Eindhoven, April 2013

Increasing the inventory turnover rate in a medium sized electronics assembly company by Bram Berkien

BSc Bram Berkien — INPG 2005 Student identity number 0588147

in partial fulfilment of the requirements for the degree of Master of Science in Operations Management and Logistics

Supervisors: Dr. ir. H.P.G. van Ooijen, TU/e, OPAC Dr. Z. Atan, TU/e, OPAC

TUE. School of Industrial Engineering. Series Master Theses Operations Management and Logistics

Subject headings: inventory, ordering, assembly-line, turnover, micro-electronics, make-to-order, stockkeeping

Preface This report describes the execution of the final project for the master program Operations Management and Logistics (OML) at Eindhoven University of Technology. The project was completed in three stages. The first two stages constituted a review of relevant literature and a research proposal which combined the findings of the literature review with the problem identified by the company where the project was conducted. This final report details the actual execution of the research proposal, describing the process, deliverables and main outcomes.

i

Abstract This report describes the development of an inventory control tool for a medium sized electronics assembly company. The company wished to increase its inventory turnover rate and specifically requested methods that could be implemented in its ERP-system. An initial analysis was conducted which revealed some phenomena and drew a general picture of the firm. Next, an inventory control tool was developed using Microsoft Excel that establishes a direct link with the ERP-system used at the firm in order to provide up-to-date information on inventory and goods movements. The tool offers a detailed analysis of the value and turnover of stocks, which can be used to identify potential problems in their inventory control. Additionally, the tool offers the option to evaluate inventory control strategies using historic data of inventory movements and levels. Implementing and subsequently testing different control strategies revealed that some items benefit from being controlled using safety stocks and reorder points based on service levels, while some specialist items that have long planned lead times should be controlled based on confirmed customer orders and human insight, which is as they have been controlled in the past. Furthermore, aggregating raw components into different groups based on their functionality and other characteristics revealed some further phenomena and appears promising for a further refinement of the suggested control strategies.

ii

Managerial summary This report describes the process of the study performed at the medium-sized electronics assembly company AME with the goal of increasing its inventory turnover rate. Inventory turnover rate is considered a practical measure that indicates how efficiently a firm uses its inventories. If this metric is low, it means that on average items reside in stock for a long time before they are used, or that some items are not used at all. If the metric is high, this generally indicates that items turn over quickly, meaning the cost of the items used in production by far exceeds the average value of inventories present. In general, low inventory turnover rates are not so much a problem themselves, but instead indicate problems may exist elsewhere. For instance, stock ordering strategies may not be optimal, or production inefficiencies may lead to considerable delays causing components and subassemblies to have a very low turnover rate. The firm had previously tried to increase its inventory turnover rate but could not work out the exact cause of its unsatisfactory low value. This study therefore conducted an analysis of stocks present at the firm and consequently used the results from this analysis to pinpoint potential problems and suggest the most feasible approaches to increase the turnover rate of inventories. The main goal identified after this analysis was the construction of a computer system that could analyze recent data on inventories from the ERP-system in use at the firm and project the effects of implementing different inventory control strategies. The analysis identified that considerable dead stocks were present at the firm, while a number of raw components can be viewed as slow moving. Additionally, minimal order quantities resulting from storing electronic components on tapes played an important role. It was suggested that current dead stocks be dealt with and future buildup of dead stocks should be prevented through stricter control of active stocks. It also appeared that safety stock levels for some items were higher than needed to offer high service levels. The implementation of different control strategies was therefore identified as one of the main options to increase inventory turnover. The firm uses an injection molding machine for the in-house production of plastic parts, which is a promising option to avoid the procurement of large shipments from Asia, which turned out to be a major factor in causing items to move slowly. Forecasting demand for raw components was problematic due to relatively short life spans for final products and complicated BOMs. Moreover, management previously did not have an analysis tool that could rapidly show a comprehensible overview of the value and inventory turnover rates for different sets of raw components, allowing it to quickly analyze inventory turnover performance. With regard to suppliers, few changes can and should be made, since the firm already works with a limited set of firms it procures raw components from, and maintains relationships with these. Inventory inaccuracies were suggested in literature as an issue that could possibly cause low inventory turnover, however the firm uses a reliable ERP-system rendering inventory inaccuracies negligible. Redesigning products to avoid

iii

using customer-specific components and instead aim for commonality proved to be difficult, since many designs are highly customized to the wishes of customers and changing components is not always possible. Restricting R&D department employees in their choice of components seems an undesirable option, since the volumes used by this department are small and freely choosing components is part of the design process. Following the conclusions from the analysis of the firm, a computer tool was developed that extracted data from the company ERP system. The tool then provided an overview of stocks present at the firm which allowed for a quick analysis by users. Furthermore, the tool offers the option to test different inventory control strategies and evaluate their expected performance in terms of increasing inventory turnover rates. Applying the tool at the firm showed how inventory value was distributed among different types of components, and also showed that a large share of this value was concentrated in a limited number of different components. Furthermore, it showed how the average inventory value and inventory turnover rate have developed throughout history. Applying this historical analysis more frequently and for smaller time spans could enable the firm to evaluate the results of certain changes in operations it has made. Additionally, some simple inventory control strategies were evaluated. First, conventional (s,nQ) policies were tested. These policies dictate service levels through safety stocks and reorder levels. It was found that in order to offer high service levels, substantial increases in inventories levels would be required, which in turn resulted in substantially decreased inventory turnover rates. This effect was explained by considering a small group of SKUs that were responsible for a considerable share of the projected average inventory value. These items turned out to have enormous lead times for delivery from suppliers and high values per piece. It was assumed that the firm is familiar with the specific characteristics of these items and already controls them in a suitable way. A hybrid inventory control strategy was then developed. The items that were projected to have enormous inventory values if they were to be controlled with the conventional (s,nQ) policy were controlled as the company had done before, while the remainder of the items were controlled using the (s,nQ) policy. This resulted in a saving of approximately eight per cent in projected inventory value. It should be expected that further fine tuning of this strategy, accounting for additional characteristics of groups of components, will result in further savings in inventory value and thus an increase in inventory turnover rates.

iv

Abbreviations AME B2B BI BOM EOQ ERP FTE FGI ITR JIT MRP MTO ODM PCB PN R&D ROP SAP SKU SMD WIP

Applied Micro Electronics Business To Business Business Intelligence Bill Of Materials Economic Order Quantity Enterprise Resource Planning Full-Time Equivalent Finished Goods Inventory Inventory Turnover Rate Just In Time Material Resource Planning Make To Order Original Design Manufacturer Printed Circuit Board Product Number Research & Development Reorder Point ERP-software Stock Keeping Unit Surface Mounted Device Work In Process

v

Contents Introduction

vii

1 Case study environment 2 Research assignment 2.1 Literature review . . . 2.2 Assignment . . . . . . 2.3 Methods . . . . . . . . 2.3.1 Interviews . . . 2.3.2 Data gathering 2.3.3 Control system 2.4 Scope . . . . . . . . . 2.5 Deliverables . . . . . .

1

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

4 4 7 7 8 8 8 9 9

3 Detailed analysis 3.1 Production operations . . . . . . . . 3.1.1 General aspects of production 3.1.2 Injection molding . . . . . . . 3.2 Supply chain . . . . . . . . . . . . . 3.2.1 Lead times . . . . . . . . . . 3.2.2 Suppliers . . . . . . . . . . . 3.2.3 Customers . . . . . . . . . . . 3.2.4 Forecasting . . . . . . . . . . 3.3 Stocks . . . . . . . . . . . . . . . . . 3.3.1 General stock analysis . . . . 3.3.2 Stock control . . . . . . . . . 3.3.3 Dead stock . . . . . . . . . . 3.3.4 Inventory turnover . . . . . . 3.4 Review of analysis . . . . . . . . . . 3.5 Strategy . . . . . . . . . . . . . . . . 3.6 Summary . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

10 10 10 11 12 13 13 14 15 15 15 19 19 21 22 23 25

4 Control tool 4.1 Functional requirements . . . . . . . . . 4.2 Procedures . . . . . . . . . . . . . . . . 4.2.1 Extraction from SAP . . . . . . 4.2.2 Formatting and filtering of data .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

26 26 27 27 29

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

vi

4.2.3 Calculations for individual SKUs 4.2.4 Analysis of stocks . . . . . . . . 4.2.5 Stock control strategies . . . . . Conclusion . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

30 33 37 40

5 Application 5.1 Implementation . . . . . . . . . . . . . . 5.2 Active inventory . . . . . . . . . . . . . 5.2.1 Applying the ABC-classification 5.2.2 Current inventory . . . . . . . . 5.2.3 Historical development . . . . . . 5.3 Dead stock . . . . . . . . . . . . . . . . 5.4 Proposed control strategies . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

41 41 42 42 43 47 47 50

4.3

6 Conclusions & recommendations

55

References

58

Appendices

60

A Kardex system

60

B Extraction of data from SAP

61

C Boundary conditions and filters

63

D Service levels and service factors

64

E ERP and control tool data comparison

65

F Conventional ABC-classification applied to firm

66

G Problematic SKUs

67

vii

Introduction The paper at hand contains the research report for the master thesis project that was conducted at a medium sized electronics assembly firm with the goal of increasing its inventory turnover. Virtually all companies strive to reduce their investment in inventories or generate higher turnover with its current inventories for both financial and operational reasons. Conducting the study at a company allowed for real-world testing of the methods that were devised. As described in the research proposal that preceded this report, the goal of the project is twofold. On an academic level, the project aims to generate new knowledge and procedures to attain the goal of increasing inventory turnover at a firm, by presenting a standard framework that in generally applicable in firms. On a more practical level, the goal is to design procedures and tools that can be directly implemented in the project environment to increase inventory turnover. This twofold nature of the project offers the perfect chance to put the findings of the study to the test. The report starts with a chapter to describe the environment where the study was conducted. It details some of its general characteristics and explains what makes the firm different from other organizations. Next, the assignment as it was undertaken is described. Attention is devoted to the literature review that was conducted in an earlier phase of the project, to the methods that were used in the main study, to the scope of the project and to the main deliverables. Consequently, a chapter is devoted to describing the detailed analysis performed of the company. This includes some early insights gathered from interviews with employees and a broad data analysis. The detailed analysis formed the basis which provided further directions to the construction of an inventory control system. Subsequently, the control mechanism that was developed is described in detail. The mechanism consists of an advanced connection between the information system used by the company, and an elaborate Excel macro. Finally, the practical use, consequences and potential improvements that can be gained by using the mechanism are described and a conclusion and recommendations for future research are presented.

viii

Chapter 1

Case study environment The project described in this report was conducted at Applied Micro Electronics ”AME” B.V. (AME), located in Eindhoven, the Netherlands. This chapter will describe the company in general terms. Later in this report, data from the ERP-system in place at the firm will be analyzed to gain a detailed perspective of the operations conducted within the company. This present chapter will focus on readily available and static data that helps to form a general picture of the company. AME is a medium sized company located in Eindhoven (The Netherlands) that provides solutions within the disciplines of research, design, development and manufacturing of electronics, electronic systems and embedded systems acting as an original design manufacturer (ODM). Currently all operations are based in one location. While its main operational activities consist of assembling electronic products, AME aims to add value to customer relationships using its ODM capabilities. Generally, its products consist of a printed circuit board (PCB) combined with a number of electronic components called surface mounted devices (SMDs), usually contained within a plastic housing. These assembled units are then incorporated in the final product manufactured by the customers of AME, which is then shipped to consumers. Some examples of final products include smart energy consumption tracking units, control panels for coffee machines and electric modules for e-bikes. In order to understand the general size and performance of the firm, its business intelligence (BI) system was used to obtain general and readily available figures. These include data on turnover, profit, total stock value and human factors. Table 1.1 shows financial figures for the company in 2011 and the expected figures for 2012 - all figures for 2012 had not yet been compiled and confirmed at the time this report was written.

Turnover R&D Turnover Products Total turnover

2011 e1,602,572 e8,989,192 e10,591,764

2012 e1,919,602 e11,354,339 e13,273,941

Table 1.1: Financial figures for AME

1

Currently, the firm employs just 32 staff members (FTE) in its production department, in addition to eight interim workers, while just under 40 are employed in its R&D department. The BI system indicates that the current stock value lies at around 2.2 million euro, while dead stock accounts for 200,000 euro. The dead stock value is the total value of all materials on stock which haven’t had a relevant movement for over a year or never at all. A relevant movement is defined as all movement types involving production orders. AME places a large emphasis on high quality, which is achieved through high automation, testability and manufacturability. The company operates across different market segments and has managed to expand its business with an average growth of 34% on an annual basis. Although competition is present the industry AME operates in, this industry is of a considerable size, allowing the firm to grow in the manner it has done. Additionally, AME aims to add value with its design of custom products and by maintaining long term relationships with customers. The company consists of two main units, which are an R&D unit and an operations unit. The R&D unit focuses on the electronic and mechanical design for customer orders, while the operations unit is occupied with the main manufacturing activities and logistics. In addition, a supporting unit handles administration, finances en information technology. The operations unit recognized that the company was not using its inventory as efficiently as it should. This efficiency is tracked by determining the so-called inventory turnover ratio, which is a dimensionless metric and calculated by dividing the cost of goods sold in a certain period over the average value of inventory. Specifically, the company wishes to increase its average inventory turnover ratio from its current level of two, to at least four but preferably six. This project originated from the desire of that unit to start using its inventory more efficiently and increase the turnover rate. The inventory turnover ratio will be further discussed in chapter 3. Orders come in exclusively from other businesses (B2B environment). Final products assembled are unique to each customer, that is, no two customers order the same product, since products are designed exclusively for that customer. Virtually all production occurs on an assembleto-order basis, meaning any finished goods will only reside within the firm for a short time and typically already be ordered by a customer. Some exceptions exist however, for customers that require a shorter lead time for their orders than could be realized otherwise. In general, the lead time quoted by suppliers for raw materials is longer than the lead time required by customers for their finished products, resulting in the need to keep a substantial raw materials inventory and the forecasting of customer demand. Ordering from suppliers occurs automatically using BOMs and forecasts for future customer demand, based on advance information shared by customers. For, critical components, new components and some other SKUs, staff handles the forecasting and ordering process. In these cases automatic ordering using MRP systems was deemed inappropriate. The production floor at AME consists of two production lines where orders are assembled. Components are stored in trays in an automated warehouse. A large percentage of these components are shipped and handled as so-called tape and reel, a standard storage method in the electronics industry that allows for the packaging of a large number (thousands) of small electronic components. An advanced ERP system (SAP) is used, additionally the firm has an efficient business information system installed. This system records the performance of several business processes, such as production operations, forecasting and supplier deliveries.

2

This study focusses on increasing inventory turnover, which was identified as an important goal by management. AME aims to be a lean and efficient organization, which means inventory should turn around relatively quickly.

3

Chapter 2

Research assignment As mentioned in the previous chapter, AME maintains a relatively large component inventory, realizing an inventory turnover rate that is significantly lower than desired. At present, this situation does not form an immediate threat to daily operations, however the extensive inventory does occupy space in the storage warehouse and most importantly it ties up capital, decreasing financial liquidity. This is a problem that is encountered in many organizations. Based on the literature review, it appears that a large gap exists between scientific research and practical studies, while an abundance of different approaches to inventory reduction are available. However, because of the excessive amount of options, it can be hard to select the best approach for a given environment. The project environment presents an excellent opportunity to test which of the approaches identified in the previous chapter is most suitable for a medium sized production company. This chapter first presents a short summary of the literature review conducted earlier. Next, the assignment that will be executed is detailed and subsequently the methods that will be used are shortly listed. Finally, the scope and deliverables of the project are stated.

