MASTER THESIS Controlling pallet flows in order to decrease inventory record inaccuracy of packaging at Avebe U.A.

(1)

MASTER THESIS

Controlling pallet flows in order to decrease inventory

record inaccuracy of packaging at Avebe U.A.

Michel Huisman

s2807599 m.huisman.16@student.rug.nl 06-41661337 Supervisor Avebe Anna Paques MSc. anna.paques@avebe.com Supervisor RuG Dr. ir. W.H.M. Alsem w.h.m.alsem@rug.nl Co-assessor RuG Dr. N.D. van Foreest n.d.van.foreest@rug.nl

Faculty of Economics and Business MSc. Technology and Operations Management

(2)

1

Abstract

In inventory management principles it is important that actual inventory levels of items correspond to the quantities recorded in the ERP-system. Otherwise inventory replenishment systems can generate faults in the automatic ordering of inventory items. If there are discrepancies found between the inventory levels and the records this is called inventory record inaccuracy.

The higher the extent in which the records are not correct, the higher the chance that demands are not observed or that demands have to be backordered if possible. The chance of lost sales increases and that is not what a manufacturing company wants. This phenomenon is reasonably studied for main products of manufacturing companies. For packaging materials, the importance of accurate inventory records is not acknowledged yet. This is however needed because the main products are often connected with packaging materials which has influence on production and handling processes. By means of a case study it was investigated which processes have effect on inventory accuracy and in which extend. One process showed up as a significant contributor to inventory record inaccuracy. It appeared that outgoing packaging flows needed most attention, so the scope was moved to these flows.

For companies dealing with inventory management of packaging materials it was investigated what typical symptoms of inventory inaccuracy are and if these are also applicable to packaging items. Four inventory error types were considered as a result of various causes. On several areas a total of 30 causes were pointed out that lead to packaging inventory inaccuracy. Some of them were unknown in inventory management theory before, but were discovered at the case in relation to the handling of packaging materials. Further deepening of this subject at the case has been led to five root causes. A root cause analysis was performed and 2 root causes were selected to focus on further.

Three different compensation methods were distinguished: prevention, correction and integration. For these methods three existing solutions, which were provided by literature, were reviewed and assessed on usability. It however appeared that the selected root causes are of a different kind and are initially not aided with those solutions. Therefore, other solutions had to be developed to prevent the causes for appearing again. It was realized that it is important that business processes are well designed and that information flows correspond to those processes. Solutions were found in the redesign of the business processes according to the outgoing packaging flow. Their corresponding procedures had to be developed to support the processes and to give them shape. Preferably some attention had to be given to sustainability aspects as well.

(3)

2

List of abbreviations

BOM Bill of Material

BPMN Business Process Modelling and Notation ERP Enterprise Resource Planning (see IS)

FIFO First in First Out

FMCG Fast Moving Consumer Goods

FY Financial Year

IP Inventory Position

IR Inventory Records

IRI Inventory Record Inaccuracy

IS Information System

LSP Logistic Service Provider

MTO Make to Order

MTS Make to Stock

OIA Overall Inventory Accuracy

RFID Radio Frequency Identification

ROP Re-order Point

SAP ERP-system (see ERP)

SKU Stock Keeping Unit

SOP Standard Operating Procedure

STO Stock Transport Order

(6)

5

Introduction

This chapter introduces this report and the motive why this research was started. It starts with the problem setting, where the environment in which the problem occurs is described. Then a high level description of the problem and its size is given. This continues with the objective that is intended to be achieved with the study, the central research question and the initial scope that is applied to set the boundaries of the research.

1.1 Problem setting

In manufacturing companies there is almost always an inbound and an outbound flow of materials. Raw materials come in and finished products are going out. Worldwide several materials are used to facilitate handling, loading and transfer of large number of packages simultaneously. Examples of highly standardized packaging for transportation are containers and pallets (Dang and Chu 2015). Moving materials stacked on a stable platform, a pallet, makes handling and transportation of goods flexible and straightforward. Finished products on a pallet can go directly from production to the customer or to a warehouse first (Figure 1). At such storage locations an inventory is created. Products do not always leave the production location in the (packed) form how customers want to receive them. Therefore, at the warehouse products can eventually undergo logistic activities such as repacking or reloading. The picture in Figure 1 is typical for such situations and shows two points where trailers are being loaded for transport, one point for direct transport to customers (left) and one for loading after storage and/or repacking at the warehouse.

