 65,35692

9780435510220 66 CHINA A Z 917,5581 918,5562 862,0714 920,7204 908,8019 901,9523 97,15846 95,35538 93,17385 89,03808 80,75269 68,55846

9780435633097 22 UK B X 326,5908 330,4138 341,724 331,4731 328,91 321,6658 77,49423 81,14615 81,29654 75,42923 69,26308 50,60731

9780435675486 22 UK C X 97,41231 97,68923 95,8974 96,63192 93,06385 91,88346 21,02962 20,97731 20,14846 19,08654 17,83577 14,87808

9780435907785 22 UK B Y 177,3804 180,1835 179,8428 178,4981 174,4854 170,0888 41,35115 40,49 40,11692 39,85077 36,70346 31,72077

9780435910389 66 CHINA C Y 471,3185 471,2854 439,387 470,3138 467,5712 464,1358 31,81923 32,415 30,02846 31,36423 30,19423 20,72808

9780435912307 66 CHINA B Z 736,7738 736,8573 690,6968 739,245 728,4069 723,2142 101,9792 98,34654 100,5758 96,39192 90,70423 75,25154

9780435912321 66 CHINA B Z 809,385 810,5046 758,695 810,6358 802,1354 793,2154 100,0631 98,58385 97,19462 99,51808 89,75038 71,92923

9780435914509 66 CHINA B Y 2802,743 2801,983 2616,09 2801,418 2788,841 2783,653 66,71615 66,40615 67,265 64,57231 60,16846 50,04385

9780435914707 66 CHINA B Y 2484,695 2483,24 2321,202 2481,287 2475,207 2470,887 50,49615 50,88385 50,49154 48,165 42,28192 36,87692

9780582026605 66 MALAYSIA A Y 1364,131 1375,078 1380,883 1358,832 1274,587 1206,22 948,2765 936,3788 935,3496 891,8669 832,6323 719,9954

9780602206604 66 CHINA A Y 1694,622 1702,861 1629,114 1704,584 1678,381 1670,872 248,2592 243,16 240,4927 236,9962 216,5938 182,2958

9780602259587 66 CHINA A Z 712,6954 714,0212 665,562 713,7181 708,5012 707,2942 48,24308 48,28615 47,62923 46,45615 41,90538 39,38962

9780602275136 22 UK A Y 925,685 931,9577 946,1964 907,4088 879,3231 853,1192 246,6165 248,8488 246,335 227,0588 198,7219 161,275

9780602290719 66 CHINA C X 92,62538 92,43192 87,5424 93,42038 90,78654 90,40731 12,78692 12,65154 12,06962 11,96269 11,19192 9,099231

9781405856461 22 UK C Y 90,15962 89,26654 89,832 90,68962 85,28692 80,94846 27,40308 27,56154 26,72192 24,18654 22,01538 20,20308

9781405856874 66 MALAYSIA C X 114,6812 112,2542 112,2808 108,6785 107,5565 98,03808 67,56308 67,24808 68,83962 62,09769 58,03154 49,58269

9781408217269 22 UK C Y 59,885 59,70192 60,246 59,66538 54,34038 50,66769 27,23731 27,23231 26,29192 25,27769 22,25423 16,99769

9781408217320 22 UK C X 134,3254 136,3462 136,1368 135,0462 129,2592 124,8804 53,39423 52,75769 52,42538 49,66885 45,50769 40,02346

9781408248744 22 UK B Z 507,8919 518,1792 511,703 519,3723 498,2381 478,4177 231,5262 238,1473 229,5854 220,0012 206,4727 195,4062

9781408248812 66 CHINA A Y 589,145 576,9277 568,966 579,7119 545,0673 517,4173 478,0677 483,1115 472,6712 461,3477 432,455 386,6662

9781408278239 66 MALAYSIA A X 24607,87 24382,67 24341,59 25683,4 22896 21931,92 12569,41 12952,27 12688,71 12354,6 11890,37 9323,482

9781447959601 22 UK A Z 26068,9 26179,31 25816,72 26221,72 25526,06 25438,01 4704,781 4895,988 4839,007 4568,122 4114,312 3014,385

9781846905063 22 UK B X 485,8931 487,1881 511,946 487,2758 481,8092 473,1512 111,0165 110,9581 104,4846 107,3504 100,1227 93,85885

9781846905094 22 UK C X 318,0335 317,3015 315,9206 317,3854 307,4192 286,2773 173,0942 183,6073 174,7315 169,9654 151,2135 132,6096

9781846905520 66 MALAYSIA B X 387,0992 382,3712 390,2924 380,0908 348,8831 339,3246 121,6558 125,2212 124,27 121,1342 104,0515 82,51731

9781846905766 22 UK B Z 1081,311 1112,494 1096,296 1116,568 1012,111 908,4523 494,1715 509,7658 495,4615 467,4288 442,1923 266,1185

9781846905971 66 MALAYSIA A Z 1922,84 1914,34 1801,969 1906,088 1856,953 1772,346 698,4488 691,9042 667,6035 684,8373 600,6815 527,9535

9781846908149 22 UK B Y 142,9304 134,7312 144,0028 141,1923 137,7258 134,5212 96,96077 97,25885 96,68885 91,90192 87,07846 74,07808