2.1

Literature review

A literature review was conducted in order to identify what research has previously been done into the subject of inventory reduction (in order to increase inventory turnover rates). The main finding of the review was that two major gaps exist in currently available literature. First of all, it was found that texts from a wide array of topics within operations research in firms have generated knowledge directly or indirectly applicable to the reduction of inventories. The approaches to inventory reduction from these respective fields have hardly been combined to yield a cohesive framework. Linking together every single approach will be impossible, but clearly it would be interesting to see how multiple approaches perform when combined. Effects may be amplified, or possibly negate each other. The second major gap found was between scientific, scholarly research and management papers. This is a traditional gap, in that managers in practice have always wondered how to apply the formal and abstract theories and methods of scholars, while scholars feel reality is too complicated to be modeled in one cohesive framework and therefore are required to make certain abstractions 4

and assumptions to be able to yield any quantitative methods. Translating the many and diverse scientific procedures to a framework that combines multiple approaches from several different fields of operations research and is feasible and understandable to managers in real-world firms is therefore identified as the most interesting goal for future research in this area. In addition to these gaps, the review yielded some general insights on the theme of inventory reduction. Concerning the most probable causes for excess inventories, the review identified three major explanations in literature. First of all, overstocking may have been a deliberate choice as a result of certain opportunities such as temporary or quantity discounts. Second, the external or internal demand rate may have seen a sudden decline. Third and foremost, ineffective and inefficient procedures may be in place. Consequently, the rationale for inventory reduction as described in literature was identified. Though some studies argued that inventory reductions may not always be beneficial, most studies agreed that keeping excess inventory is a bad thing. The reasons they stated were that it ties up company funds, it introduces stock holding costs, it can mask problems in operations and reduces the flexibility of production. It should be noted that literature generally does not cite any representative values for inventory turnover ratios for different industries or types of companies. It appears that this ratio is relatively company-specific and drawing comparisons between several companies is not common. Finally, the approaches several studies have suggested were investigated. The most important and promising options and areas of improvement in terms of potential inventory reductions and feasibility will be briefly repeated here. • Firms can reduce the amount of funds invested in inventory by reducing the value of those inventories, for instance by keeping raw materials on stock instead of finished goods (Primrose (1992)). • Safety stocks can be reduced by using better forecasts, reducing supplier lead time, changing batch quantity or decreasing service levels (Gips (1998), Ploss and Wight (1967), Ritchken and Sankar (1984)). Naturally, reducing service levels may not be a desirable course of action. • Inventory inaccuracies should be minimized through careful stock keeping (Gips (1998), Tersine and Tersine (1990), Harrington, Lambert, and Vance (1990), Ernst, Guerrero, and Roshwalb (1993)). • Supplier selection and management deserves special attention (Gips (1998), Tersine and Tersine (1990), Forslund and Jonsson (2007), Zhao and Lau (1992), Holstr¨om et al. (2002), Cachon and Fisher (2000), Jahnukainen and Lahti (1999), Goffin, Szwejczewski, and New (1997)). Strategies include using fewer suppliers, using two suppliers to procure one item, using nearby suppliers and sharing info with suppliers. • Delivery of supplies should be investigated (Gips (1998), Tersine and Tersine (1990), Finkin (1989), Chadwick and Waddington (1982)). This includes reducing the lead time suppliers quote, obtaining more frequent delivery, increasing the reliability of deliveries in terms of variability of lead time and careful inspection of the quality and quantity of supplies as soon as possible after receiving them. • Firms can opt to introduce or perfect using the JIT philosophy for manufacturing of customer orders and procurement of components (Tersine and Tersine (1990), Kinney and

5

Wempe (2002), Sepehri (1986)). • Several authors advise delayed product differentiation and using commonality in components (Tersine and Tersine (1990), Lee and Tang (1997), Baker, Magazine, and Nuttle (1986). • Forecasting of future customer demand is offered as an option by many studies (Tersine and Tersine (1990), Gips (1998), Gardner Jr (1990), Jacobs and Wagner (1989), Shoesmith and Pinder (2001)). • Special attention can be paid to slow moving items or obsolete and dead stock (Gips (1998), Razi and Tarn (2003), Watson (1987), Croston (1974), Johnston, Boylan, and Shale (2003)). Dead stock should be identified and scrapped. Another option suggested by Sox et al. (1997) is not keeping any stock at all for slow moving parts, giving high priority to their procurement once a demand occurs • Tersine and Tersine (1990) suggests returning unused (obsolete or extremely low demand) components to suppliers. • A phenomenon not often discussed is MRP nervousness (Rantala and Hilmola (2005)). This can lead to both under stocking and over stocking and results from the sensitivity of MRP systems to small changes in certain parameters. It can be countered by fixing the production schedule for a longer time, increasing the forecasting horizon and applying lot-for-lot production. • Careful selection and evaluation of the inventory control policy that is used. (Song (2000), Jahnukainen and Lahti (1999), Gips (1998), Williams and Tokar (2008), Zhao and SimchiLevi (2006)). This includes evaluating if the inventory policy that is used is actually suitable for the specific environment, whether applying as assemble-to-order or make-toorder strategy may be used and whether an (R,nQ) policy can be applied. These options can all contribute to inventory reductions. • Inventory consignment concerns not having to pay suppliers for their deliveries until after these have actually been used in production or sold to customers (Wallin, Rungtusanatham, and Rabinovich (2006), Corbett (2001)). Effectively, physical inventories will not be reduced this way, but company funds are freed up. Naturally, suppliers may not be too eager to participate in such a method since it means delayed payments for them. • Some studies emphasized dealing with general inefficiencies and inaccuracies in operations (Tersine and Tersine (1990)). Such defects can include inaccurate bills-of-materials and an excessive amount of rejects at quality control and customer returns. Finally, Chadwick and Waddington (1982) suggests that fear of under stocking with managers may play a role in inventory accumulation. While the focus of this project is not on the psychological aspects of inventory reduction, one can imagine that using the proper and reliable methods and tools may aid managers in taking the right decisions.

6

2.2

Assignment

Based on the gaps identified in literature and using the opportunity offered by the challenge faced by the firm, the main research goal can be formulated: Considering the logistic approach of a medium sized high-tech assembly company on a supply chain level (backward and forward) and the logistical and financial parametrization on a material level, investigate how inventory can be managed in a more efficient way, increasing the inventory turnover rate. Or, stating a clear assignment: Develop an inventory control strategy that can be used to increase the inventory turnover rate at a firm with main characteristics similar to those of AME, as described in chapter 1. This assignment is complex due to the large number of different SKUs kept by the firm, and the fact that new products are introduced regularly, making the forecasting of customer demand difficult. Additionally, minimal order quantities play a significant role in the ordering of raw components. The assignment is novel in that it combines several of these characteristics that complicate increasing the efficiency with which inventory is used, and aims to simplify the analysis in order to aid managers. This simplification is done by focussing on developing a decision tool with a clear interface. In addition, a number of sub assignments are defined. • Determine the characteristics of the suppliers used by the firm. • Characterize the component delivery processes. • Investigate customer orders and forecasting. • Determine the characteristics of the inventories kept at the firm. • Determine which of the options for inventory reduction suggested by the review of literature are most appropriate to the project environment. This will be further addressed in section 5.3. • Develop a method to implement the most feasible options at the firm. • Compare the results of these methods and determine which methods are most effective in a setting as described in the chapter on the case study environment.

2.3

Methods

After discussing the main objectives and previous literature in the previous sections, the present section will present the methods to be used. In summary, these methods include the extraction of data sets from the company information system SAP and subsequently analyzing it, interviewing company employees and constructing an extensive system using Microsoft Excel macros that will allow for an up to date and cohesive analysis of stocks, in addition to proposing control strategies and subsequently testing them. The project will start with an initial analysis of the firm and interviewing key personnel. Subsequently, a control tool will be developed that will aid in achieving the objective of increasing inventory turnover.

7

2.3.1

Interviews

Several employees will be interviewed in order to gain some perspective of the most important operations at the firm. Naturally, these interviews will not serve as a source of hard data, but they provided some useful directions for further investigation, and can also be expected to highlight certain constraints, limitations and directions that would definitely not be worth pursuing. Among the staff members to be interviewed are the warehouse manager, the main procurement officer, a staff member involved in client contact and the staff member responsible for introducing injection molding at the firm. The main outcomes of these interviews will be considered in the initial analysis in chapter six.

2.3.2

Data gathering

The firm operates a well-known enterprise resource planning system (SAP). This system is widely integrated in the operations of the organisation and allows for easy access to virtually all needed data. These can be accessed through an interface using a set of commands. Data can subsequently be exported to a spreadsheet. Alternatively, data can be directly and automatically imported to Microsoft Excel through the use of an advanced macro. Initially, the more conventional method of manually exporting from SAP will be used to perform an initial analysis of the firm. Later in the project, the control tool will be augmented with automatic exporting from SAP to Microsoft Excel.

2.3.3

Control system

The main portion of the work on the project will be dedicated to constructing the control tool that AME can use to actively monitor its stocks. The tool should allow the firm to identify which individual components and groups of components are not turning around as quickly as most of the stock and therefore deserve special attention. In addition, the tool should allow for an analysis of dead stocks. Most importantly however, it should include functionality to determine and test several different control strategies for the inventory kept at the firm. These strategies can be aimed at specific low performers in terms of inventory turnover, or at high value components. The control tool should show the potential benefits and drawbacks of each strategy. This functionality will allow decision makers at the firm to select the most suitable strategy based on its perceived value and potential drawbacks. Specifically, the strategies should be aimed at increasing the inventory turnover rate through the reduction of the average raw component inventory value. This can be accomplished by fine-tuning safety stocks and changing the control policy for individual raw components, or for groups of components that share certain characteristics. Strategies can be tested by applying them on historic demand data for the past year. Where this control tool could genuinely shine is in its direct link to the company ERP-system. This coupling would make the tool up-to-date and relevant at all times and enable for an evaluation of changes in control policies, in addition to providing the option of reviewing performance in periods further back in time, due to the extensive data logging of the ERP-system.

8

2.4

Scope

The scope of this assignment will be limited to considering just the firm itself, instead of attempting to optimize a multi echelon system. Since a number of different approaches will be tried, and each approach concerns several distinct areas of inventory management, only the company itself can be considered. At this point in time, it may make sense to consider all items within the company, since no data is available yet. However, this may prove to be an elaborate task. Therefore, first of all, the items that possibly contribute most to the problem should be isolated. These items can logically be expected to be the slowest moving items within the company stocks. Subsequently, further analysis will be directed towards these items.

2.5

Deliverables

In terms of deliverables, several different items are to be created. Naturally, this current report was one of them. Furthermore, a concrete decision support tool is to be constructed that implements the methods selected. This will help the validation of the methods and to test the implementability of them. The tool will also serve as a useful product for the firm where the research is conducted. Specifically, the tool will allow for the parametrization of important decision variables in the ERP system operated by the company, which aids in improving inventory turnover rate.

9

Chapter 3

Detailed analysis Before the development of an inventory control system could be started, a detailed analysis of the company inventory and processes was conducted. This analysis involved talking with company employees who had tasks that were relevant to inventory turnover and utilizing the ERP-system implemented at the firm to gather data. Using these two methods, a good overview of operations at the firm was acquired, that allowed for deciding on an exact strategy to be used to increase inventory turnover. This chapter will consider production operations, the main suppliers used, the most important customers of the firm and characterize the stocks held. The findings of these sections are then summarized and finally, based on this initial analysis, the most viable strategies to increase inventory turnover will be considered. Additionally, potential strategies that do not seem promising given the specific characteristics of the company will also be discussed. The outcomes of individual interviews with company employees will not be discussed separately but have instead been used to determine which data to gather and which areas to focus on.

3.1

Production operations

This section describes production operations at AME, both in a general sense and by addressing some specific aspects that are unique to this particular firm.

3.1.1

General aspects of production

AME currently operates one production location, which it expects to retain in the near future. It has grown from one production line to two and will add a third one by 2015. The same holds for the number of assembly areas it operates. Since 2006, the firm has tripled its number of unique types of end-products and projects this number to increase by another 100 per cent up to 2015. In total, the firm shipped around 700,000 finished products in 2012, compared to 30,000 in 2006 and a projected number of two million by 2015. Production runs generally take well under a day, once the run has been completed all finished items and raw components are removed from the production floor. Therefore, WIP-levels can be assumed to be low and negligible for the analysis of inventory turnover. In general, few issues occur in production. 10

Nearly all production at AME is started based on customer orders, whether these are confirmed by customers - based on actual production schedules or on forecasts - or on human insight at AME. That is, oftentimes, managers know based on past periods how to estimate the demand by a specific customer for a specific product. For these reasons, production operations at AME should be described as make-to-order (MTO). Even if no actual order for a production batch was placed, the items will generally reside only shortly in the finished goods inventory (FGI) before being consumed by an incoming customer order. Another factor that plays a huge role in the nature of the production process is that all finished products are unique to one customer. In other words, two customers do not order the same finished product. The firm attaches a large amount of value to customer service in terms of delivery performance and therefore wishes to minimize production shortages. Thus far this has translated into significant safety stocks of raw materials, which will be investigated later. For now however, it will be interesting to see whether or not production shortages are indeed rare. The business intelligence report for material shortages shows only a minor number of occurrences that date as far back as 2010. An important aspect is that in addition to producing products, mostly through assembly, AME designs products for customers. Firms often turn to AME to draw up technical designs of products and develop and test prototypes. This means that up to a certain level AME determines which components will be used in final products. Its R&D department makes technical drawings of products, specifying which components are placed where in the design. In most cases, AME subsequently manufactures the product. Naturally, customers will frequently have extensive design requirements, necessitating the use of specific components. Essentially then, AME designs the product and determines which parts are to be used, but oftentimes certain restrictions are imposed by the design demands made by the customer. A phenomenon inherent with the design of new products is the process of iteration. Designing the technical layout of a product is a process that is executed step by step, with requirements and designs changing along the way. The placement of components is not fixed until production has actually started and the number and even type of components used can change up to a very late stage. However, in order to be able to meet customer lead times, at a certain point the procurement of the required components will need to be started. The danger associated with this mechanism is that occasionally, parts will be ordered that are rendered obsolete because the BOM for the new product is changed in a late stage. Although ordering components that will not be used in production is a regrettable practice, this is part of the process of designing of new products.

3.1.2

Injection molding

For some time now, the firm has been using an injection molding machine to manufacture plastic parts. The specific machine present at AME can manufacture parts that weigh between 25 and 150 grams within the dimensions of 30 by 30 centimeters. These parts generally cost anywhere between a few cents to a few euros and were previously ordered from external suppliers, which were mainly located in Asia. This substantial distance meant components were shipped by sea, resulting in a considerable lead time and large minimal ordering quantities. In practice this resulted in numerous pallets of plastic parts standing around idle in the warehouse for a long time until they were used in production. Naturally, this situation was detrimental to inventory turnover.

11

Although the firm needs some more time to become totally familiar with designing molds and scheduling production on the machine, using it for in-house production of plastic components offers some substantial advantages over out sourcing their production to Asia. First of all, these components are usually relatively expensive compared to surface mounted devices (SMDs) that are stored on tape & reel. This makes it beneficial to be able to choose production quantities instead of being limited to minimal order quantities imposed by external suppliers. Due to a low internal lead time for components produced using the injection molding machine, components can actually be produced on a make-to-order basis instead of ordering to stock as in the case with out sourcing production. Naturally, certain lead times and minimal economically justifiable production quantities still apply, however these are smaller than for externally produced components. Furthermore, producing in-house comes with the additional benefits of allowing for substantially better quality control and independence from external suppliers and uncontrollable factors such as shipment issues. Production using the injection molding machine requires a raw resource named granular plastic. Several different types of granular plastic exist, however generally a limited set of types is used in production. This means the same type of plastic can be used for different manufactured components. This offers substantial benefits in controlling inventory of the granular plastic. Another advantage of this raw inventory source is that it is considerably more compact than the products it is manufactured into, which makes for easy and efficient storage. In order to determine the potential benefits in terms of more efficient inventory control the injection molding machine could provide, a company employee familiar with its characteristics was interviewed. This yielded several important figures. First of all, the lead time for components manufactured on the injection molding machine is composed as follows: • Four weeks are needed to design the mold that will be used. This mold is specifically designed for the single component it will produce. Molds are relatively expensive objects, but have to be designed and produced regardless of whether or not the component is manufactured in-house. Molds generally cannot be re-used for different components. • After designing the mold, an additional three weeks are needed to produce the mold. • Once the mold has been designed and produced, on average a production run will take two to three weeks to be scheduled and executed. This period of two to three weeks can be considered as the actual lead time for a component that has been in production before. The period needed to design and manufacture the mold will coincide with the overall designing of the final product. It should be noted that the production schedule for the machine is currently generally determined for up to six weeks ahead. However, at present no clear procedure is in place to regulate the scheduling of production runs. Therefore, for the sake of both economic considerations and optimizing inventory control, future research should be performed to develop a scheduling strategy for the injection molding machine.

3.2

Supply chain

This section describes the supply chain AME operates in. Attention will be devoted to both suppliers and customers, to lead times and to forecasting. AME aims to optimize the supply chain it operates in by communicating extensively with both suppliers and customers. It utilizes

12

a supplier evaluation system that allows the firm to determine how reliable a supplier is. This system places a penalty both on delivering late and delivering early.

3.2.1

Lead times

A challenging aspect of the supply chain AME is positioned in is that the average component lead time is around 16 weeks, while customers demand a lead time of 3 weeks for their orders. Naturally, after having received procured components the firm also needs some production and transportation time which account for another two weeks of lead time on average. This explains why stocks have to be held in the warehouse and some form of forecasting and safety stocks are in place. With respect to supplier lead times, an interesting pattern is visible. Figure 3.1 shows the frequency with which a planned lead time (based on supplier predictions) occurs for all SKUs.

Figure 3.1: Count of lead time Interestingly, major peaks are visible at seven days, fourteen days and 56 days. Clearly most suppliers quote lead times as an integer number of weeks, with a limited number taking more than two months to ship components.

3.2.2

Suppliers

AME operates using a limited number of suppliers. It is Company policy to order from known suppliers if possible, mostly for administrative and relationship reasons. Ordering from a limited set of suppliers makes it easier to arrange deals and limits administrative work. Naturally, at times components can not be ordered from this small group of suppliers and need to be procured elsewhere. Table 3.1 shows the most important suppliers for AME in terms of delivered value in the past six months. Ordering from these suppliers occurs in two different manners. For about 60 per cent of deliveries, forecasts are constructed automatically and forwarded to suppliers. These forecasts are purely based on material requirements, taking into account planned production runs and planned lead times. For other items, manual orders are placed. These items could be necessary for new

13

Vendor Name Supplier A Supplier B Supplier C Supplier D Supplier E Supplier F Supplier G Supplier H Supplier I Supplier J

Delivered Value e766,755 e389,587 e352,028 e273,226 e242,037 e214,108 e182,086 e168,630 e139,727 e112,859

Share 17.3% 8.8% 7.9% 6.2% 5.5% 4.8% 4.1% 3.8% 3.1% 2.5%

Table 3.1: Vendors used by AME product introductions, for production runs that experienced issues or in the case of changed material requirements.

3.2.3

Customers

Similar to the situation with its suppliers, AME works with a fairly limited group of customers who buy a large portion of its products. Table 3.2 shows the order share for the top customers who order from AME. Customer Customer Customer Customer Customer Customer Customer Customer

A B C D E F G

Share 13.2% 12.1% 9.1% 8.4% 6.9% 6.5% 6.5%

Table 3.2: Customer orders share Clearly, the firm works with a limited set of customers - just over fifty -, where an even smaller subset of these account for a large portion of the total order value. As mentioned before, AME places a high value on delivery performance. It utilizes a scoring system that takes into account whether orders were delivered on time, late or early. A certain delivery window is defined, meaning orders can be delivered early or late by a limited amount of days. If orders are delivered outside of this window, that is early or late by too many days, a penalty is incurred. Naturally, a larger penalty is placed on tardiness. Table 3.3 shows the delivery performance for the past four months. From this table it is obvious that the company is able to offer its customers close to perfect service.

14

Month March 2013 February 2013 January 2013 December 2012

Delivery score 100% 100% 100% 99%

Table 3.3: Delivery performance of AME to its customers

3.2.4

Forecasting

Forecasts are normally based on customer projections, general industry and economic trends and human insight. Although the firm keeps track of forecasting accuracy up to some degree by registering the forecast and actually delivered number of finished products, these data are hardly usable for the present study. First of all, although it is possible to translate these forecasts to demand for raw components from the definition of the BOM for each of the individual finished products, this procedure is extremely cumbersome. Additionally, the nature of ordering at AME makes forecasting a complicated procedure. Customers often only order a finished product for a limited period of time, typically in the range of one to six months. This makes it hard to detect an actual pattern in demand through the use of traditional trend analysis algorithms. AME applies a combination of customer supplied forecasts and human insight to construct predictions for demand for finished products. Components are procured partly based on these forecasts especially in the case of customer specific raw components - and partly using safety stock and reorder points - for components that are used in multiple finished products.

3.3

Stocks

Stocks play an extremely important role for AME. The firm aims to offer perfect service to its customers in terms of both quality and delivery performance. For this reason, significant stocks are kept at the company premises. This section will present a general stock analysis, followed by an analysis of dead stocks.

3.3.1

General stock analysis

This subsection will describe the several aspects associated with inventories kept at AME. First of all, storage types, space utilization and coding systems will be described. After these frameworks have been detailed, the stock itself can be evaluated in terms of value. Storage types A major distinction in the present inventory can be made by looking at the manner in which a component is stored. In the electronics assembly industry, it is extremely common to place small electronics devices such as resistors on so-called tape&reel, which is a cylindrical device that can contain thousands of these parts. The advantages of storing components in this manner are that it allows for easy processing on specialized machinery, and storage becomes easier as well. An obvious disadvantage is that these components can often only be procured in large quantities. An alternative to storing these components on tapes is to store them on so-called 15

tubes or sticks. However, these devices can only carry a limited number of components, and significantly increase the processing time required. Human operators will have to change carriers on processing equipment more frequently. An additional aspect of storing raw components on tapes is that for some production runs one type of component is required multiple times at the same time. In this case, often, multiple tapes carrying the same component are loaded into the machine to allow for more efficient and faster production. This entails that safety stocks for some components will actually need to be higher than required by conventional optimization procedures. If safety stocks and reorder levels were determined without taking the just mentioned mechanism into account, it is possible that the number of components required could fit onto one tape. However, since the component could be needed in production multiple times at the same instance, multiple tapes carrying the same component are required. Dividing the number of components that should be stocked according to safety stock calculations among multiple tapes is no option, since the number of components one tape contains is fixed by suppliers. Most other components used in assembly are stored in a more conventional manner. These components include casings, small motors and printed circuit boards. AME currently stocks just under three thousand different raw components, accounting to a total number of 50 million components used in production annually. This number is expected to have tripled by 2015. Space Besides the most obvious reasons for reducing inventory investments - to avoid tying up funds and to make operations more efficient - another reason could be to free up storage space. AME is a company that grows rapidly as can be concluded from the figures thus far. In order to investigate the relevance of the aforementioned advantage, consider table 3.4, which shows the utilization of storage space in the automated storage warehouse present in the firm. This automated warehouse in a Kardex system, an example of which can be seen in appendix A. This warehousing systems works with vertical modules which contain trays. These trays can contain any components up to a certain size, in any form of storage - both tapes and conventionally stored items. The numbers displayed in table 3.4 should be considered as slots in these trays. Please note that in addition to storing components in this automated unit, the company also keeps components in a number of other storage locations, such as dry cabinets and on conventional pallets. Typ 10 20 921 922

Storage type name Rollenmagazijn Rollenmagazijn Stock transfers (StLoc) Posting change area

Occupied 2,858 4,405 1 0

Empty 722 676 0 1

Utilization 79.83% 86.7% 100% 0%

Table 3.4: Space utilization From this table it becomes obvious that although there is some space available in both Rollenmagazijn storage locations (automatic storage facility), when looking forward into the future this capacity will most certainly not be sufficient. It should be noted that at the time of writing this report, the firm has expanded its warehouse to a vacant neighboring building. However, given the projected future growth, utilization of storage space will eventually become an issue.

16

Coding systems AME utilizes several coding systems within its ERP-system to indicate what kind of component a certain SKU is, and where it is kept. First of all, SKUs are identified by so-called product numbers, or PNs. These are in the format xxxx-xxxx-xxxx, where each x represents an Arabic number. The first four digits are used to indicate the type of this particular SKU. Table 3.5 lists the options for these types. From 1000 2000 3000 4000 5000 6000 6900 7000 8000 8100 8200 9000 9100 9999

To 1999 2999 3999 4999 5999 6899 6999 7999 8099 8199 8999 9099 9998 9999

Type of SKU Passive Discrete Analog Digital Electromechanical Sub Assemblies Purchased Sub Assemblies Finished Products Packaging Material Computers & Accessory Consumable Goods Office supplies Assembly tools Remaining

Table 3.5: SKU groups Table 3.6 shows some examples of these part numbers, along with a description of the part. Part number 1130-0007-1200 4500-0001-0003 6190-0401-0105 6318-1100-0000

Description 1W 5% Thick Film 2512 120E Single Chip Dual USB UART/FIFO, LQFP-48 ZPC Sleutel 500W EC Motor Controller

Table 3.6: Examples of part numbers Furthermore, several locations exist within the firm warehouse, to indicate in what kind of storage a component is currently kept. Table 3.7 lists the most relevant options. Several additional locations exist, however for the scope of this study these are not relevant. Additionally, in order to identify the importance of each of these locations, figure 3.2 depicts the frequency with which the most important transfers from one location to the other occur. If a particular type of transfer is not depicted in the figure it does not occur a significant amount of times on an annual basis. In this case, the transfer does not contribute to relevant streams that should be considered in the analysis of inventory turnover rate. Stock value Finally, after introducing the systems used to classify inventories, its actual current value will be considered. Although the stocks kept at the firm will change with time, inspecting them 17

Code 10 20 100 901 902 911 914 916 921 922 998 999

Location Non-tape warehouse stocks Tape warehouse stocks Production Goods receival Goods receival Kostenplaats Outgoing Outgoing Stock transfer Stock transfer Taping Stock differences

Table 3.7: Stock locations Figure 3.2: Frequency of most important transactions

at a fixed moment in time can draw a fairly clear picture and provide guidelines for further research. A list of all stock present at the firm was extracted from the company ERP-system (SAP). From this list, all SKUs that had a value of zero listed were removed. These are mostly SKU-numbers that are not in use any more. What remained were 9195 different SKUs. For these SKUs, several analyses were executed. First of all, several groups of SKUs were identified based on the value of each component. For these groups, the number of SKUs and the value per group are listed in table 3.8. This table clearly shows that a large amount of SKUs are relatively inexpensive, but these groups represent a large amount of value. Furthermore it should be noted that for some SKUs, no inventory is kept at the firm currently due to stock outs or the temporary absence of demand for the raw component in production runs if it is procured to order, which explains the relatively low value present for some of the more expensive groups.

18

Value per piece (euro) 0 to 0.1 0.1 to 0.5 0.5 to 1 1 to 2 2 to 5 5 to 10 10 to 20 20 to 50 50 to 100 100 to 500 500 to 1000 More than 100

Number of SKUs 2,210 1,417 1,149 554 662 536 468 567 316 513 90 57

Value (euro) e195,826 e488,228 e210,804 e393,327 e261,076 e210,664 e156,334 e204,540 e63,467 e147,646 e14,804 e20,339

Number of components 15,439,558 2,280,809 303,798 282,679 85,577 31,495 11,462 8,066 956 943 25 6

Table 3.8: Value of SKUs

3.3.2

Stock control

The firm utilizes different control policies for different groups of items, depending on whether the items are customer specific and based on their annual demand and value, among other things. For most items, the firm uses either a conventional (s,nQ) policy or uses human insight when ordering. If an (s,nQ) policy is used, this means that once the stock level for a SKU drops to a certain level - called the reorder level - an integer multiple of Q items is ordered. This implies that for most items, pack sizes are determined by suppliers. For instance, some items are delivered in case packs, while - as mentioned earlier - most surface mounted devices are stored on tapes. Furthermore, suppliers often define minimal order quantities. This means suppliers will not deliver less than a specified amount of components in one order. Finally, stocks are continuously monitored, facilitated by the inventory tracking system used at the company. For some materials stored on tape, an interesting phenomenon plays a role. For some production runs, multiple tapes are required in order to allow production to start, since the same component is used multiple times in the design. Switching one tape to be used multiple times on the same product is not an option here. Therefore, even though one tape may hold enough components in total to produce a run, multiple tapes are required in order to allow production to proceed efficiently. This sometimes causes the reorder level for SKUs to be unnaturally high. When using human insight, decisions about when and how many items to order are based on the experience of managers, who know about developments in industry and the general economy, agreements with suppliers and specific mechanisms in production and ordering.

3.3.3

Dead stock

Dead stocks have been identified in the literature review as detrimental to inventory turnover figures. Stocks that have not been used in production for a full year are generally considered to be dead stock. At AME, a slightly different definition of dead stock is used, for financial accounting reasons. Every three months, stocks are analyzed to determine which components should be labeled as dead stock and processed accordingly, usually implying the components are removed from the warehouse and stored in a separate location. The frequency with which stocks are identified as dead stocks will be increased to once per month, the company is currently

19

working on making this change happen. Instead of considering all stocks that have been used in production in the last year as active stock, and all stocks that have not been used as dead stock, AME uses a more gradual definition to determine whether or not a component should be considered as such, which is visible in table 3.9. The firm uses this method in order to recognize that as the time that has passed since a component was last used in production increases, the chances of the component being used in production again should be considered smaller. The risk of the component becoming obsolete increases gradually, which is acknowledged with this detailed definition of dead stocks. Days since last used 0 to 90 91 to 180 181 to 365 Over 365

Discount accounting value by 0% 50% 75% 100%

Table 3.9: Discounting of stock This definition is fairly strict and will indeed allow the firm to keep a close eye on slow moving components. It should be noted that in literature, the definition used by the firm is not common. It appears that the definition used by the firm is mostly a financial one, and is not relevant to operations. Figure 3.3 shows the stock value present in each of these groups at AME. Figure 3.3: Last move date of raw components

Additionally, table 3.10 shows a comparable picture, specific to dead stock components. Moreover, table 3.11 shows the value of dead stocks represented by both tape and non-tape components. For the definition of these types and locations, please refer to table 3.7. It should be noted that in this overview, if a component does not reside in the storage warehouse, it is not identified by its storage type but instead by its location. This is the way the system implemented at AME, while it would have been more convenient to still identify these components by their type instead of their location. This table illustrates that nearly all dead stocks are of the non-tape type (010). It should be 20

Year last moved 2007 2008 2009 2010 2011 2012 Never Total

Total value e313 e25,288 e64,286 e20,379 e70,902 e33,061 e73,333 e287,566

Table 3.10: Total value of components last moved in year Type 010 020 100 901 902 911 914 Total

Number of SKUs 1274 50 9 1 18 7 1 1360

Total value in group e274,158 e6,307 e2,501 e33 e2,213 e2,225 e126 e287,566

Average value per SKU e215 e126 e278 e33 e123 e318 e126 e211

Table 3.11: Dead stock per type further investigated which items represent the largest share of these dead stocks, and thereby determine which items are more prone to be slow moving SKUs. This analysis - which identifies characteristics dead stock components have in common - should be included in the design of the control tool that will be developed. Clearly, the firm holds substantial dead stocks. Dead stocks represent about ten per cent of total inventory, and although the value of these stocks is depreciated in a financial sense, they still represent a former investment in inventory. Therefore, they have a significant influence on the inventory turnover rate. On a more general level, clearly, investments in inventory have been made that possibly should not have been conducted. A large fraction of dead stocks are cheap components, which are generally components stored on tape. Another large fraction of these stocks are expensive components, such as PCBs and rare or specialized components. Talks with company employees revealed that some components have actually been stored in the warehouse for multiple years without having ever been used for production. This indicates that ordering of raw components may be conducted on a fairly loose basis, whereas more strict control is clearly justified. The firm aims to move stocks that have been identified as dead to a separate storage location. Virtually all dead stocks are raw components, meaning that work in process and finished goods inventory are not relevant in the analysis and reduction of dead stocks.

3.3.4

Inventory turnover

In this project, inventory turnover ratio is by far the most important metric considered. It is defined as the ratio of the cost of goods sold to the average value of inventory. Usually, this ratio is calculated over the period of one year. That is, the average inventory and the total cost of goods sold during one whole year are determined. 21

The systems at AME offer two options to determine the inventory turnover rate. First of all, the firm uses the business information system BI, which offers a clear graphical user interface allowing managers to obtain quick summaries of important performance metrics. This portal lists inventory turnover rate in its logistics report. At the time of checking this figure, depending on whether inventory turnover is extrapolated or based on prognosis, the company business information system indicates this figure to be equal to 2.35 or 2.93. Extrapolated inventory turnover is calculated by extrapolating the cost of goods sold for the current year and dividing that by the stock value. The projected inventory turnover is calculated by multiplying the prognosis for the coming 12 months with (1 - 0.48) and dividing that by the current stock value, as defined in the company business intelligence portal. These metrics seem to be rather arbitrary. In addition, SAP maintains a list of inventory turnover rate for all individual SKUs. This list is relatively meaningless without also considering the value per piece of each of these SKUs, and the value currently present of this SKU. The list only displays inventory turnover rate for every SKU without weighing it for the importance of the SKU. Naturally, the list can be used to identify which SKUs definitely do not pose a problem, and which SKUs may deserve special attention. However, the list can not be used to aggregate SKUs into groups, since the information required for this aggregation is lost when the ratio is calculated. It will then need to be extracted from an other SAP table. An other major disadvantage of these figures for inventory turnover rate of individual SKUs is that no clear documentation is available in SAP about how exactly they are calculated. SAP offers some basic explanation about inventory turnover rate but does not list the exact algorithm or discuss exception handling.