1871,972 1871,848 1840,112 1894,629 1806,934 1768,373 519,9419 530,109 521,2636 505,2142 470,8778 373,4381

Page 119: Towards the Optimal Inventory Review Intervals

119

Table 15: KPI “% Delivered on time versus Requested”

ISBN LT Supplier ABC XYZ 439 440 441 442 436 443 444 445 446 447 469 470

9780433004806 66 CHINA B X 0,861176 0,859216 0,865633 0,867059 0,855294 0,867451 0,865137 0,875863 0,849246 0,826884 0,834664 0,764323

9780433026587 66 CHINA C X 0,726154 0,721154 0,726954 0,735385 0,728077 0,701923 0,665572 0,707787 0,674338 0,653117 0,655959 0,559393

9780433026822 66 CHINA C Y 0,767308 0,758462 0,737079 0,758462 0,761538 0,769615 0,720006 0,736406 0,739915 0,735751 0,685922 0,619518

9780433028840 66 CHINA B X 0,755769 0,745 0,7386 0,735769 0,755 0,753846 0,742527 0,728482 0,754358 0,75566 0,727046 0,694289

9780433034452 66 CHINA C X 0,685 0,678462 0,655933 0,693846 0,664615 0,660385 0,653489 0,614622 0,630325 0,612201 0,592465 0,539014

9780435026578 22 UK A Z 0,8275 0,8635 0,865104 0,812267 0,819716 0,821354 0,846132 0,802888 0,820534 0,790457 0,701275 0,521415

9780435026950 66 MALAYSIAA X 1 1 0,99871 1 0,996923 0,997444 0,998462 0,999502 0,998289 0,996441 0,992273 0,947435

9780435030810 66 MALAYSIAC Y 0,445 0,446923 0,459667 0,465385 0,459615 0,488846 0,418267 0,44405 0,44004 0,431355 0,414781 0,40262

9780435031329 22 UK A Z 0,891154 0,891154 0,895016 0,886154 0,860385 0,893077 0,862696 0,891366 0,854224 0,829445 0,792633 0,669654

9780435032050 22 UK B Z 0,829231 0,871538 0,857868 0,852692 0,852308 0,879231 0,859002 0,823576 0,836501 0,808024 0,711198 0,513016

9780435041151 22 UK C X 0,471667 0,450417 0,460957 0,444236 0,454194 0,460757 0,438489 0,446231 0,43281 0,426342 0,391283 0,323278

9780435046873 22 UK A X 0,99 0,987308 0,9892 0,987692 0,988846 0,982308 0,986923 0,990769 0,989615 0,984421 0,97831 0,933639

9780435075705 66 MALAYSIAB X 0,825385 0,798462 0,795716 0,823077 0,808846 0,797308 0,792086 0,791568 0,802678 0,792401 0,780112 0,768507

9780435075750 66 MALAYSIAB Y 0,89 0,887692 0,864295 0,893846 0,889231 0,883077 0,882221 0,891298 0,878768 0,871379 0,852877 0,815089

9780435075781 66 MALAYSIAB Z 0,778462 0,79 0,784066 0,774615 0,784615 0,788462 0,778214 0,78045 0,751413 0,765538 0,74199 0,694157

9780435075804 66 MALAYSIAB Z 0,785385 0,781538 0,766415 0,793846 0,791923 0,761923 0,759887 0,780858 0,774981 0,739024 0,694438 0,655718

9780435076337 66 MALAYSIAB Y 0,94 0,936538 0,93266 0,928462 0,933462 0,929231 0,921799 0,924465 0,925474 0,926682 0,911765 0,844564

9780435118105 22 UK B Y 0,798846 0,776154 0,792827 0,811154 0,808077 0,802308 0,805141 0,784258 0,786525 0,787864 0,74783 0,712256

9780435125936 22 UK B X 0,856538 0,862692 0,88431 0,871923 0,857692 0,855385 0,871181 0,868725 0,85393 0,849834 0,821219 0,757

9780435208622 66 MALAYSIAA Y 0,998077 0,997692 0,984143 0,997692 0,998077 0,998846 0,997494 0,997268 0,996153 0,995241 0,985225 0,936519

9780435208653 66 MALAYSIAA Z 0,930769 0,930769 0,903115 0,920385 0,928462 0,931923 0,91744 0,90956 0,891917 0,886534 0,885514 0,80705

9780435224875 66 CHINA A X 0,996154 0,999615 0,9978 0,999615 0,996154 0,997308 0,999231 0,999203 0,997544 0,998645 0,994246 0,94835

9780435308056 66 CHINA A X 0,931942 0,922823 0,929029 0,925004 0,920689 0,873107 0,934615 0,933846 0,934123 0,931403 0,924675 0,896293

9780435312305 22 UK B X 0,808077 0,774231 0,791 0,771538 0,808462 0,783846 0,770186 0,787165 0,796298 0,754617 0,715957 0,55363

9780435357535 22 UK A X 0,989231 0,986538 0,9886 0,99 0,987308 0,985769 0,989495 0,985086 0,986101 0,988557 0,970624 0,882503