3.4

Review of analysis

The previous sections described several specific aspects of the company under investigation. This current section will provide some conclusions and draw a bigger picture of the most important performance aspects. AME is a an electronics assembly firm with an annual turnaround of approximately ten million euro and employs around 70 FTE. Its average stocks are worth two million euro, while its dead stocks account for over 200,000 euro. Ordering occurs based on forecasts of customer orders, and partly based on actual orders if suppliers lead time is sufficiently small. Part of the inventory is controlled using (s,nQ) control policies with safety stocks, while other SKUs are ordered when deemed necessary. In terms of inventory value, the value per component is skewed to the edges. A large part of SKUs is extremely cheap, these are tape & reel components that are ordered in large quantities. At the other end of the spectrum are relatively expensive parts, such as printed circuit boards. This division is visible both in the overall stock and in the dead stock. Nearly all stocks are raw inventory, while in general WIP stock, finished products, service and tooling stocks are negligible. Dead stocks account for about ten per cent of inventory value and a large portion of these dead stocks are significantly older than one year. This could point to a legacy issue, where components have been ordered a number of years ago that still have not been used in production and will most likely not be used in the future. A large portion of the inventory moves slowly, suggesting reorder quantities may be too large. AME delivers its products to a small number of customers that make up a big portion of demand, while the same holds true for its suppliers.

22

3.5

Strategy

Reflecting on the insights presented in the previous sections, this section will discuss the approaches suggested in literature to increase inventory turnover, and how they could be applied in the environment as described. Most importantly, the approaches that appear to be suitable will be highlighted, while the approaches that should not be considered feasible any more given the boundary conditions as described above will also be discussed. The main points that should be deducted from the roundup in the previous section are that the unsatisfactory inventory turnover ratio at AME is caused both by a legacy issue resulting in dead stock and by slow moving raw stock that is not properly controlled. It is clear that dead stock accounts for a significant share of the value of inventory present at the firm. Dealing with dead stock should involve two actions. Firstly, stocks currently identified as dead stock should be disposed of and labeled as such. This entails taking them out of the general warehouse and storing them separately. This frees up storage space and makes the dead stock more visible. Second, it should be investigated why these components have turned into dead stock. The first step translates to curing the symptoms of the disease, while the second step entails actually preventing the disease. In-house production using the injection molding machine was identified as a possible major contributor to increasing inventory turnover rates. Producing plastic components in-house instead of procuring them from other companies offers the option to eliminate large lead times, high minimal order quantities, supply uncertainty and could increase (in-house) ordering frequencies and therefore decrease average stocks of these components. However, current company policy is already aimed at increasing the utilization of the injection molding machine, meaning no policy change is needed. The fact that AME now uses the machine to produce parts that were previously procured from abroad means it has eliminated a factor that was a major contributor to slow turnover rates and buildup of dead stock in the past. Additionally, since implementing an inventory control strategy aimed specifically at incorporating the injection molding machine can be seen as a rather extensive task, this option should be considered out of scope. However, it can serve as a useful direction for future research for managers at AME. Forecasting plays an important role at AME. Since the lead time demanded by its customers is often significantly smaller than that offered by suppliers, a strategy involving both keeping stocks at the firm and forecasting future customer demand is used. AME currently maintains records of the quality of its forecasts, where the projected customer demand for a certain month is compared with the actual demand for that month. Some problems with these forecasts are that they are only specified for the final product, so a BOM would have to be used to translate demand for the final product into demand for raw components. The company does not maintain a mentionable WIP or finished goods inventory. Furthermore, from the forecasting time span it is also clear that demand for products usually only lasts for a small number of periods, making accurate forecasting a challenge. While forecasting is absolutely an area that deserves attention with respect to increasing inventory turnover, it should be considered out of scope for this current project. From the analysis it appears that many inventory items are not controlled in an optimal way, or are not even controlled by a clear and parameterized inventory control policy - instead they are controlled using human insight -, resulting in excessive dead stocks and an inventory turnover rate management is not content with. Especially for expensive items and slow moving items introducing such a control strategy should be a priority. An approach should be taken where based on recent demand and volatility of demand a suitable safety stock, reorder level and reorder 23

quantity are determined. This strategy should incorporate specific characteristics of SKUs, such as its value, and differentiate between several distinct groups of SKUs. In summary, a proper control strategy for all of the SKUs currently ordered should be developed, that can be easily implemented in the ERP-system used at the firm. Inventory turnover has been a largely invisible performance metric at the firm. It is mentioned in an automatically generated performance report, but it is unclear which company-specific factors determine this metric. Management should be offered a tool to investigate which SKUs and groups of SKUs are responsible for the inventory turnover rate. This should help identify trends and problems and help management take steps to prevent dead stock from building up and prevent slow moving items from having a detrimental impact on average inventory turnover. This system will also have the potential to project the results of implementing different control policies and help management quantify the potential savings of changing its inventory strategies. Literature suggests to use nearby and reliable suppliers to decrease variability, uncertainty and length of lead times. Additionally, using or negotiating a higher ordering frequency and decreasing minimal order quantities are feasible options. Using a limited number of suppliers allows for better relationships and room to negotiate more beneficial agreements. However, given that AME already operates using a limited number of preferred suppliers and is often bound by design of the product, not a substantial amount of room for supplier selection and negotiation is available. Therefore, a future project conducted at the firm could be devoted to optimizing the supply chain for AME, choosing the most appropriate suppliers in terms of location, product quality and delivery characteristics, and possibly optimizing the sharing of demand information throughout the supply chain. However, for this present research this aspect should be considered too extensive. Another issue suggested noteworthy in literature is that of inventory inaccuracies. During daily operations all sorts of events occur that can result in the registered inventory balance showing a deviation from what is actually physically in stock. Specifically, at AME, components can be lost during production, resulting in a larger number of components used than planned, while the difference can not be registered due to the manner in which components are stored. However, from inventory counting reports constructed based on annual counting procedures, the company has gotten a fairly good grip on keeping accurate inventory. Therefore, inventory inaccuracies can be assumed to have a negligible influence on inventory turnover at AME. Connected to these inventory inaccuracies is the phenomenon of entering incoming shipments into the inventory records. When supplies are received by truck, they are temporarily stored in a separate room, until an operator finds the time to perform an incoming quality and quantity check. In practice, shipments spend only a limited time in this room, meaning any issues with quality or quantity will not be detrimental. Beforehand, using commonalty in final products was suggested as a potential way to make the procurement of raw components more efficient. However, since many final products use customer specific components, and the few items that are common in multiple products are usually low value SMDs, commonality should not be expected to have a huge potential for increasing inventory turnover. In addition to suggesting to strive for commonality, literature mentions that customer specific stock is a potential risk. Should this particular customer go out of business, the stock could become virtually useless. Though AME maintains a large inventory that is customer specific, there are no viable methods to change this phenomenon. Customers often require very specific and custom designs, leaving little room to design for commonality. Moreover, the R&D department of AME has a fair amount of freedom in ordering components to

24

be used for prototypes. Though this could result in some buildup of stock that will not be used in the foreseeable future, only small volumes of raw components are involved and the department actually needs some freedom in ordering, since it facilitates the development of state-of-the-art products. The R&D department sometimes changes the design of final products in a relatively late stage on the introduction of a new product, specifically after procurement of components has started. This cycle is inherent to the introduction of new products, but does entail that the BOM for a product could change late in the project, resulting in the ordering of components that in the end are not used. It was established that the core and most expensive elements in a design are usually retained and therefore small alterations involving low value and common components do not have a detrimental impact on inventories. Furthermore, the analysis conducted in this chapter showed that rejects in production and delayed production runs due to part of the BOM for a final product being out of stock are hardly an issue. These issues could have resulted in a slightly slower turnaround of inventory, even though the effect would have been small. However, lowering the stock for one raw component could result in shortages for that SKU, possibly resulting in delayed production. These could in turn result in stocks for the other components to be used in the same production remaining at their old level for longer than planned. However, the effects of this phenomenon should be considered marginal. The utilization of components that are stored on tapes is very specific to the electronics assembly industry. These storage devices result in interesting minimal order quantity problems. Though storing these devices on tubes or sticks is possible, this is hardly an alternative for using tapes. The phenomenon of ordering components stored on tapes is inevitable and should be considered as given.

3.6

Summary

This chapter investigated the relevant performance characteristics of AME and subsequently discussed which strategies to increase inventory turnover are feasible given these characteristics. It can be concluded that the disposal and prevention of dead stock should play a pivotal role in attaining this goal. This can be accomplished through more effective stock control policies and by enabling managers to keep a closer eye on inventory turnover rates, through providing them with an easily accessible tool. The inventory control tool will make inventory analysis a faster analysis and makes inventory turnover rates a more visible metric. Additionally, the tool should help parameterize inventory control strategies, making stock control a more clear and defined process. The next chapter will detail how this tool was developed.

25

Chapter 4

Control tool The previous chapter identified that the main product of this project is a tool that will aid in the control of stocks at AME. The tool will allow for an up-to-date analysis of inventory kept at the firm, in addition to offering the option to determine the probable effect of changing inventory control policies for certain SKUs. The system is linked with the company ERP-system in order to automate importing of the most recent data and supplies the user with an evaluation of stocks and the projected performance of alternative inventory control strategies. The main analysis performed in this inventory control tool could have been done manually, however this consumes a tremendous amount of time. The major advantage of automating this analysis is that it can easily be performed whenever needed. This chapter describes the development of this system. It starts with listing the functional requirements of the system. Next, the main procedures implemented in the system are discussed. These include the extraction of data from the company ERP system, the formatting and filtering of data, performing calculations for individual SKUs, analyzing stocks and testing inventory control strategies. Finally, a short conclusion is given.

4.1

Functional requirements

The most important outcome for the company in this project is a stock control strategy that offers parametrization of decision variables that can be implemented in its ERP-system. Therefore, it is crucial the inventory control tool incorporates up-to-date information from the ERP-system and offers the user parameters that can be readily implemented in the same ERP-system. Furthermore, the system should offer a projection of the effects of adopting one of the suggested inventory control strategies and its associated parametrization of decision variables. These effects may include investment in inventory and thus potential savings, projected inventory turnover and expected drop in service levels. Control strategies are entered into the tool through changing part of the programming code of the macro which drives the tool. Naturally, allowing users to choose control strategies from a predefined list would be more user friendly, however for the purpose of this present study such interfaces are too elaborate. Additionally, the system should offer an evaluation of all active stock items, identifying slow moving SKUs and specific groups with similar characteristics in terms of value and turnover 26

speed. This functionality allows decision makers to keep a close eye on SKUs that may eventually become dead stock, and identify patterns. For instance, it may become clear that specific types of components are prone to turning over relatively slowly. Figures for inventory turnover rate are available from both the company ERP-system and its business intelligence interface. However, these figures are either specified for individual SKUs or for the overall inventory. Therefore, it is not possible to consider groups of SKUs using these figures. Additionally, poor documentation is available on how these figures are actually calculated. Therefore, it was deemed necessary to implement a custom calculation of inventory turnover rate. In addition, the system should offer an evaluation of dead stocks. This evaluation may include the same features as the evaluation of the active inventory, but should additionally show the size of dead stock. Importantly, the tool should include a module that calculates inventory turnover rates based on data extracted from the company ERP system. The methods currently available to obtain these inventory turnover rates are insufficient for the purpose of this project. This can be explained from the way inventory turnover rates are calculated. In order to determine the inventory turnover rate for a group of SKUs, the total cost of goods sold of these items is divided by the average inventory value for these items in the past year. The systems currently available at the company can only display the total inventory turnover rate for all items combined, or for each item individually. In both cases, it is not possible to aggregate or split up these rates into multiple groups of SKUs. This operation would allow for an analysis of groups of inventory, identifying which items are turning over slower than others. These items should then be controlled in a more effective manner in order to reduce their average inventory and thereby increase the overall inventory turnover rate. For these reasons, the tool is complemented with a custom inventory turnover rate calculation procedure, which maintains the figures needed to be able to aggregate SKUs into custom defined groups at all times. With regards to the implementation, it would be extremely useful if the methods developed were applicable on a more general level, outside of the current company setting. By using standard software options and commonly agreed upon definitions and assumptions, this general applicability is guaranteed. Additionally, in order to be usable for managers, the tool should offer some sort of interface which can easily be used to quickly view the most important figures and control the execution of the system. For this purpose, a control panel sheet should be incorporated in the system. Chapter 5 will further discuss the implementation of the tool.

4.2

Procedures

In this section, the exact procedures used in the inventory control system will be detailed step by step, from extraction of data from SAP to the processing of this data to outputting the data to the user.

4.2.1

Extraction from SAP

Whenever the user starts the control tool, after initializing variables, a connection with the ERP-system used by AME is established. The ERP-system SAP stores all information in socalled tables, which are interconnected. Normally, a user would log into an interface and use certain commands to access data from these tables, presented in an easily understandable format. However, for the purposes of this tool, direct access to the source of the data, which are the tables, is required. Therefore functions are available in Excel macros that establish a direct connection 27

with these tables. This connection ensures that data on inventory is always up to date and the analysis is as relevant as possible. Figure 4.1 presents a basic image of the design of the flow of information through the control system. Figure 4.1: Basic information flow

For a more detailed discussion of the extraction of data from specific SAP tables, please refer to appendix B. When extracting data from SAP, a function is used that requires certain parameters. Naturally, the table that will be accessed should be specified. Furthermore, the number of entries (rows) to be retrieved is specified. For this tool, a large number that will certainly include all available entries is chosen. It is relevant to note that should the firm grow considerably in the future, a close eye has to be kept on this figure. It is not unthinkable that a number that is deemed to be sufficiently large at present will exclude some entries in the future, if the company grows tremendously. Next, for each table, boundary conditions and filters are chosen. Later on, when the procedures for determining inventory turnover rate and implementing control strategies are discussed, the filtered data will be used. It is important that the correct filter be applied to the extracted data at an early stage, since this can significantly reduce processing time for the main algorithm. It is more time-efficient to filter when extracting from SAP, as compared to filtering the data once it has been extracted and stored in an Excel-sheet. The latter takes significantly more processing power.

28

Appendix C lists the filters and boundary conditions used. A final comment should be made. The extraction of data from SAP is a fairly time-consuming process. In the case of some tables, over 100,000 records need to be extracted. Moreover, as the firm grows, both the number of material movements and the number of SKUs stocked will increase tremendously, resulting in even larger extraction operations if the tool is executed. Therefore, it is advised that a function be added to the tool that will identify the most recent record previously extracted and only extract the movements that occurred after this date.

4.2.2

Formatting and filtering of data

The data that is extracted from SAP is in a specific format and suffers from some formatting artifacts that have to be repaired. For instance, the part number for SKUs is often followed by three spaces for no obvious reason at all. These artifacts can result in errors when running the system and therefore have to be removed. The same holds true for the formatting of dates, which are required to be in a specific format before an analysis can be conducted. As described before, all incoming stock movements in the past year have to be obtained from the ERP-system. Based on the coding of stock locations as described in chapter 3 this would entail all movements from locations 901 and 902 to locations 010 and 020. Unfortunately, some additional locations exist, and it is not always clear whether or not a certain stock movement contributes to the value of inventory for a certain SKU. Therefore, instead of focusing on location type, an alternative approach was chosen. The company ERP-system assigns a movement type to each stock movement. The specific movement types relevant to this project are 101, 712 and 262, as described previously. These are any movements coming in from outside of the company, into raw stock or directly onto the production floor. By only considering these movements, the actual number of raw components coming into the warehouse on a monthly basis can easily be found. Filtering for these movement types is done at the moment data is extracted from SAP. Additional filtering of data occurs at the time of extraction from SAP, as described before. SAP offers a table that contains the types of all SKUs, that is, shows whether each SKU is stored on tape or in a conventional piece-by-piece manner. However, for some SKUs, no type is available. Table 4.1 shows the count of SKUs for each category. Type Non-tape Tape Not available Total

Count 6,028 2,111 3,780 11,919

Table 4.1: Count of stock types There seems to be a significant number of SKUs that do not have information on their type available. However, further investigation revealed that SKUs that show no type information are not stocked currently. This indicates that these SKUs could be old parts that are currently not stocked any more but still have information stored in the ERP-system. Additionally, these could be components that will be stocked in the near future, with company employees still requiring additional information to complete the entry in the ERP-system. Finally, these could be items that are not part of normal production operations but instead support operations, for instance stationery. In the end, the decision to eliminate any SKUs that do not have type information

29

associated with them is justified. Any item that falls in one of the three aforementioned categories does not require strict control or monitoring, or in the case of new component introduction cannot be controlled tightly since insufficient data is available. After all tables have been formatted correctly and filtered, a loop is executed that compiles a list of all SKUs that have a type associated with them, along with relevant static data such as planned lead time and value per piece. This list is only compiled when the static tables are refreshed. Each time the actual system is run, it will use this list of SKUs as input and execute its analysis for every SKU present in the list.

4.2.3

Calculations for individual SKUs

After data has been extracted from SAP and transformed into the right format, the calculations of important metrics for individual SKUs can start. For each SKU present in the previously compiled list, an analysis is performed. The user can choose whether the analysis should be performed for the most recent period or for a chosen past year. In case the most recent period is chosen, this means the analysis will be performed up to the last full month, due to the updating frequency of the historic inventory position table in SAP. In case the analysis in performed for a past year, it will be executed for the period from January till December of that year. Next, the analysis performed for each SKU will be described. First of all, relevant dynamic data for the SKU is retrieved from the tables extracted from SAP at the start of running the system. This includes the active stock for the SKU and all incoming deliveries of the SKU over the past year. From SAP, all monthly closing inventory balances in the past year are retrieved. This procedure deserves some special attention, since the data obtained from the SAP table are not complete. Inventory levels are sometimes not available for every month. The goal is to obtain inventory levels for thirteen months. This will allow for the calculation of the number of items used in twelve months, through the simple equality: Inventory at end of month = inventory at start of month + incoming deliveries − items used First of all, inventory levels are searched for each of the thirteen months needed. Next, if an inventory level for a certain month can not be found, the level of the previous month is used. This process is repeated if the previous month does not have an entry in the SAP table either. The approach is valid, since SAP occasionally does not add an entry in the table if the inventory level did not change in a whole month. The data extracted from SAP include entries for 24 months before the current date, which should allow for sufficient room to go back in time and fetch the inventory level of a preceding month. If in this near history an inventory level can not be found, the level is assumed to have been zero all along. This approach appears to have an obvious drawback: if the item was ordered before this period and has never been used in production since then - making it dead stock - it will be assumed to have had zero inventory in the past year. However, this potential flaw is dealt with when calculating the relevant metrics, described below. First of all, the average annual inventory for the item is calculated. Thirteen data points are available,which are then used to determine the average inventory throughout the year by means of the standard calculation of an average number. Next, the number of components used of a particular SKU is determined for each month. This calculation is executed as follows:

30

Components used = M aximum of (Inventory at start of period + components bought − inventory at end of period, 0) By introducing the max function in this calculation, the problem of misclassifying dead stock as described previously is solved. The SKUs that are in fact dead stock and have not been used in production for a whole year will show zero components used for every month through the inclusion of the max function. Had the function not been included, the result of this calculation would have been a negative number, since stocks for the SKU had been at a level of zero until the point an entry was created in the SAP table to reflect the actual inventory level. Furthermore, the total number of items used in production and procured from suppliers in the past year is determined by accumulating all monthly figures. After these calculations have been performed, the inventory turnover rate is determined by dividing the total value of items used in production by the average annual inventory. If no inventory was present, this division is invalid and instead the inventory turnover rate is set to be zero. Figure 4.2 shows the data that is needed to determine the number of components used and the average inventory. Figure 4.2: Timeline

Where Ii = inventory level at start of periodi INi = inf low of components in period i OU Ti = outf low of components in period i Using these data, the average inventory for one year is calculated using: Average inventory =

13 P i=1

Ii 13

While the total number of items used in production is found using: Items used =

13 P i=1

OU Ti =

13 P

Ii + INi − INi+1

i=1

The approach described above utilizes a number of assumptions and suffers from some drawbacks, which will be listed and justified below. • Most importantly, the calculation of inventory turnover rate is not exact. This metric is approximated by considering intervals of the size of a month. Ideally, the inventory turnover rate should be calculated by tracking the exact inventory level for a SKU throughout an entire year and then determining the average level. However, one can easily see that doing this for a large number of SKUs can become an extremely time-consuming task, even when

31

using smart computer programs. Therefore, instead of constantly monitoring inventory levels, it may be more efficient to consider small intervals, measuring stock levels at a number of points throughout the year. Intervals could be a day, a week or a month in length. However, unfortunately, the frequency with which SAP registers inventory levels poses a serious restriction on choosing the length of this interval. No intervals smaller than one month can be considered. Using these fairly long intervals will result in a loss in accuracy in the calculation of the average annual inventory, and thus result in a diminished accuracy in the calculation of the inventory turnover rate. Although no alternative seems to be readily available, this loss in accuracy should be quantified. This verification of the usability of this form of calculating the inventory turnover rate will be performed later in this chapter. • The calculation of average annual inventory assumes that each month is of an equal length, naturally this is not true. However, the loss in accuracy associated with this assumption should be considered marginal. • The definition of whether a component has been used in production is not entirely crystal clear. For instance, when a set of components is sent to a production line to be used in a production run, they are not immediately consumed but instead could reside for a period of time before they are actually used. However, this is not considered serious enough to have a detrimental impact on accuracy. Subsequently, it is determined in which of the groups described in chapter 3 the SKU should be placed. This will allow for a more detailed analysis later. Next, SKUs are classified as dead stock or active stock. The definition of dead stock used here is extremely simple. If the SKU has not been used in production in the past full 12 months, it is considered as dead stock. In this case, the SKU is copied to the list of dead stock SKUs, which will be analyzed later. If the SKU has been used in production over the past full 12 months it is considered as active stock and placed in the list of active SKUs, which will also be analyzed at a later stage. After the lists containing the active SKUs and the dead stock SKUs have been drafted, the system then continues to analyzing these lists. It should be mentioned that although the main output of the program will be shown on the control panel sheet, these lists also contain extremely relevant data for decision makers. It allows them to analyze SKUs on an individual level, and identify separate SKUs that are weak performers. It is important to check whether or not the right data have been extracted from the company ERP-system, and if the basic calculations performed in the control tool yield results comparable to figures available in SAP. To test this, a number of SKUs are selected on a random basis to identify if the correct figures are present in Excel. This analysis is performed for five random components that are on the list for active stocks. The random selection was performed by using a random number generator and subsequently considering the SKUs placed in these rows. The results can be found in appendix E in tables E.1 and E.2. From the results, it is clear that the figures are extracted from SAP correctly, and simple calculations to determine the number of components consumed and procured are executed in the correct manner, albeit with some minor deviations which should be attributed to the definition of using a component that is used in the control tool. Though these actions seem simple and trivial, it is absolutely vital that these figures be valid, since all further calculations are based on them. Especially with regard to the consumption of components in production and the delivery of components, the removal of movements that should not be considered in calculations proved to be time-consuming and far from obvious.

32

4.2.4

Analysis of stocks

After all SKUs have been processed by placing them on a list of active stock or dead stock and determining their inventory turnover ratio - which naturally is equal to zero for dead stock -, the inventory kept at the firm will now be analyzed at an aggregate level. This is a three-way process: first, an analysis is performed for the active stock, which is all SKUs that have been used in production somewhere in the past year. Next, a number of control strategies are proposed and evaluated. Finally, dead stocks are analyzed. Active stock The analysis of active stocks is an extensive and functional part of the control system. It includes the calculation of safety stocks, aggregating SKUs according to a number of different procedures and showing the turnover performance of each of these aggregated groups. After this initial analysis, control strategies will be presented and implemented. In order to deliver a general overview of stocks present at the firm, some general figure are supplied. These include listing the total value of inventories present, the total value of active SKUs and the total value of dead stock items. Subsequently, the total average inventory, total cost of goods sold and the total inventory turnover rate are calculated. The inventory turnover rate for any SKU is given by the straightforward formula IT R =

Cost of goods sold Average inventory

Furthermore, for a group of SKUs, inventory turnover rate is given by: N P

Cost of goods soldi

i=1 N

IT R = P

Average inventoryi

i=1

Where the range {i...N} represents the group of SKUs considered. It is important that this metric be calculated in the right manner, since these figures will allow the firm to determine whether or not its actions concerning inventory control actually lead to the desired effect, which is increasing inventory turnover rate. Two separate checks are performed. First, the figures are compared to the figures available for each SKU in the company ERP-system. Subsequently, the total inventory turnover computed is compared to the figure available in the company business intelligence system. The ERP-system in use at the firm (SAP) keeps track of inventory turnover rates for each SKU separately. While these figures are useful and can come in handy to identify individual SKUs that are slow movers, it does not allow for easy aggregation of SKUs into groups. When the ratio is calculated, the two figures that were used - average inventory and cost of goods sold of the SKU - are lost. Without these two figures, it is not possible to aggregate SKUs into groups. For this reason, the present control system was developed. However, figures for individual SKUs as calculated by SAP should still be considered to be accurate in most cases. For this reason, the figures calculated using the control tool and the figures obtained from SAP are compared. Before this comparison could be conducted, the data had to be severely filtered. For 169 SKUs, SAP calculated an extremely high inventory turnover rate - 99999 - while the control tool yielded a result of zero. Naturally, these were SKUs that had an average inventory of (close to) zero

33

but were still used in production. These SKUs were omitted from the comparison, since the tool correctly identified these SKUs and since the difference is simply a numerical issue that does not impact the overall results. Secondly, for as many as 11543 SKUs, the control tool did not calculate an inventory turnover rate. These are SKUs that were listed as dead stock and consequently the turnover rate extracted from SAP is also equal to zero. Finally, some 226 SKUs showed a turnover rate of zero in the control tool while SAP calculated their turnover rate as larger than zero. These are SKUs that - for numerical reasons and due to design choices and assumptions - were calculated as having an average inventory level of zero and were artificially set to have an inventory turnover rate of zero to avoid divisions by zero. Omitting these SKUs from the comparison is unfortunate since it does reveal a potential weakness of the approach. However, since these are SKUs that on average do not represent a large investment, the error is considered negligible. Finally, the figures can be compared. Around twenty outliers were omitted from these results, since they showed an absolute deviation in inventory turnover rate of larger than 100. These outliers showed either a substantially high inventory turnover rate in SAP or in the control tool, while its counterpart was also high but less high by one order of magnitude. These outliers obscured the results and were left out from the analysis. One may wonder whether these data points should indeed be considered outliers. Closer investigation revealed that these were SKUs that were only in stock for a short amount of time. Due to the fact that the way the inventory turnover rate is calculated in the control tool suffers from some loss in accuracy as a result of its assumptions, these outliers can be considered as insignificant. In addition, these SKUs are items that turn over extremely fast, meaning they could not be responsible for the low inventory turnover ratio the firm is experiencing. Figure 4.3 shows the deviation between the turnover rates as obtained from the control system and from the company ERP-system. Figure 4.3: Deviation in inventory turnover rate for filtered SKUs

This deviation is determined by subtracting the inventory turnover ratio calculated by the control tool from the inventory turnover ratio obtained from SAP.

34

IT Rdeviation = IT RSAP − IT Rcontrol

tool

From this figures it is obvious that some major deviations exist between the two results, however for the majority of SKUs the two results are fairly similar. Considering the outliers, the differences can be explained by the method which is used to determine these rates. SAP considers the actual date movements occur, while the control system developed can only consider movements on a cumulative monthly basis. In SAP, if a SKU is used in production nearly immediately after is it delivered to the firm, its inventory turnover rate will be extremely high. On the other hand, in the control tool, its average inventory will likely be zero, since the inventory level for SKUs is only updated once a month instead of real time. The tool contains a safety measure to avoid divisions by zero (average stock level). In these cases, the inventory turnover rate is assumed to be zero. This artificial measure does not influence the overall average inventory turnover rate, but does result in SKUs that move extremely fast show an inventory turnover rate of zero. However, this should not prove to be a serious problem. Extremely slow moving SKUs that did move in the last year will never show up having a turnover rate of zero, making it easy to distinguish between the two types, for instance by filtering any SKUs that have zero turnover rate. Additionally, the most likely reason a SKU will show up having an inventory turnover rate of zero is if it has not moved at all within the last year, making it dead stock. However, in this case, the SKU is moved to the list of dead stock instead of being classified as active stock. For these reasons, the calculation of inventory turnover rates for individual SKUs is considered to be accurate. Some deviations exist due to the different method of calculation, however overall the results for the two methods of calculation are similar enough to allow the manual calculation to be used in further research. As a second form of verification, the inventory turnover rate as presented in the business intelligence interface used by the firm is investigated. This turns out to be 2.35 in case the cost of goods sold is extrapolated for the current year, and 2.93 in case the cost of goods sold is projected for twelve months head at the time of executing the check. The inventory turnover rate as determined using the tool for the same period is 5.06. This is a major difference, however several comments should be made. First of all, the firm uses different methods to determine the turnover rate. These methods are both focused on the future. One of them extrapolates the cost of goods sold up to the point of calculation for the whole current year, while the other projects the cost of goods sold for twelve months ahead. Additionally, the tool utilizes a method that is an approximation, so some additional minor deviations should be expected. Finally, the control tool considers the past full twelve months, while the company business intelligence system considers data from the past months of the present year and extrapolated or projected demand data for the future. The deviation between these two measures is substantial. The comments made can explain part of the difference, however a large gap between the methods still exists. Interestingly, a minor gap also exists between the two methods used in the business intelligence system. This difference can be explained from the fact that one method extrapolates data for the past months in the current year for the whole year, while the other utilizes projections of customer demand. A final comment should be made about inventory turnover. While a significant difference exists between the methods in the business intelligence system and the inventory control tool, the differences between the figures from SAP and the control tool were only small. These last two methods are far more similar in nature, in that they consider data for the past twelve months. Additionally, the nature of the inventory turnover ratio makes it a good measure to compare different firms and departments, or make comparisons through time for the same firm. The exact value for the inventory turnover rate in a firm does not bear significant meaning. For this reason, even though a substantial deviation exists between the figures from the business

35

intelligence system of the firm and the figure obtained from the control tool, since the method used to determine the rate in the tool adheres to the definition of inventory turnover rate and uses correct data as input, the figures found using this method can still be compared amongst each other. This way, the relative benefit of using certain inventory control strategies can be evaluated and the turnover rates of different groups of raw components can be compared. It is fairly common when analyzing large populations to make a classification in order to identify the most important individual entities. Specifically, in the case of inventory analysis, it is highly relevant to identify the most valuable SKUs so that a control strategy can be designed tailored to these specific items. One such classification is the ABC-division. The premise of this procedure is that a population is divided into three groups: group A, group B and group C. The first group are the small selection of items that represent a substantial amount of value. This value can be defined in a number of ways, which will be addressed later. Group B represents a somewhat larger selection of other items that represent a smaller amount of value than the group A items, while group B is a large selection of items that represents only a small portion of the total value of the population. In the present case, since the main goal of the project is to lower the average value of inventory present at the firm, the definition of value that should be used is that of current inventory value. The reason for choosing this definition is that in this manner, the actual and most current situation is considered. A potential drawback to this definition is that if the inventory value of a certain SKU is highly volatile throughout time, this item may be assigned to different groups over time. However, for the purpose of this study, the items that are most important in determining a low inventory turnover are the items that are expensive and move only slowly on an annual basis. These items will therefore not have a highly volatile inventory position. Alternative definitions of value of SKUs could have been annual average inventory value, average yearly demand expressed in value or value per piece. However, these definitions suffer from not being up to date and from not representing the actual inventory investment in the SKU. Although the choice for actual current inventory value does have a drawback, it is most suitable for the current study, as discussed above. Consequently, the exact parameters for the classification have to be chosen. There is no strict definition for these parameters, although the most commonly used definition is as depicted in table 4.2. Chapter 5 will discuss the application of the ABC-classification in the present company. Group A B C

Theoretical % of value 70% 25% 5%

Theoretical % of SKUs 20% 30% 50%

Table 4.2: ABC-classification Furthermore, SKUs are aggregated into groups according to storage type (tape & reel or conventional) and type of component, by looking at its product number. These have been described previously in chapter 3 and do not need require further explanation. It should be noted that the number of groups used to divide SKUs based on their product number range is limited at present. The firm uses an extensive tree to further distinguish components, down to a very detailed level. Implementing this tree in the control tool will likely be useful but requires an extensive effort. This further detailing of component types into groups should be considered a possible direction for future research.