9780435389215 22 UK A X 0,982308 0,985385 0,9852 0,988462 0,987692 0,985 0,982692 0,982616 0,983846 0,982693 0,977058 0,932108

9780435468606 22 UK A Z 0,872308 0,865128 0,875696 0,862051 0,876923 0,862862 0,873983 0,868493 0,840448 0,831002 0,810942 0,744493

9780435501105 22 UK B Z 0,840385 0,840385 0,844692 0,816154 0,834615 0,831154 0,798315 0,855278 0,819586 0,794241 0,769419 0,646939

9780435510220 66 CHINA A Z 0,735769 0,740385 0,699669 0,741154 0,736538 0,725 0,708718 0,733744 0,73209 0,701158 0,708517 0,654683

9780435633097 22 UK B X 0,706923 0,698077 0,7232 0,712308 0,724231 0,738077 0,711149 0,733122 0,728912 0,695299 0,669698 0,597546

9780435675486 22 UK C X 0,629231 0,592308 0,5924 0,561538 0,607692 0,566154 0,553925 0,553472 0,567344 0,561821 0,542977 0,448756

9780435907785 22 UK B Y 0,858077 0,850769 0,850383 0,848077 0,855 0,849231 0,868918 0,85497 0,848756 0,839822 0,838041 0,787454

9780435910389 66 CHINA C Y 0,643462 0,643462 0,628194 0,633077 0,628846 0,650769 0,640171 0,636324 0,645785 0,596567 0,607007 0,503436

9780435912307 66 CHINA B Z 0,867308 0,893077 0,868551 0,878846 0,893846 0,898462 0,86993 0,877384 0,880643 0,869628 0,848366 0,797042

9780435912321 66 CHINA B Z 0,896538 0,895385 0,876417 0,904615 0,875769 0,905385 0,873958 0,875696 0,896887 0,868772 0,854575 0,785876

9780435914509 66 CHINA B Y 0,957308 0,952692 0,942418 0,946923 0,952308 0,956538 0,939072 0,961523 0,942764 0,955324 0,921428 0,868218

9780435914707 66 CHINA B Y 0,919615 0,927308 0,913592 0,933462 0,940769 0,926538 0,936642 0,923104 0,927744 0,916117 0,892844 0,832561

9780582026605 66 MALAYSIAA Y 0,999614 1 0,997853 0,996648 0,994374 0,963027 1 1 1 1 0,998371 0,973484

9780602206604 66 CHINA A Y 0,999615 0,998462 0,990138 0,999615 0,999231 0,998846 0,998515 0,997975 0,993342 0,992721 0,978426 0,928688

9780602259587 66 CHINA A Z 0,732308 0,741538 0,736584 0,743077 0,761154 0,750769 0,738592 0,734159 0,730107 0,736085 0,676948 0,654685

9780602275136 22 UK A Y 0,998846 0,997308 0,996516 0,999231 0,998077 0,998462 0,998846 0,997912 0,99782 0,993063 0,961282 0,864849

9780602290719 66 CHINA C X 0,425385 0,429231 0,4356 0,410769 0,444231 0,428846 0,398652 0,411384 0,411679 0,382649 0,391231 0,357707

9781405856461 22 UK C Y 0,730385 0,735 0,730175 0,717308 0,754615 0,753846 0,711918 0,726545 0,70413 0,702935 0,642389 0,565448

9781405856874 66 MALAYSIAC X 0,729703 0,765372 0,743002 0,770385 0,715265 0,724285 0,724033 0,705801 0,694307 0,696349 0,688165 0,625796

9781408217269 22 UK C Y 0,756526 0,751154 0,754213 0,747623 0,748816 0,721401 0,75888 0,733622 0,713979 0,713686 0,654758 0,557439

9781408217320 22 UK C X 0,787692 0,793077 0,791632 0,788077 0,795385 0,789846 0,808157 0,789903 0,771633 0,768365 0,714966 0,68369

9781408248744 22 UK B Z 0,923077 0,929615 0,908853 0,916154 0,918077 0,916923 0,914974 0,918341 0,920489 0,913533 0,895631 0,850703

9781408248812 66 CHINA A Y 0,985132 0,98848 0,977024 0,975384 0,958371 0,912369 0,999028 0,995385 0,998462 0,996481 0,993379 0,971381

9781408278239 66 MALAYSIAA X 0,992308 0,996154 0,9942 0,995385 0,991605 0,984972 0,997084 0,991507 0,988785 0,992762 0,992486 0,928368

9781447959601 22 UK A Z 0,919231 0,938846 0,9196 0,923846 0,931538 0,915385 0,922712 0,918512 0,902116 0,899509 0,854922 0,697937

9781846905063 22 UK B X 0,797692 0,778846 0,770591 0,780385 0,772692 0,798846 0,790315 0,775633 0,78822 0,771251 0,750024 0,717061

9781846905094 22 UK C X 0,749164 0,745309 0,762387 0,770212 0,758096 0,748036 0,74325 0,723961 0,732803 0,737269 0,730017 0,631994

9781846905520 66 MALAYSIAB X 0,923991 0,920932 0,886897 0,897895 0,905524 0,849512 0,880008 0,881166 0,896167 0,858209 0,826629 0,695664

9781846905766 22 UK B Z 0,887308 0,855618 0,861622 0,8699 0,852296 0,830538 0,869225 0,847245 0,833529 0,835189 0,753339 0,542013

9781846905971 66 MALAYSIAA Z 0,898462 0,928462 0,924393 0,951154 0,924615 0,914231 0,910136 0,895197 0,902904 0,875741 0,884591 0,839874

9781846908149 22 UK B Y 0,898518 0,906081 0,90537 0,885383 0,859971 0,832522 0,921538 0,922133 0,913496 0,911911 0,886049 0,853214

Page 120: Towards the Optimal Inventory Review Intervals

120

B: “% Delivered on time versus Requested” KPI comparison between the

two replenishment policies regarding ABC classification In this Appendix, a more extensive analysis is implemented to compare the “% Delivered on time

versus Requested” KPI results regarding the ABC classification according to the different product

clusters. (Regarding the ABC classification, the items are classified based on demand (consumption)

rate: Inventory control is based on a form of Pareto analysis. The inventory items are divided into

three categories (A, B, and C), according to a criterion such as revenue generation, turnover, or value.

Typically, 'A' items represent 20 percent in terms of quantity and 75 to 80 percent in terms of the

value. )

Further, to shed more light on the different product clusters, the author decided to visualize the

separate product clusters in order to be able to compare the A products in the current policy to the A

products in the Min/Max policy, the B products in the current policy to the B products in the

Min/Max policy, and the C products in the current policy to the C products in the Min/Max policy.

“% Delivered on time versus Requested” KPI results regarding the ABC classification:

Comparison between A products:

Figure 62: “% Delivered on time versus Requested” KPI averages for A products for the current policy.

Page 121: Towards the Optimal Inventory Review Intervals

121

Figure 63: “% Delivered on time versus Requested” KPI averages for A products for the Min/Max policy.

In the figures presented above, the x-axis represents the IRIs in days and the y-axis the service level

in percentage expressed by the KPI “% Delivered on time versus Requested”. Hence, for example, the

1st dot on the scatter plot presented here in the previous graph refers to around 0.95% average

service level for all A products for an IRI of 1 day for the Min/Max policy (7th scenario). Hence, again

the y-axis represents the IRI in days and at the same time the number of scenarios that have run. For

the last graph of the Min/Max policy for example, each one of the dots in the scatter plot intersects

vertically the x-axis at the value of the IRI in days that corresponds to the scenarios 7 to 12.

“% Delivered on time versus Requested” KPI results regarding the ABC classification:

Comparison between B products:

Figure 64: “% Delivered on time versus Requested” KPI averages for B products for the current policy.

Figure 65: “% Delivered on time versus Requested” KPI averages for B products for the Min/Max policy.

Page 122: Towards the Optimal Inventory Review Intervals

122

“% Delivered on time versus Requested” KPI results regarding the ABC classification:

Comparison between C products:

Figure 66: “% Delivered on time versus Requested” KPI averages for C products for the current policy.

Figure 67: Delivered on time versus Requested” KPI averages for C products for the Min/Max policy.

Page 123: Towards the Optimal Inventory Review Intervals

123

C: “% Delivered on time versus Requested” KPI comparison between the two

replenishment policies regarding XYZ classification In this Appendix, An analysis is implemented to compare the “% Delivered on time versus Requested”

KPI results regarding the XYZ classification according to the different product clusters. Regarding the

XYZ classification, the XYZ classification is an indicator of the items ‘demand predictability and

expected forecast accuracy. Hence, the items are classified according to demand variability in a way

that:

X items include all items in which use is relatively constant and fluctuates only rarely. The

probability of correct predictions is very high. (high forecast accuracy expected)

Y items include all items with substantial fluctuations in demand due to seasonal reasons or

because of trends in product use. The probability of correct predictions is medium.

Z-products finally are all articles with highly irregular use. The reliability of predictions in this

case is low (low forecast accuracy expected).

Moreover, product clusters are visualized in plots separately in order to be comparable: the X

products in the current policy are compared to the X products in the Min/Max policy, the Y products

in the current policy to the Y products in the Min/Max policy, and the Z products in the current policy

to the Z products in the Min/Max policy.

“% Delivered on time versus Requested” KPI results regarding the XYZ classification:

Comparison between X products:

Figure 68: “Delivered on time versus Requested” KPI averages for C products for the Min/Max policy.

Page 124: Towards the Optimal Inventory Review Intervals

124

Figure 69: “% Delivered on time versus Requested” KPI averages for X products for the Min/Max policy.

Figure 70: “% Delivered on time versus Requested” KPI averages for Y products for the current policy.

Page 125: Towards the Optimal Inventory Review Intervals

125

Figure 71: “% Delivered on time versus Requested” KPI averages for Y products for the Min/Max policy.

Figure 72: “% Delivered on time versus Requested” KPI averages for Z products for the current policy.

Page 126: Towards the Optimal Inventory Review Intervals

126

Figure 73: “% Delivered on time versus Requested” KPI averages for Z products for the Min/Max policy.