36

The analysis that is conducted for the total stock is repeated for SKUs that have been assigned to the A-group (highest value) separately. These SKUs are of high interest and therefore deserve a closer look. Presenting a separate analysis for these items allows managers to closely monitor the components that have the highest impact on performance. Dead stock The analysis of dead stock plays a slightly smaller role in the system, since dead stocks are really a result of dysfunctional control of active stocks. However, firms should actively pursue dealing with dead stocks, since they do still represent an inventory investment, can take up storage space and hide problems. The dead stock analysis is mostly focused on showing the size of the problem, and which particular groups of SKUs represent the biggest part of this problem. The control panel will show an overview over dead stocks, including its total value, the number of SKUs and the impact of dead stock on the inventory turnover rate. Furthermore, a distinction is made between SKUs stored on tape and SKUs stored piece-by-piece, and between the different component groups as distinguished in chapter 3. For each of these groups, the value it represents and the number of SKUs in the group are shown. It should be noted that the component groups can be further divided into groups, in the end actually distinguishing between very specific groups of components. This way, managers can very precisely identify what type of components are mostly responsible for the value represented by dead stocks. Finally, the control panel will show the ten SKUs which represent the most value in dead stocks. These items deserve special attention from management.

4.2.5

Stock control strategies

A major function of the control system is simulating the effect of implementing distinct strategies to control inventories. These strategies are evaluated based on data from the last twelve full months. No reliable and deterministic data for demand for raw components in the future is available, for this reason historic data from the past twelve months is used. This data is not perfectly representative. However, since the demand for raw components can be expected to keep growing due to the projected increase in customer demand for finished products, it is vital to ensure that any data used is as recent as possible. Future research and additions to the tool may focus on evaluating control strategies based on projections for demand during the coming months. At present, however, data for average inventories for all SKUs and the number of components of each SKU used in production in the past year are used. By determining what the average inventory for each SKU would have been, given that a different ordering policy was used, the expected inventory turnover rate can be calculated. Additionally, a projection for the total investment in inventory can be given. These projections are based on two major assumptions. • Total demand during the past year does not change. Furthermore, demand occurs at a constant and linear rate. Naturally, this is not true in reality. Demand for a fair share of raw components is rather lumpy, with demand for certain parts occurring in peaks and no demand occurring in between production runs. However, especially for the most common components, demand is relatively continuous. • Lead times as planned in the company ERP-system are exact. That is, the figures are not projections but orders will actually arrive in the exact time specified. 37

These ordering policies can be based on a number of different factors. For instance, it is possible to base the policy on the value of the component. However, it would be more conventional to distinguish between a number of groups of SKUs, which are each controlled in a specific way. These groups can be based on the average value of inventories for the component over the last year (using the ABC-classification) or on more static characteristics such as the type of component (using the product number groups) or storage type (tape & reel or conventional storage). In order to allow for the implementation of control strategies, several different values have to determined. These include safety stock values based on set service levels, and economic order quantities. However, it should first be checked whether or not safety stocks will actually be necessary for the SKU considered. Some clues can be found in both the static and dynamic characteristics of SKUs. First of all, consider planned lead time for the SKU. If this is configured as zero in the ERP-system, stocking any components of the SKU should be considered uncalled for. Zero lead time may indicate special delivery procedures for the SKU, or an internal process that yields the component. Additionally, some SKUs show up having had zero average inventory over the past twelve months, while having been used in production at least once. This indicates special ordering procedures are in place for the SKU, where shipments are only ordered when needed and immediately consumed in production. Especially for expensive specialist components this is true. Therefore, for these items, no safety stock should be set. After having eliminated these SKUs from further calculations (setting their safety stocks to zero), the remaining SKUs are considered. Safety stock calculations are based on past demand and variability of past demand. The procedure to determine these two values is relatively simple. In order to determine the average monthly demand for a SKU, the total number of items sold in the past twelve months is divided by twelve, as one would expect. Average monthly demand =

Annual demand 12

Next, variability of demand is determined. In the case of determining safety stocks, this is expressed by the standard deviation. For discrete random variables, this metric is defined as s N P σ = N 1−1 (xi − µ)2 i=1

Where N is the number of observations, xi is the demand for observation i and µ is the average monthly demand. Normally, the common definition of standard deviation assumes that sufficient observations are available. Due to the method of determining past demand, only twelve observations are available in the control tool. This qualifies as a relatively small sample, which is why a different relation is used, which applies Bessel’s correction. This correction uses degrees of freedom instead of sample size, dividing by N − 1 instead of N Next, the actual safety stock is determined. This is done using a simple definition, which assumes lead times are deterministic, since generally planned lead times as quoted by suppliers are reliable. Only uncertainty in demand is considered. Saf etystock = σ ∗ k Where σ represents the standard deviation of demand for the SKU and k represents the service factor, which is defined by the choice of a service level. Table D.1 in appendix D shows the service factors resulting from the choice of a certain service level. The service level is defined as the probability that demand for SKUs during an order cycle can be fulfilled from stock. In other words, it indicates the probability that demand during lead time 38

will be less than or equal to the amount of stock present at the time of ordering. The definition assumes that the reorder point is positive and inventory is monitored continuously - meaning no stock outs can occur prior to reordering. Since AME utilizes an inventory tracking system that should be considered reliable and up to date, and it aims to provide perfect service and prevent back orders, these assumptions hold. This definition places a large emphasis on the stock out event, disregarding the amount back ordered. Other options for the definition of the service level include measures that take into account the amount of products back ordered and measures that account for the time that units are back ordered. For the purposes of this study, a measure that only considers the event of back ordering is sufficient, since back orders are to be prevented, regardless of their size and time span. Choosing a suitable service level can be a challenging task for firms. Naturally, a firm will aim to meet customer orders from stock. However, stocking sufficient items to reach high service levels can prove to be extremely costly. For instance, increasing the service level from 99.5% to 99.99% results in a 44% increase in inventory investment. Therefore, even though firms like AME aim for high customer service, the trade-off has to be carefully considered. Inventory investments can be expressed in monetary units, however it is hard to directly express the benefits of high service levels in such units. Companies may try to slowly decrease the investments in inventory from a level that yields a desirable service level to where negative effects start occurring to see what service level suits them best. It should be mentioned that when using these values for service levels it is assumed that demand is normally distributed. It is obvious that this is not the case for all SKUs, since some components are only used in a small number of assemblies. This means demand for the item may be largely dependent on one customer. The customer could always order in standard amounts, and not place any orders at all in some months. This yields a distorted view of standard deviation. However, for items that experience greater demand from multiple customers, a normal distribution can be assumed. Chapter 5 will further investigate this phenomenon. Next, using the safety stock levels, the reorder point (ROP) can be determined. This is expressed in a simple manner as ROP = Saf ety stock + E(demand during leadtime) All of which are expressed in number of components. The expected demand during lead time is found through E(Demand during leadtime) =

planned lead time 365

∗ annual demand

Where demand during lead time and annual demand are expressed in number of components, while planned lead time is expressed in days. This definition suffers from a major drawback: safety stocks and reorder levels can only be determined based on historic demand. Forecasts for demand for raw components are too unreliable to be used in present calculations. In order to determine the best quantity to order when this ROP is reached, the conventional method known as economic order quantity (EOQ) will be used. This method is based on the premise that an optimum exists where both ordering costs and holding costs are in balance. The method is extremely simple to use and requires few parameters as input. For this reason, it should also be considered a major simplification of reality, but for the purpose of this present study the method will suffice. The control tool can easily be adapted to allow for the application of more advanced methods to determine reorder quantities. Applying this method to the current situation means these two costs have to be quantified, which can be a troublesome operation in

39

practice. This issue will be addressed in chapter 5. For now it is assumed that these costs are known, which yields the following formula: q EOQ = 2DS H With D as the annual demand, S as the fixed costs per order and H as the holding costs. Chapter 5 will describe in detail which inventory control strategies were implemented in the system and subsequently evaluated for the demand occurring over the past year.

4.3

Conclusion

This section described the construction of the inventory control tool. It considered the demands on the functionality of the tool, the information flow structure, the main procedures used in the algorithm and the verification of these procedures. Importantly, some major assumptions were made and justified, and through several methods of verification the tool was shown to function properly.

40

Chapter 5

Application This final chapter is dedicated to the application of the control tool in the research environment. The added value of the tool is examined by using it to analyze the inventory turnover performance of the firm under investigation. In addition this chapter serves as a useful product for the firm. First, some general statements about the application of the system are made. Next, subsequent sections focus on characteristics of the active inventory, which control strategies are proposed to increase inventory turnover and what their projected benefits are and finally dead stock is analyzed. Mainly, this chapter will show how a manager at a firm would use the tool, drawing conclusions from the data presented on the control panel. Additionally, this chapter will show that the control tool developed does indeed complete the assignment as stated in chapter two. The first section in this chapter will shortly discuss how the tool would be implemented at a firm. Next, the active inventory of the company is analyzed. This includes an analysis of the current active inventory, and the historic development of the active inventory kept at the firm. Next, dead stocks are analyzed. Finally, a number of control strategies is implemented, tested and adapted.

5.1

Implementation

Due to the nature in which the control tool has been developed, it is relatively easy to implement at other firms. SAP is an ERP system that is used by an extremely large number of companies, while Microsoft Excel is a software package found on virtually all computers. Some remarks should be made with regard to implementing the tool at any company. First of all, it should be checked whether the fields in SAP tables conform to the definition of these in the inventory control tool. Sometimes, SAP offers multiple options to define certain metrics such as component values, for instance in the case of electrical wiring that can be measured in length, weight or number of pieces. Companies may have chosen to customize the implementation of SAP to fit their specific operational characteristics. In this case, careful consideration of the units used in SAP and in the control tool is required. Naturally, in case a firm does not have SAP records that date back more than a year - for instance if the firm has implemented SAP for the first time in the past year - the tool will not function.

41

Technically, implementation comes down to saving the Excel macro file to a computer that has access to the company SAP database. In the macro, the user name and password of the user have to be specified, in addition to several parameters that specify the implementation of SAP in the company. Consequently, managers can run the tool by pressing a button in the control panel sheet of the tool. After taking some time to extract all required information from SAP and performing a range of calculations, the system will then show an analysis of stocks on the control panel sheet. Additionally, it will provide projections for stock value and inventory turnover rates for the total stock and aggregated groups of SKUs for the control strategies that are defined in the macro that runs the system. These strategies can be changed in the macro programming code. For these strategies, the sheet that contains all active SKUs lists the safety stock and reorder point for each SKU. These parameters can then be easily entered into the company ERP system, effectively implementing the chosen control strategy. A control panel was developed, allowing users to easily control the tool and view important figures. Appendix ?? shows a screen shot of part of this control panel.

5.2

Active inventory

First of all, the active inventory present at the firm is considered. The tool presents the user with a clear overview of important figures, allow them to get an understanding of the composition of active stock, the value each group of SKUs represents and how fast these SKUs are moving. Both the current situation and development throughout history will be considered in order to evaluate if certain trends are visible.

5.2.1

Applying the ABC-classification

Before analyzing the inventory kept at the firm, the ABC-classification discussed in chapter 4 is applied to the stock present at the firm. The definition of the A, B and C groups will now be considered. The definition of these groups most often used in research was discussed in chapter 4. Appendix F shows how this definition fits the company. This model for the ABC-classification does not fit the inventory under investigation well. Especially group A seems to be a rather bad fit. Naturally, it would seem advantageous if group A represented relatively more value that anticipated, however the distribution of SKUs among the other groups will show some deviations in this case. An alternative model is defined as shown in table 5.1. Table 5.2 shows the results of this adapted model as applied to the active inventory of AME. Group A B C

Theoretical % of value 66.6% 23.3% 10.1%

Theoretical % of SKUs 10% 20% 70%

Table 5.1: Alternative definition of groups This time, the model is a good fit. Evidently then, at AME, a significant amount of value is represented by a small amount of SKUs.

42

Group A B C

Actual % of value 66.64% 22.35% 11.02%

Actual % of SKUs 8.61% 15.38% 75.99%

Table 5.2: Alternative groups applied

5.2.2

Current inventory

Next, the current situation at AME is investigated by considering the figures presented on the overview sheet of the control tool. These figures should be expected to be similar to those presented in chapter 3. This is actually the main strength of the control tool. It allows managers to make a thorough analysis of inventories, which would normally be conducted manually, but can now be executed at any given moment. Looking at the general data for the inventory kept at the firm, consider table 5.3. Current total value Number of SKUs Inventory turnover rate Annual average total value Value used in production

e1,965,282 3,452 4.75 e1,784,345 e8,477,446

Table 5.3: General inventory data This yields figures comparable to those presented in chapter 3, except for the inventory turnover rate. However, the difference in that metric was explained in chapter 4. Otherwise these figures do not show any unexpected issues. After this general analysis, an in-depth analysis of stocks and inventory turnover rates will be presented. First of all, the storage type of components is investigated. As chapter 3 explained, components are either stored on tape, or in a conventional manner. Different storage types could lead to different turnover rates due to minimal order quantities, for instance. Table 5.4 presents some figures for both of these storage types.

Tape Non-tape

ITR 5.18 4.14

Stock value e1,147,232 e818,050

Share of stock value 58.37% 41.63%

SKUs 1,974 1,478

Share of SKUs 57.18% 42.82%

Table 5.4: Storage types From this table it is clear that components stored on tape constitute a slightly bigger part of the inventories kept at the firm, both in terms of number of different SKUs and in terms of stock value. Additionally, they turn around slightly faster than non-tape components. Overall, this table does not offer any shocking information, other than that in spite of the minimal order quantities that place huge constraints on the ordering of tapes, these components turn over slightly faster than other components. However, this can be explained from the fact that these components are used in many different products. Next, consider figure 5.1. This figure presents the total value each type of component currently represents. Special attention should be drawn to the value of the group finished products, which 43

is virtually zero. This confirms the earlier statement that no finished goods inventory is kept. Furthermore, the most valuable groups can be identified. Figure 5.1: Value per component type

More interestingly, figure 5.2 displays the inventory turnover rates for each group of components. It should be noted that in this case an inventory turnover rate of zero corresponds to an extremely high inventory turnover rate, where division by zero is a risk. Therefore, the groups purchased sub assembly and finished products should be considered as extremely fast moving and therefore do not form a problem. Additionally, the sub assemblies group turns rapidly as well. Comparing this figure with the data from figure 5.1 can help identify which groups of components deserve special attention. The slowest moving components appear to be in the first five groups of components, and in the last group - consumable goods. However, of these groups, only passive components, analog components, digital components and electromechanical components represent a substantial inventory investment. Therefore, these groups should be targeted as the main drivers of inventory turnover. Finally, figure 5.3 depicts the number of SKUs present in each of the component type groups. What is extremely interesting to see here is that the analog and digital component groups only contain a limited number of different components, while the passive and electromechanical groups contain a substantially bigger number of components. This is interesting knowledge since it entails that by implementing a more efficient control mechanism for a fairly limited set of components, substantial increases in inventory turnover rate can be obtained. As an additional recommendation, it should be noted that these component groups can be

44

Figure 5.2: Inventory turnover rate per component type

further specified and detailed by using the component tree used in the firm. This allows for the creation of a far larger amount of groups. That way, an even more targeted analysis can be performed, which should result in the firm being able to tackle the small subset of components that are responsible for a slow average turnover. It would enable managers to observe patterns, for instance certain specific types of components may be prone to slow turnover rates. Next, to further investigate stocks and determine which items are the absolute most important determinants of inventory turnover, the ABC-classification is applied. The metric used to determine these groups is the current inventory present of each SKU. Table 5.5 shows the results of applying the ABC-classification to the current inventory of the firm.

A B C

ITR 4.34 4.19 6.5

Stock value e1,309,850 e439,063 e216,371

Share of stock value 66.65% 22.34% 11.01%

SKUs 306 524 2,622

Share of SKUs 8.86% 15.18% 75.96%

Table 5.5: ABC-classification of stocks This table does not reveal any shocking facts. However, after dividing the inventory into these A, B and C groups, another analysis similar to the analysis performed for the total stock is executed. This will show which components in the A-group - which makes up most of the value present at the firm - are most important. These components should therefore be considered the 45

Figure 5.3: Number of SKUs per component type

most important inventory, which needs to be managed extremely closely. Table 5.6 shows the inventory turnover rate, value and number of SKUs in each group of components, with each of these components being included in the A group of the ABC-classification. Group Passive Discrete Analog Digital Electromechanical Sub Assemblies Purchased sub assemblies Finished Products Packaging Material Computers & Accessory Consumable Goods

ITR 2.77 5.15 3.41 4.43 3.42 8.34 0 0 0 4.52 2.22

Stock value group A e181,229 e46,595 e173,022 e318,685 e241,547 e281,977 e0 e0 e0 e10,156 e56,634

Share 13.84% 3.56% 13.21% 24.33% 18.44% 21.53% 0% 0% 0% 0.78% 4.32%

SKUs 39 13 45 59 71 61 0 0 0 4 14

Table 5.6: Figures of component types in group A The table shows some interesting facts. First of all, there are 39 passive components which make up around e180,000 of stock value - out of a total of about e1,800,000 - and have a slow turnover rate of only 2.77. The same holds true - but to a lesser degree - for 45 analog components and 71 electromechanical components. These specific SKUs can easily be found using the control tool, and subsequently should be controlled in an appropriate manner. 46

5.2.3

Historical development

By extracting information for a different time span than the past twelve months, the control tool allows for an analysis of historical development of important figures such as inventory value and turnover rate. Some comments should be made to place this historic analysis in the proper perspective. First of all, the control system considers the list of SKUs that are currently listed in the firm ERP-system and attempts to retrieve historic data for these SKUs. Naturally, some of these SKUs were not yet in use in the firm in the past and will therefore show up as not having been used once in production. For this reason, the number of SKUs classified as dead stock cannot be counted. Additionally, in 2007, the firm first started maintaining inventory records in SAP. Therefore, the analysis starts after this year, since no inventory turnover rate can be determined for a year that does not have a fully recorded history available. Consider table 5.7, which presents an overview of several overall metrics through time. Year 2008 2009 2010 2011 2012

Average inventory value e689.845 e952,645 e1,276,036 e1,270,444 e1,765,922

SKUs 2,319 2,446 2,848 3,425 3,479

Value used e3,675,806 e3,736,839 e5,728,816 e6,946,359 e8,414,406

ITR 5.33 3.92 4.49 5.47 4.76

Table 5.7: Historic overview of performance Note that these figures are determined by considering all SKUs that were used in production during the year considered, that is, were active stock and not dead stock at the time. Figure 5.4 presents the development of the inventory turnover rate through history. Interesting enough, there seems to be hardly any change. Please note that these turnover rates were determined at the end of each year. Determining them on a monthly basis for a year back may show a slightly different result, but the overall trend seems to be fairly constant, turnover rate has not increased significantly over time. Furthermore, figure 5.5 depicts the development of average inventory value and the value of components used during one year. As could have been expected for a firm that is growing rapidly, both increase at a rapid and linear rate. Using the control system, a more elaborate analysis of historical developments related to inventory and turnover rate can easily be made. Especially in the case where elaborate changes in the control strategy used by the firm have been made in history, it will be extremely interesting to be able to look back in time and verify whether or not these policy changes have resulted in the desired effects occurring. The analysis can be further detailed by also considering the component type groups identified earlier, and expanding these groups further into the tree maintained by the firm.

5.3

Dead stock

The control tool offers a final section devoted to dead stocks. Naturally, in a way, once a component has turned dead stock it is already too late. However, by keeping a close eye on

47

Figure 5.4: Development of inventory turnover rate

Figure 5.5: Development of stock value and value of components used

48

which components turn dead stock and how much value is present there, some indications of which components are more likely to become dead stock eventually can be obtained. Figure 5.6 shows the value of dead stocks of each component type. Figure 5.6: Dead stock value of component types

This figure shows that by far and large, electromechanical components represent the most value. One possible explanation for this is that AME ordered substantial amounts of plastic components from Asia. For these orders, large minimal order quantities were in place, resulting in a buildup of stocks. Luckily, the firm is currently in an advances stage of producing these components in-house using injection molding techniques. Additionally, it appears that sub assemblies and passive components represent another substantial share of dead stock value. Next, consider figure 5.7, which shows the number of SKUs present in each group of dead stock components. It appears that electromechanical SKUs are far more likely to turn into dead stock. At the same time, although these components represent a large share of dead stock value, the value each of these individual SKUs represents is small, since a large amount of SKUs are dead stock. Finally, with regard to reducing dead stocks, a number of options have been suggested in the literature review. For instance, dead stocks could be sold back to suppliers or sold on the open market to other firms. To reduce the chance of inventories turning into dead stock, it is possible to postpone payment to suppliers until the parts are actually used, although most suppliers will not be eager to participate in such an agreement. Additionally, the firm should absolutely investigate the option of lowering the minimal order quantity products are ordered in, even if it may have to pay a small additional fee. It should be stressed once again that although disposing of dead stocks should be pursued, these stocks are always the result of a failure in the management of inventories in an earlier 49

Figure 5.7: Number of dead stock SKUs per component group

period. Therefore, dead stocks should be prevented instead of disposed of. his can be done by implementing the proper control strategies, which are presented and tested in the following section.

5.4

Proposed control strategies

In addition to analyzing the characteristics of current stocks, the control tool allows for the calculation of expected performance given that a certain control strategy was used to control inventories during the past twelve months. This functionality can offer useful insights. By frequently running the tool - after a month has concluded - proposed control strategies are rendered more useful since they then incorporate the most recent available figures of demand for components. Running the tool more often than this currently does not improve the analysis in terms of relevance, since due to the limitations of the company information system, only data for months that have finished can be considered. This section will describe the search for a suitable strategy to control inventories at AME. First of all, the current performance is listed again, in table 5.8. This will serve as a basis for comparison. Note that currently, the control of components is performed in a number of different ways. For some components, safety stocks and reorder points are defined, whereas for other components purchasing is based on expected future demand for the item, whether based on forecasts made by customers or on manual forecasts based on experience and insight. Control strategy Current

Stock value e1,965,282

Turnover rate 4.75

Table 5.8: Current performance

50

First of all, with regard to the calculation of the economic order quantity (EOQ), an important assumption has to be made. No data on the costs of ordering and holding is readily available. The firm has made agreements with most of its suppliers that no delivery costs are paid, or in some cases are added to the cost per unit ordered. However, this does not mean that ordering costs should be assumed to be zero, since any time a component is ordered it still requires processing by a human operator upon arrival. Additionally, the costs for storage mostly result from the costs of installing and subsequently operating an advanced warehousing system. For these reasons, and for simplicity, it is assumed that the costs of ordering components and the costs of holding are of equal magnitude. This is a major simplification, however firms can easily change these numbers. Since the present purpose is to test the implementation of the tool, this should prove to be a reasonable assumption. If these costs are of equal magnitude the formula for EOQ reduces to: √ EOQ = 2D Next, some simple strategies are tested. These strategies assume that all components are controlled in the same way, which is by setting a service level, and determining the associated safety stock and reorder points. Reorder quantities are rounded up to respect minimal order quantities as set by suppliers. Table 5.9 shows the projected stock value and inventory turnover rate using these strategies. Proposed control 75% service level 80% service level 90% service level 95% service level 99% service level

Projected value e1,660,018 e1,928,842 e2,626,773 e3,196,552 e4,288,004

Saving e12,4327 e-144,496 e-842,427 e-1,412,206 e-2,503,658

Relative saving 6.97% -8.1% -47.2% -79.1% -140.3%

Projected ITR 5.11 4.4 3.23 2.65 1.98

Table 5.9: Different service levels in (s,nQ) strategy with safety stocks and EOQ These results are quite revealing. If stock were to be controlled using a standard (s,nQ) strategy, retaining the same investment in inventory would result in service levels between 75 and 80 per cent. Arguably, this is unacceptably low. If a higher service level had to be attained, this would require a 140 per cent increase in inventory investment, resulting in a dramatic drop in inventory turnover rate. Clearly, an (s,nQ) strategy with high service levels is not suitable to the demand processes for raw components at AME. Further investigation revealed an interesting phenomenon. A relatively small group of SKUs is responsible for a disproportional share of the inventory investments that would be required if the proposed (s,nQ) control policy were applied. What all these SKUs have in common is that the lead time planned for their procurement is incredibly large, ranging from 56 days at least to as many as 720 days - two complete years. With lead times this large, it comes as no surprise that applying an (s,nQ) strategy with safety stocks based on service levels yields incredibly high inventory investments. Table G.1 in appendix G shows some of the SKUs that require enormous inventory investments assuming service levels of 90 per cent. In addition to having a large lead time, most of these items show an annual demand that is considerably higher than their average inventory level. Furthermore, these items have extremely large standard deviations in demand. These characteristics combined suggest that these components are only needed a limited number of times each year, but in large quantities if they are demanded. Furthermore, the long lead times mean huge safety stocks should be kept in order

51

to attain high service levels, however this makes (s,nQ) control infeasible due to the enormous associated costs. For these reasons, a different control strategy should be implemented for these problematic SKUs. Specifically, it should be assumed that the firm currently procures these items to order, that is, purchases these specific SKUs only when it has confirmation from customers that the products the parts are used in will indeed be ordered. For these items, that control strategy should be considered more feasible than the (s,nQ) control proposed. In this case, it can be assumed that the number of items used annually, and the average inventory level do not change compared to the considered historic period if the current control strategy is retained. This just leaves the question of which items to control as before, and for which items to implement (s,nQ) control as suggested in the initial attempt. At this point it makes sense to involve the ABCclassification to determine which items should be excluded from the proposed (s,nQ) control and instead be controlled as the company does at present. As discussed before, the ABC-classification is implemented to consider the items that see the largest share of cost of goods sold annually. However, in the current situation, an unconventional problem is faced. One would normally want to offer high service levels on items placed in group A, while groups B and C are considered less important and therefore require less tight control. In the present situation, the problem can be described as: How to control stocks in such a way that items that see a large variation in demand and often have extremely long planned lead times in such a manner that a reduction in costs - through a reduction in average inventory - is achieved? Specifically, what is the most suitable way to distinguish whether an item should be controlled using a service level, or by forecasting and ordering to demand - the way the company currently controls the inventory of these items? In this case, it makes sense to modify the definition of the ABC-classification. Specifically, it should be modified so that the SKUs that would be controlled using extremely large safety stocks and reorder points should not be controlled using the proposed (s,nQ) control but instead in the manner they were previously controlled, which is simply described as ordering to forecast demand. These forecasts are constructed both through communicated customer demand and by manual insight. Table 5.10 shows the projected stock value and inventory turnover rate when using an inventory control strategy that uses the aforementioned ABC-classification. That is, the items designated to be in group A - constructed based on projected inventory value if safety stocks are used - will be controlled as before, while groups B and C are controlled using a 90 per cent service level. This way, the items that would be controlled using enormous safety stocks are controlled as they have been during the last year, while items that will not have huge projected inventories are controlled using safety stocks. This strategy is justified since the items that would have enormous projected average inventories are special items in terms of lead times, storing considerations and production characteristics. These items are controlled in a specific way at present. Controlling them using safety stocks and reorder points would be sub-optimal. Proposed control ABC

Projected value e1,645,956

Saving e138,389

Relative Saving 7.76%

Projected ITR 5.15

Table 5.10: Results for ABC strategy This strategy is expected to obtain a decrease in average stock value of nearly eight per cent. The potential drawback naturally is that service levels may be lower than before, even though as shown previously in this chapter, the current control of inventories translates to a service level between 75 and 80 per cent, while the present suggested strategy controls most items with 52

a service level of 90 per cent. The strategy should be further fine-tuned to obtain additional savings in inventory investments, for instance by combining it with control aimed at specific groups of components. Subsequently, since lead times seem to be an important determinant in projected stock value when using safety stocks, the option of making the control strategy depend on the planned lead time for each SKU was also investigated. In case the planned lead time was larger than a specified limit, the SKU was controlled as before, while if the planned lead time was smaller, the SKU was controlled using safety stocks and a 90 per cent service level. The assumption here is that for relatively long lead times, the firm already uses specialized control, mostly based on human insight, which cannot be replaced by a sturdy control using specified safety stocks and service levels. Table 5.11 shows the projected stock value and inventory turnover rate for this strategy, using a number of different limits for the lead time. Proposed control Lead time > 7 days Lead time > 56 days Lead time > 80 days Lead time > 90 days

Projected value e1,701,474 e1,868,450 e1,867,508 e1,881,250

Saving e82,871 e-84,104 e-83,162 e-96,904

Relative Saving 4.64% -4.71% -4.66% -5.43%

Projected ITR 4.98 4.54 4.54 4.51

Table 5.11: Results for lead time strategy Surprisingly, this strategy does not yield satisfactory results. One would expect that by focusing on items that have large planned lead times, enormous safety stocks can be avoided. However, this approach seems to suffer from omitting stricter control for some of the components with long planned lead times but predictable demand and average stocks that were previously high already. In summary, the strategy that seems the most promising is the one that uses the ABC-classification to distinguish the items that would not benefit optimally from using safety stocks and service levels and controlling these as before. This somewhat resembles the strategy of cherry-picking, where a specific operation is only applied on the entities that would benefit most from it. The strategy would most probably benefit substantially from further fine-tuning, possibly combining it with the strategy targeted at different component groups, thereby tailoring it to address those specific components that have the most detrimental impact on inventory turnover. Some remarks should be made with regard to the assumptions made while working towards the implementation of the control tool, and the sensitivity of the results to these assumptions. A major assumption was that demand for raw components from the production floor is normally distributed. This assumption is not realistic for a large number of components, especially the ones that are only used sporadically in an extremely limited number of different final products. A considerable amount of raw components is only used in one final product. For components that are used often and in multiple final products, the assumption that demand is normally distributed is more realistic. Overall, the assumption that demand is normally distributed should be expected to have an impact on the results. However, the final control strategies accounted for the fact that some components are only used sporadically and in a limited number of products, and recognized that the control mechanism the firm already used for these items should be assumed to be effective. For these items, the safety stocks based on service levels (for which the assumption of normally distributed demand is used) do not play a role. Instead, mostly the items that see a large annual demand are controlled using service levels. For these items, the assumption of normally distributed demand is more realistic. 53

Another major assumption was made with regard to demand and production shortages. It was assumed that when projecting how average stock levels would have been in the past year using the suggested control policies, demand during that year is equal to its actual observed pattern, and production runs were completed as they were in reality. It is not hard to imagine that if different control strategies were applied in the past twelve months, production may have proceeded differently. For instance, shortages of raw components could have occurred. Additionally, the scheduling of production jobs may have been adapted following these shortages in order to maximize utilization of the two production lines by postponing jobs that could not be started as a result of these shortages. However, since service levels overall did not change significantly following the introduction of adapted control strategies and control of some of the more expensive and specialized components was retained as before, it should be considered reasonable to assume that overall, production schedules remained relatively unchanged. Concluding the discussion of the application of the control tool, several remarks with regard to its practical use can be made. First of all, due to the integration with SAP and the accessible implementation in Microsoft Excel, the system can be easily implemented at other firms. Furthermore, firms often have custom information systems that extract information from SAP and present it in a summarized and visual manner, similar to the control panel included in the tool developed in this project. These visual overviews are often constructed using Excel macros. This means that the tool developed in this project can easily be migrated to most custom information systems in use at other firms. AME has already shown an interest in integrating the tool in their own business information system graphical summaries. The tool is useful to firms because it makes it easy for managers to keep a close eye on inventory turnover rates for specific groups of items. At the same time, it offers them the functionality to investigate what the possible effects in terms of change in inventory turnover rate are of implementing different inventory control policies, based on historic demand. That is, the tool can be used to predict the effect of changing inventory control policies before actually implementing them. This is extremely useful in practical situations, since changing inventory control can be a considerable operation and possibly have serious negative effects if an improper stock control policy is chosen.