In the figures presented above, the x-axis represents the IRIs in days and the y-axis the service level

in percentage expressed by the KPI “% Delivered on time versus Requested”. Hence, for example, the

1st dot on the scatter plot presented here, in the last graph refers to around 0.85% average service

level for all Z products for an IRI of 1 day for the Min/Max policy (7th scenario). Hence, again the y-

axis represents the IRI in days and at the same time the number of scenarios that have run. For the

last graph of the Min/Max policy for example, each one of the dots in the scatter plot intersects

vertically the x-axis at the value of the IRI in days that corresponds to the scenarios 7 to 12.

Page 127: Towards the Optimal Inventory Review Intervals

127

D: “% Delivered on time versus Requested” KPI comparison between the

two replenishment policies regarding Lead Time classification In this Appendix, an analysis is implemented to compare the “% Delivered on time versus Requested”

KPI results regarding the two different lead times that have been examined for both the

replenishment policies.

As it has been stated, the author decided to vary the lead time by choosing three different suppliers:

The UK litho-production supplier with a lead time of 22 working days that correspond to one month.

Further, the China supplier and the Malaysia supplier were selected and both have a lead time of 66

working days that correspond to 3 months.

Hence, product clusters are visualized in plots separately in order to be comparable: The products

with a lead time of 22 days in the current policy are compared to the products with this lead time of

22 days in the Min/Max policy. Further, products with a lead time of 66 days in the current policy are

compared to the products with this lead time of 66 days in the Min/Max policy.

The following figures show this comparison.

“% Delivered on time versus Requested” KPI results regarding the Lead time: Comparison

between products with LT = 22days:

The following two figures present the “% Delivered on time versus Requested” KPI averages for all

products that have a LT=22days for the current policy and for the Min/Max policy respectively. The

x-axis represents the IRIs in days and the y-axis the service level in percentage expressed by the KPI

“% Delivered on time versus Requested”.

Figure 74: “% Delivered on time versus Requested” KPI averages for products with LT=22days for the current policy.

Page 128: Towards the Optimal Inventory Review Intervals

128

Figure 75: “% Delivered on time versus Requested” KPI averages for products with LT=22days for the Min/Max policy.

“% Delivered on time versus Requested” KPI results regarding the Lead time: Comparison

between products with LT = 66days:

The following two figures present the “% Delivered on time versus Requested” KPI averages for all

products that have a LT=66 days (3 months) for the current policy and for the Min/Max policy

respectively.

Figure 76: “% Delivered on time versus Requested” KPI averages for products with LT=66 days for the current policy.

0,6

0,65

0,7

0,75

0,8

0,85

0,9

0 5 10 15 20 25 30 35 40 45

Inventory Review Intervals in days

“% Delivered on time versus Requested” KPI averages for products with LT 22 days for Min/Maxt policy

Page 129: Towards the Optimal Inventory Review Intervals

129

Figure 77: “% Delivered on time versus Requested” KPI averages for products with LT=66 days for the current policy.

0,6

0,65

0,7

0,75

0,8

0,85

0,9

0 5 10 15 20 25 30 35 40 45

Inventory Review Intervals in days

“% Delivered on time versus Requested” KPI averages for products with LT 66 days for Min/Max policy

Page 130: Towards the Optimal Inventory Review Intervals

130

E: “Inventory final product” KPI comparison between the two

replenishment policies regarding ABC classification In this Appendix, an extensive analysis is implemented to compare the “Inventory final product” KPI

results regarding the ABC classification according to the different product clusters. Further, to shed

more light on the different product clusters, the author decided to visualize the separate product

clusters in order to be able to compare the A products in the current policy to the A products in the

Min/Max policy, the B products in the current policy to the B products in the Min/Max policy, and

the C products in the current policy to the C products in the Min/Max policy.

the “Inventory final product” KPI results regarding the ABC classification: Comparison

between A products:

In the following figures, the x-axis represents the IRIs in days and the y-axis the inventory level in

expressed by the KPI “Inventory final product”. Hence, for example, the 1st dot on the 1st scatter plot

that follows refers to around 4000 units for all A products for an IRI of 1 day for the current policy

(1st scenario). Hence, again the y-axis represents the IRI in days and at the same time the number of

scenarios that have run. For the second graph of the Min/Max policy for example, each one of the

dots in the scatter plot intersects vertically the x-axis at the value of the IRI in days that corresponds

to the scenarios 7 to 12.

Figure 78: ” Inventory final product” KPI averages for A products for current policy

Page 131: Towards the Optimal Inventory Review Intervals

131

Figure 79: “Inventory final product” KPI averages for A products for Min/Max policy

the “Inventory final product” KPI results regarding the ABC classification: Comparison

between B products:

The following figures visualize the “Inventory final product” KPI behavior for the B products at the

current policy and then at the Min/Max policy.

Figure 80: ” Inventory final product” KPI averages for B products for current policy

Page 132: Towards the Optimal Inventory Review Intervals

132

Figure 81: ” Inventory final product” KPI averages for B products for current policy

the “Inventory final product” KPI results regarding the ABC classification: Comparison

between C products:

The following figures visualize the “Inventory final product” KPI behavior for the C products at the

current policy and then at the Min/Max policy.

Figure 82: ” Inventory final product” KPI averages for C products for current policy

Page 133: Towards the Optimal Inventory Review Intervals

133

Figure 83: “Inventory final product” KPI averages for C products for Min/Max policy

0

100

200

300

400

500

600

0 5 10 15 20 25 30 35 40 45

Inventory review Intervals in days

"Inventory final product" KPI averages for C products for Min/Max policy

Page 134: Towards the Optimal Inventory Review Intervals

134

F: “Inventory final product” KPI comparison between the two

replenishment policies regarding XYZ classification In this Appendix, an analysis is implemented to compare the “Inventory final product” KPI results

regarding the XYZ classification according to the different product clusters. Regarding the XYZ

classification, the XYZ classification is an indicator of the items ‘demand predictability and expected

forecast accuracy.

Moreover, product clusters are visualized in plots separately in order to be comparable: the X

products in the current policy are compared to the X products in the Min/Max policy, the Y products

in the current policy to the Y products in the Min/Max policy, and the Z products in the current policy

to the Z products in the Min/Max policy.

“Inventory final product” KPI results regarding the XYZ classification: Comparison between X

products:

Figure 84: “Inventory final product” KPI averages for X products for current policy.

Figure 85: “Inventory final product” KPI averages for X products for Min/Max policy.

Page 135: Towards the Optimal Inventory Review Intervals

135

“Inventory final product” KPI results regarding the XYZ classification: Comparison between Y

products:

Figure 86: “Inventory final product” KPI averages for Y products for the current policy.

Figure 87: “Inventory final product” KPI averages for Y products for Min/Max policy.

Page 136: Towards the Optimal Inventory Review Intervals

136

“Inventory final product” KPI results regarding the XYZ classification: Comparison between Z

products:

Figure 88: “Inventory final product” KPI averages for Z products for the current policy.

Figure 89: “Inventory final product” KPI averages for Z products for the Min/Max policy.

Page 137: Towards the Optimal Inventory Review Intervals

137

G: “Inventory final product” KPI comparison between the two

replenishment policies regarding Lead Time classification In this Appendix, an analysis is implemented to compare the “Inventory final product” KPI results

regarding the two different lead times that have been examined for both the replenishment policies.

The products with a lead time of 22 days in the current policy are compared to the products with this

lead time of 22 days in the Min/Max policy. Further, products with a lead time of 66 days in the

current policy are compared to the products with this lead time of 66 days in the Min/Max policy.

“Inventory final product” KPI results regarding the Lead time: Comparison between products

with LT = 22days:

Figure 90: “Inventory final product” KPI averages for products with LT=22days for the current policy.

Figure 91: “Inventory final product” KPI averages for products with LT=22days for the Min/Max policy.

Page 138: Towards the Optimal Inventory Review Intervals

138

“Inventory final product” KPI results regarding the Lead time: Comparison between products

with LT = 66 days:

Figure 92: “Inventory final product” KPI averages for products with LT=66 days for the current policy.

Figure 93: “Inventory final product” KPI averages for products with LT=66 days for the Min/Max policy.

Page 139: Towards the Optimal Inventory Review Intervals

139

9 Bibliography Anderson, D., Sweeney, D. & Williams, T., 2002. An Introduction to Management Science:

Quantitative Approaches to Decision Making. tenth red. Ohio: South-Western Publishing Company

Cincinnati.

Anon., 2015. Creating insights in supply chains, Rotterdam : Macomi B.V..

Anon., 2015. Operations and Supply Chain Terms. [Online]

Available at: http://scm.ncsu.edu/scm-articles/scm-terms

[Geopend 1 10 2015].

APICS, sd Supply chain management. [Online]

Available at: http://www.apics.org/dictionary/dictionary-information?ID=4202

Axsater, S., 1993. Optimization of order‐up‐to‐s policies in two‐echelon inventory systems with

periodic review.. Naval Research Logistics, 40(2), pp. 245-253.

Ballou, R., 2004. Business Logistics/Supply Chain Management. Englewood Cliffs, NJ: Prentice Hall.

Benneyan, J. C., 1994. An introduction to using computer simulation in healthcare: patient wait case

study.. Boston, Massachusetts: Harvard community helath plan.

Blanchard, S., B. & Fabrycky, W., 1998. Systems Engineering and Analysis. 3d red. New Jersey:

Prentice Hall Inc..

Cachon, G. & Fisher, M. (., 1997. Campbell soup’s continuous replenishment program: Evaluation and

enhanced inventory decision rules.. Production and Operations Management, 6(3), pp. 266-276.

Chan, 2003. Performance Measurement in a Supply Chain. International Journal of Advanced

Manufacturing Technology, pp. 534-548.

Chase, R., Aquilano, N. & Jacobs, R., 2001. Fundamentals of Operations Management For Competitive

Advantage. ninth red. New York: McGraw-Hill.

Chopra, S. & Meindl, P., 2007. Supply ChainManagement: Strategy, Planning and Operation (3rd ed.).

New Jersey: Pearson Prentice Hall.

Clevert, et al., 2007. Cost analysis in interventional radiology—A tool to optimize management costs.