54

Chapter 6

Conclusions & recommendations This report described the execution of a project to increase the inventory turnover rate at a medium-sized electronics assembly company, through the use of a control tool that uses up-todate data extracted from the company ERP-system to analyze inventories and proposes appropriate inventory control strategies. A detailed analysis of the firm was conducted, which was used to guide the direction for further research. Based on the analysis, the most promising approaches for increasing the inventory turnover rate were discussed, along with providing reasons for why certain common approaches should not be expected to yield satisfactory results. Next, an inventory control tool was developed which was subsequently implemented at the firm, in order to test the methods applied and to obtain some interesting conclusions. The tool provided some directions for management to use in increasing inventory turnover. Specifically, it indicated which types of components had a low inventory turnover and high inventor value. It also provided an analysis of dead stocks, however most importantly the tool offered the option to implement and test specific inventory control strategies. Though the projected increase in inventory turnover was not substantial, the strategies do provide some directions for further investigation and indicate which policy change will and will not lead to desirable effects. In this section, conclusions that can be drawn from the project and presented, and some recommendations for future research are suggested. Conclusions Importantly, this study was mainly focused on decreasing the average inventory, mostly for expensive SKUs. However, from the definition of inventory turnover rate it is obvious that increasing this metric can be achieved in another way, which is through increasing the cost of goods sold while maintaining the average inventory at the same level. In practice, this translates to using inventories more efficiently, which is mainly achieved from a production perspective. It should be noted once again that although increasing inventory turnover rate is a worthy goal, a low inventory turnover rate is always the consequence of other problems, not a cause of problems.

55

Therefore, firms should strive to investigate these causes and prevent them from causing a low inventory turnover rate. Clearly, this project has been a highly relevant study, with less emphasis placed and its academic rigor. However, some generalizations can be made in order to make the methods applied relevant in a wider and more general context, whilst also positioning the study in literature and comparing it to other companies and industries. What makes the study interesting is that is was conducted at a relatively small but rapidly growing high-tech assembly company. The sheer fact that the firm is growing rapidly means investments in stocks increase at a high rate, increasing the danger to procure components that eventually turn into dead stock. Furthermore, the growth in volume of finished products, number of different products produced and customers served also means that historical figures for demand are hardly usable. The company aims to be highly efficient in its use of inventories, preferring to maintain low safety stocks, few dead stock and overall run an extremely well organized and clean warehouse. Finally, what distinguishes this firm (and other electronics assembly firms) from for instance the retail sector, is that it operates using rather large lot sizes. This can be attributed to the use of storage devices called tape & reel, which carry up to thousands of small surface mounted devices. Though often extremely cheap per piece, some of these SMDs can be fairly expensive, making minimal order quantities and rounding values a highly relevant issue. In the case of retail firms, case pack sizes are often considerably smaller. Reviewing the steps taken from start to end - which can and should be repeated at any other firm - the following summary can be provided. First of all, an initial analysis of the firm should provide directions for potential sources of problems - slow moving items or dysfunctional inventor control - and point out any unique characteristics of the firm. Next, a system should be developed that will use recent data from the ERP-system used at the firm, to generate up-to-date reports of the status of the inventory kept. Naturally, the system should be tested and verified. Finally, the system is implemented and used to analyze inventories currently kept, and to implement and test specific control strategies. In determining the proper control strategies, the analysis performed at the start of the project should be kept in mind, to account for any unique characteristics of the company and to target the problematic phenomena that were identified previously. Positioning the study in literature, it can be classified as providing a moderately hands-on procedure, due to its high relevance resulting from the fact that the study was conducted at and for an actual firm. However, the steps followed and framework provided should be considered general enough to make the procedure applicable in other firms and industries. Recommendations Finally, some recommendations for future research should be made. First of all, in chapter 3 it was already mentioned that the injection molding machine should be appropriately used in production schedules. Some optimization can be done in order to ensure that lead times are short, utilization of the machine is high and stocks are controlled. Furthermore, the control tool can be optimized in its code as programmed in the macro to deliver faster results. Mainly, it should check to see which movements records are new and only extract those from SAP. In terms of actually increasing inventory turnover rates, the firm should investigate whether instead of reducing stock levels, these can be retained to achieve higher output. Specifically, since the firm expects to grow significantly in the coming years, it should study whether it can limit the increase in stocks while increasing its output of finished products. The analysis implemented in

56

the control tool can be further specified by incorporating a deeper level of the tree the firm uses to classify its raw components. Finally and most importantly, the control strategies suggested can be significantly refined. This refinement should be done through combining the strengths of separate strategies, incorporating the deeper level of the component type tree and adding human insight to account for subtleties a control policy may not be able to process.

57

References Baker, K., Magazine, M., & Nuttle, H. (1986). The effect of commonality on safety stock in a simple inventory model. Management Science, 982–988. Cachon, G., & Fisher, M. (2000). Supply chain inventory management and the value of shared information. Management Science, 46 , 1032–48. Chadwick, L., & Waddington, A. (1982). Inventory control: how to reduce stockholding costs. International Journal of Retail & Distribution Management, 10 (3), 61–62. Corbett, C. (2001). Stochastic inventory systems in a supply chain with asymmetric information: Cycle stocks, safety stocks, and consignment stock. Operations research, 487–500. Croston, J. (1974). Stock levels for slow-moving items. Operational Research Quarterly, 123–130. Ernst, R., Guerrero, J., & Roshwalb, A. (1993). A quality control approach for monitoring inventory stock levels. Journal of the Operational Research Society, 1115–1127. Finkin, E. (1989). How to limit inventory expenses. Journal of Business Strategy, 10 (1), 50–53. Forslund, H., & Jonsson, P. (2007). The impact of forecast information quality on supply chain performance. International journal of operations & production management, 27 (1), 90–107. Gardner Jr, E. (1990). Evaluating forecast performance in an inventory control system. Management Science, 490–499. Gips, J. (1998). Mission impossible? the boss wants to double our inventory turns. Hospital materiel management quarterly, 20 (2), 34. Goffin, K., Szwejczewski, M., & New, C. (1997). Managing suppliers: when fewer can mean more. International Journal of Physical Distribution & Logistics Management, 27 (7), 422–436. Harrington, T., Lambert, D., & Vance, M. (1990). Implementing an effective inventory management system. International Journal of Physical Distribution & Logistics Management, 20 (9), 17–23. Holstr¨ om, J., Fr¨ amling, K., Kaipia, R., & Saranen, J. (2002). Collaborative planning forecasting and replenishment: new solutions needed for mass collaboration. Supply Chain Management, 7 (3), 136–45. Jacobs, R., & Wagner, H. (1989). Reducing inventory system costs by using robust demand estimators. Management science, 771–787. Jahnukainen, J., & Lahti, M. (1999). Efficient purchasing in make-to-order supply chains. International journal of production economics, 59 (1), 103–111. Johnston, F., Boylan, J., & Shale, E. (2003). An examination of the size of orders from customers, their characterisation and the implications for inventory control of slow moving items. Journal of the Operational Research Society, 54 (8), 833–837. Kinney, M., & Wempe, W. (2002). Further evidence on the extent and origins of jit’s profitability effects. Accounting Review , 203–225.

58

Lee, H., & Tang, C. (1997). Modelling the costs and benefits of delayed product differentiation. Management science, 40–53. Ploss, G., & Wight, O. (1967). Production and inventory control: principles and techniques. Textbook . Primrose, P. (1992). The value of inventory savings. International Journal of Operations & Production Management, 12 (5), 79–92. Rantala, L., & Hilmola, O. (2005). From manual to automated purchasing: case: middle-sized telecom electronics manufacturing unit. Industrial Management & Data Systems, 105 (8), 1053–1069. Razi, M., & Tarn, J. (2003). An applied model for improving inventory management in erp systems. Logistics Information Management, 16 (2), 114–124. Ritchken, P., & Sankar, R. (1984). The effect of estimation risk in establishing safety stock levels in an inventory model. Journal of the Operational Research Society, 1091–1099. Sepehri, M. (1986). Just-in-time, not just in japan: case studies of american pioneers in jit implementation. APICS . Shoesmith, G., & Pinder, J. (2001). Potential inventory cost reductions using advanced time series forecasting techniques. Journal of the Operational Research Society, 1267–1275. Song, J. (2000). A note on assemble-to-order systems with batch ordering. Management Science, 46 (5), 739–43. Sox, C., Thomas, L., & McClain, J. (1997). Coordinating production and inventory to improve service. Management Science, 1189–1197. Tersine, R., & Tersine, M. (1990). Inventory reduction: preventive and corrective strategies. International Journal of Logistics Management, The, 1 (2), 17–24. Wallin, C., Rungtusanatham, M., & Rabinovich, E. (2006). What is the right inventory management approach for a purchased item? International Journal of Operations & Production Management, 26 (1), 50–68. Watson, R. (1987). The effects of demand-forecast fluctuations on customer service and inventory cost when demand is lumpy. Journal of the Operational Research Society, 75–82. Zhao, L., & Lau, H. (1992). Reducing inventory costs and choosing suppliers with order splitting. Journal of the Operational Research Society, 1003–1008.

59

Appendix A

Kardex system Figure A.1: Example of a Kardex storing system

60

Appendix B

Extraction of data from SAP This appendix details which information is extracted from the ERP-system. Figure B.1 shows a detailed overview of which tables from SAP are used. Figure B.1: Detailed information flow

Importantly, two different types of extraction from SAP can be identified in this image. One of these is the extraction of static data, while the other is the extraction of dynamic data. Some of the data to be used in the tool should be considered relatively static. These data include planned lead times - as indicated by suppliers-, the value of one component of a certain type

61

and planned safety stocks. While these figures may change over a longer period of time, they should be considered constant. When executing the main analysis tool, these data are therefore not extracted. Instead, when using the tool, decision makers should occasionally opt to refresh these static data. Especially in the case where significant changes to some of these static data have been made in SAP, managers should refresh the extracted static data. The system keeps track of when the last refresh was performed. Examples of cases where static data should be refreshed are when suppliers have been changed, which could result in different lead times, and when new components are added to inventory. The second type of extraction is that of dynamic data. These figures should be up to date at all times and are therefore extracted each and every time the system in ran. Examples of such data are current inventory balance and recent stock movements. Next, each piece of information extracted from SAP will described, according to which SAP-table it originates from. First, the static data retrieved from SAP originate from: • MARC (plant data for material): this table contains general information on all SKUs known in the ERP-system. For the purposes of this tool, the data retrieved are PN, lead time, safety stock, minimal order quantity and fixed lotsize. • MBEW (material valuation): this table registers total stocks for each SKU. The data extracted from SAP are PN, value and value per. Value per designates for which quantity the value field is listed. For instance, in some cases, value is not listed per component but instead per 100 components. • MLGN (material data for each warehouse number): another table that contains general information on all SKUs, however with a focus on more technical aspects. This table is used to identify whether or not a component is stored on tape. The data retrieved are PN and type, where type can either be 010 (non-tape) or 020 (tape). The dynamic data extracted from SAP originate from: • MBEW: this is the same table used in the extraction of static data. However, in case of executing the actual control tool, different data is extracted. The actual current inventory balances are retrieved from SAP. • MARDH (master storage location segment: history): SAP records the historic inventory balances for each month in this table. The data extracted are PN, date and stocks. Inventory balances are recorded on a monthly basis. • MKPF (header: material document): this table records transaction numbers and dates, among other data. These two metrics are extracted, the purpose of which will be explained later. • MSEG (document segment: material): this table contains transaction information. The data extracted are transaction number, PN and quantity.

62

Appendix C

Boundary conditions and filters • MARC: no filter is needed, since all SKUs known in the ERP-system should be included in the extraction. • MBEW (static): no filter is used, values of all SKUs are required. • MBEW (dynamic): no filter is used, active stocks for all SKUs are required. • MARDH: a filter is required that will ensure that only the most recent inventory balances are extracted. Balances are needed for thirteen months, from the last month until thirteen months before that. The balances in the MARDH table are recorded at the last day of each month. This means any analysis can only take place up to the last full month. Additionally, for some months and for some SKUs, no inventory balances appeared to have been recorded. It turned out that in this case no stock movements had occurred for that specific SKU in that specific month. Therefore, in order to ensure that for every SKU data is available for thirteen months back, two full years of balances are extracted for each SKU. This way, should a SKU not have moved for a significant amount of time (up to one whole year) at a particular time, by going back in history the correct balance for each month can still be found. The exact procedure used to fill in blank periods will be explained in the next subsection. • MLGN: no filter is required, the types of all SKUs have to be retrieved. • MKPG: this table is filtered to only include movements that occurred during the last twelve full months. • MSEG: the filter for the MSEG table (stock movements) is rather complicated. First for all, stock movements are filtered by type. Only types 101, 712 and 262 are retained. SAP uses a three-digit code to identify the type of movement that occurred. These three types include all movements that involve incoming deliveries. The table has to be filtered for incoming deliveries because these are used to determine the inventory turnover rate. Next, the remaining stock movements have to be filtered for the correct movement dates, which is during the last twelve full months. For this purpose, the range of movements extracted from the MKPG table is used, since the MSEG table does not include date information. The lowest and highest transaction numbers in the MKPG table associated with the relevant time span are fed to the MSEG query as boundary conditions, resulting in all incoming stock movements being obtained. 63

Appendix D

Service levels and service factors

Service level 50% 60% 65% 70% 75% 80% 85% 90% 91% 92%

Service factor 0 0.25 0.39 0.52 0.67 0.84 1.04 1.28 1.34 1.41

Service level 93% 94% 95% 96% 97% 98% 99% 99.5% 99.9% 99.99%

Service factor 1.48 1.55 1.64 1.75 1.88 2.05 2.33 2.58 3.09 3.72

Table D.1: Service levels and factors 64

65

Type 10 20 20 20 10

Type 10 20 20 20 10

PN 5090-0054-0033 1229-0006-7004 5900-0003-0007 3340-0002-0009 5070-0019-0010

PN 5090-0054-0033 1229-0006-7004 5900-0003-0007 3340-0002-0009 5070-0019-0010

Val PP 0,06 0,04 1,81 3,6579 15,95

Val PP 0,06 0,04 1,81 3,6579 15,95

MOQ 1000 1000 300 2000 200

RV 500 1000 300 2000 10

Stock 606 1643 116 1301 125

Value 36,36 65,72 209,96 4758,9 1993,8

# Used 479 9238 1434 4726 1251

# Bought 0 7741 1327 4267 1400

MOQ 1000 1000 300 2000 200

RV 500 1000 300 2000 10

Stock 606 1643 116 1301 125

Value 34,66 63,13 209,72 4758,97 1993,62

# Used 729 9557 1251 4726 1251

# Bought 0 7741 1327 4267 1400

Table E.2: Data for sample of SKUs obtained from SAP

LT 56 56 56 42 56

Table E.1: Data for sample of SKUs obtained from tool

LT 56 56 56 42 56

Current SS 0 0 0 0 0

Current SS 0 0 0 0 0

ERP and control tool data comparison

Appendix E

66 Theoretical % of SKUs 20% 30% 50%

Actual % of SKUs 9.81% 25.36% 64.8%

Table F.2: Conventional groups applied

Actual % of value 69.98% 25.02% 5%

Table F.1: Conventional definition of groups

Theoretical % of value 70% 25% 5%

Group A B C

Group A B C

Conventional ABC-classification applied to firm

Appendix F

67

PN 4900-0001-0003 4190-0003-0003 3900-0005-0003 8210-0006-0040 6618-1200-3402 2060-0001-0025 4100-0002-0021 4100-0002-0029 4110-0001-0004 5220-0001-0001

LT 360 180 720 720 56 360 360 182 182 360

# Used 62050 18834 35319 18 10140 917151 23877 180721 12118 34685

# Average 5085 350 5973 10 50 73290 2914 17380 2464 5712

Avg stock val 10660,19 1057 6689,76 4790 149 3664,5 2389,48 10428 9215,36 8510,88

StdDevDem 40295,45 25858,77 56342,82 95,49 5389,19 513403,5 30129,48 36978,02 5312,7 12436,67

Table G.1: Sample of problematic SKUs

Val used 130082 56878,7 39557,3 8622 30217,2 45857,6 19579,1 108433 45321,3 51680,7

Problematic SKUs

Appendix G

ITR 12,2 53,8 5,91 1,8 203 12,5 8,19 10,4 4,92 6,07

ROP 90% SL 112778 42387 141789 158 8454 1561743 62116 137445 12842 50129

Avg Inv 90% SL 52578 33599 72259 122 11896 658156 39166 48382 7098 16219

Avg Val 90% SL 110224,52 101468,98 80930,08 58438 35450,08 32907,8 32116,12 29029,2 26546,52 24166,31