European Journal of Radiology 61, p. 144–149.

Coyle, J., Bardi, E., Jr., L. & C.J., 2003. The Management of Business Logistics: A Supply Chain

Perspective. 7th red. Ohio: South-Western..

Croom, S., Romano, P. & Giannakis, M., 2000. Supply chain management: an analytical framework for

critical literature review. European Journal of Purchasing & Supply Management,6(1), pp. 67-83.

Croxton, K. L. et al., 2001. The Supply Chain Management Processes. The International Journal of

Logistics Management, Vol 12, Number 2, pp. 13-35.

Dalkey, N. & Helmer, O., 1963. An experimental application of the delphi method to the use of

experts.. Management Science, 9(3), p. 458–467.

Davis, T., 1993. Effective supply chain management. Sloan Management Review, pp. 35-46.

Dietz, J., 2006. Enterprise Ontology: Theory and Methodology. Berlin: Springer.

Page 140: Towards the Optimal Inventory Review Intervals

140

Dietz, J. L., 2006. Enterprise Ontology: Theory and Methodology.. ISBN 35400291695 red. sl:Springer.

Dijk, E., Leeuw, S. & Durlinger, P., 2007. Voorraadbeheer in perspectief: Zeven invalshoeken van het

vak.. Deventer: Slimstock B.V..

Doorewaard, H. & Verschuren, P., 2010. Designing a research project. the Hague: Eleven

International Publishing.

Dorris, M., 2013. http://www.digitalgov.gov/. [Online]

Available at: http://www.digitalgov.gov/2013/12/02/whats-your-strategy-operational-excellence-

product-leadership-or-customer-intimacy/

[Geopend 2 October 2013].

Elsayed, E. & Boucher, T., 1994. Analysis and Control of Production Systems. London: Prentice Hall .

Everette, S. & Gardner, J., 1990. Evaluating forecast performance in an inventory control system.

Management science , 36(4), pp. 490-499.

Farahani, R. Z., Rezapour, S. & Kardar, L., 2011. Logistics operations and managemen: concepts and

models. London : Elsevier.

Fawcett, S., Ellram, L. & Ogden, J. :., 2007. Supply Chain Management: From Vision to Implemention..

New Jersey: Pearson Education, Inc.

Fuller, J. & Martinec, C. L., 2005. Operations Research And Operations Management: From Selective

Optimization To System Optimization. Journal of Business & Economics Research, Volume 3, Number

7.

Gaur, V., Kesavan, S., Raman, A. & Fisher, M., 2007. Estimating demand uncertainty using judgmental

forecasts.. Manufacturing & Service Operations Managemen, 9(4), pp. 480-491.

Goetschalckx, M., 2011. Supply Chain Engineering(International Series in Operations Research &

Management Science). sl:Springer.

Goor, A. & Weijers, S., 1998. Poly Logistiek Zakboekje, Arnhem: PBNA.

Harrington, L., 1996. Untapped savings abound. Industry Week, Volume 15th July, pp. 53-58.

Hillston, J., 2003. COMPUTER SCIENCE 4 & MSC: MODELLING AND SIMULATION. [Online]

Available at: http://www.inf.ed.ac.uk/teaching/courses/ms/notes/note14.pdf

[Geopend 20 October 2015].

Hoekstra, S. & Romme, J., 1991. Integral logistic structures: developing customer-oriented good flow.

London: McGraw-Hill.

Hopp, W. & M.L, S., 2007. Factory Physics: Foundations of Manufacturing Management. 3rd red.

Chicago: McGraw Hill.

Houlihan, J., 1988. International Supply Chains: A New Approach. Management Decision 26(3), pp.

13-19.

Hyndman, R. J. & Koehler, A. B., 2006. Another look at measures of forecast accuracy. Internatuional

Journal of Forecasting , 22(4), pp. 679-688.

Janssen, F., Heuts, R. & Kok de, T., 1998. On the (R, s, Q) inventory model when demand is modelled

as a compound Bernoulli process.. European Journal of Operational Research, 104(3), pp. 423-436.

Page 141: Towards the Optimal Inventory Review Intervals

141

Janssen, T., 2016. Enterprise Engineering: Sustained Improvement of Organizations. Switzerland:

Springer.

Jensen, P. & J.F., B., 2003. Operations Research Models and Methods. sl:Copyright 2003.

Johnson, J., Wood, D., Wardlow, D. & Murphy, P., 1999. Contemporary Logistics. seventh red.

Prentice Hall: sn

Johnson, M. & Pyke, D., 2001. Supply Chain Management.. p.edited by C. Harris and S. Gass. red.

sl:Encyclopedia of MS/OR.

Kapmeier, F., 2006. Research Based on Two Pillars: Combining Qualitative Empirical Social Research

and Simulation in Strategic Management, Stuttgart: Universität Stuttgart,Betriebswirtschaftliches

Institut.

Kleijnen, J., 1995. Theory on Methodology: Verification and validation of simulation models.

European Journal of Operational Research , Volume 82, pp. 145-162.

Kleijnen, J., 2005. Supply chain simulation tools and techniques: a survey. International Journal of

Simulation & Process Modelling, 1(Nos. ½).

Krajewski, L., Ritzman, L. & Malhotra, M., 2007. Operations management: Process chain and value

chains. New Jersey: Pearson Education..

La Londe, B. J., Masters & M., J., 1994. Emerging Logistics Strategies: Blueprints for the next century..

International Journal of Physical Distribution and Logistics Management 24(7), pp. 35-47.

Lambert, D., Cooper, M. & Pagh, J., 1998. Supply Chain Management: Implementation Issues and

Research Opportunities. The International Journal of Logistics Management, Vol. 9 Iss: 2, pp. 1-20.

Lau, Xie & Zhao, 2008. Effects of inventory policy on supply chain performance:. Computers &

Industrial Engineering , pp. 620-633.

Law, A. & Kelton, W., 1991. Simulation Modeling and Analysis. 2nd red. New York: McGraw-Hill, Inc..

Lee, H., Padmanabhan, V. & Whang, S., 1997. The Bullwhip Effect in Supply Chains. Sloan

Management Review, 38(3)(Spring ), pp. 93-102.

Liu, Fang & Song, J.-S., 2012. Good and bad news about the (S, T) policy.. Manufacturing and Service

Operations Management, 14(1), pp. 42-49.

Marklund, J., 2011. Inventory control in divergent supply chains with time‐based dispatching and

shipment consolidation.. Naval Research Logistics, 58(1), pp. 59-71.

Mentzer, J. T. et al., 2001. Defining supply chain management. Journal of business logistics 22(2), pp.

1-25.

Olhager, J., 2003. Strategic Positioning of the Order Penetration Point. International Journal of

Production Economics, 85(3), pp. 319-329.

Ostrom, E., 2008. Doing Institutional Analysis Digging Deeper than markets and hierarchies.

HAndbook of New Institutional Economics, pp. 819-848.

Reid, R. & Sanders, N., 2007. Operations Management: an integrated approach. 3rd red. New York:

John Wiley & Sons.

Page 142: Towards the Optimal Inventory Review Intervals

142

Sargent, R., 2007. Verification and Validation of Simultion Models.. Syracuse, N.Y., Proceedings of the

2007 Winter Simulation Conference.

Shahin, A. & M.A, M., 2007. Prioritization of key performance indicators: An integration of analytical

hierarchy process and goal setting. International Journal of Productivity and Performance

Management, 56(3), pp. .226 - 240.

Shang, K., Tao, Z. & Zhou, S., 2015. Optimizing reorder intervals for two-echelon distribution systems

with stochastic demand.. Operations Research, 63(2), pp. 458-475.

Silver, E. A. & Peterson, R., 1985. Decision systems for inventory management and production

planning. New York: John Wiley & Sons Inc..

Simchi-Levi, D., Kaminsky, P. & Simchi-Levi, E., 2003. Managing the Supply Chain: The Definitive Guide

for the Business Professional. New York:: McGraw-Hill..

Smart, P. A., Maddern, H. & Maull, R. S., 2009. Understanding business process management:

Implications for theory and practice. British Journal of Management, 20(4), pp. 491-507.

Stadtler, H. & Kilger, C., 2008. Supply Chain Management and Advanced Planning: Concepts, Models,

Software, and Case Studies. sl:Springer.

Sylver , E., Pyke, D. & Peterson, R., 1998. Inventory management and production planning and

scheduling. New York: Wiley.

Tako, A. & Robinson, S., 2012. The application of discrete event simulation and system dynamics in

the logistics and supply chain context. Decision Support Systems, 52(4), p. 802–815.

Thissen, W., Enserink, B. & van der Lei, T. (. J., 2008. Teaching problem formulation in technology,

policy and management education: A Systems approach.. Portugal, Engineering Management

Conference.

Treacy, M. & Wiersema, F., 1993. Customer intimacy and other value disciplines. Harvard Business

Review, 71(1), pp. 84-93.

Verbraeck, A., 2010. Discrete modellen deel 2 - Discrete simulatie. sl:sn

Vermorel, J., 2014. www.lokad.com. [Online]

Available at: https://www.lokad.com/min-max-inventory-planning-definition

[Geopend 1 10 2015].

Verschuren, P. & Doorewaard, H., 2010. Designing a Research Project. The Hague: Eleven

International Publishing.

Visser, H. & Goor, A., 2004. Werken met Logistiek. Groningen/Houten: Wolters-Noordhoff..

Waters, C., 2003. Inventory Control and Management. 2nd red. Chichester: John Wiley & Sons Ltd..

Webster, F., 1992. The Changing Role of Marketing in the Corporation. Journal of Marketing, 56(4), p.

1–17.

Wild, T., 2002. Best Practice in Inventory Management. 2nd red. Oxford: Butterworth-Heinemann

(imprint of Elsevier)..

Wolstenholme, E. F., 2003. Towards the definition and use of a core set of archetypal structures in

system dynamics.. Syst. Dyn. Rev., 19(7-26), p. doi: 10.1002/sdr.259.

Page 143: Towards the Optimal Inventory Review Intervals

143

Yin, R., 2009. Case Study Research: Design and Methods. Thousand Oaks: Sage Inc.