Figure 1.1: Material flows after leaving the production environment

(7)

6

possible. The chance of lost sales increases and that is not what a manufacturing company wants. Knowing this, they shall always try to keep their inventory record inaccuracy as low as possible. The terms ‘inventory inaccuracy’ and ‘inventory record inaccuracy’ (IRI) are both used for the same thing: a discrepancy of inventory level between the inventory records in the information system and the real available stock. In this research both terms are used interchangeably. Inventory inaccuracy can have various reasons. The root causes of inventory inaccuracy and their influence on the performance of inventory systems are studied by several people in the last ten years (Rekik 2011). Rekik (2011) classified various main error types under which inventory inaccuracy can be generated, these were all proven by his study. To get inventory inaccuracy in control, several compensation methods were described by Uçkun and Canan (2008). That the awareness of inventory inaccuracy is important, is presented by the examples described by DeHoratius and Raman (2008). They show that decision support tools such as automated replenishment systems could fail because of inventory inaccuracy.

It appears that inventory inaccuracy has been studied widely for core-products (products which are the main selling items) but not for packaging materials. However, managing the inventories of both items are not necessarily identical. And that while packaging can significantly affect the efficiency of logistic systems and activities such as manufacturing, distribution, storage and handling throughout the supply chain (Dang and Chu 2015). That said, packaging has in contrast with core-products the following characteristics:

 Packaging is a by-product; a secondary product derived by a (logistic) production process.  Is presenting considerable less value than a core-product.

 It is often considered a consumable item, which is written off and disposed after use. Inventory record management of packaging is however needed because:

 Availability and accuracy of packaging inventory is a prerequisite for optimal flow control of core-products because the two are highly connected.

 Lots of packaging may still represent an appreciable value.

 Inefficient packaging flows require more manpower or other resources than is justified. In this study a method is going to be developed to improve the accuracy in inventory records of packaging materials. It will be investigated if it has the same causes and if known compensation methods for core-products are also applicable for packaging materials. The research is managed to find eventual differences and which (known) methods can be applied to prevent the forming of inventory record differences. This will help companies to improve current processes so that the information system’s inventory records related to packaging materials equals the physical inventory. The objective of this study is to develop a business process for optimal pallet management. This will be done by performing a case study, underpinned by literature.

(8)

7 performed in which causes of IRI are identified and analyzed. Subsequently in chapter 5 potential solutions are developed for the main problems that were selected and rated on effort and impact. Then recommendations are given in chapter 6. Chapter 7 contains a discussion in which critical aspects of the research are reflected. The report finishes with conclusion statements and the answering of the main research question.

1.2 Problem description of the case

The case was found in the potato starch factory Avebe U.A. (hereafter Avebe), which struggles with inventory inaccuracy of their packaging materials, specifically pallets. Discrepancies are observed between the pallet stocks as administered in the Enterprise Resource Planning (ERP) system and the actually present pallet stock. A discrepancy can be both positive or negative. Sometimes there are less pallets present compared to the records and sometimes there are more. It can be seen that at some sites the differences are mostly positive and at other sites the differences are mostly negative. Other locations exhibit both differences. For effective inventory record management every physical action related to materials should be go together with an administrative action. When for example pallets are used for production, this usage should be recorded in the system. Otherwise a difference in physical stock and recorded inventory will be immediately created. Next to this first example causing inaccuracy, there are a lot more known causes of IRI which will be presented in the theoretical background section (Chapter 3). A more detailed profile of the case will be described in the chapter case study (section 4.1).

At some sites of Avebe physical inventory is counted regularly to check the actual pallet inventory. Other locations do not maintain any registration of inbound or outbound pallet flows at all. This lack of registration leads to incorrect inventory levels in the ERP-system. In the last five years, an average of 21.000 pallets were written-off each year without knowing where these pallets had exactly gone to. In Figure 1.2 the net pallet stock discrepancy of all Dutch locations is given per year and also its represented value. For the years 2014 and later, also the positive stock discrepancies were taken into account. This turned the negative net discrepancy into a positive net discrepancy. That is why Figure 1.2 shows three years of less pallets and two years of more pallets than recorded.

Figure 1.2: Measured discrepancy in pallet stock with their values

€ -300.000 € -200.000 € -100.000 € -€ 100.000 € 200.000 € 300.000 -50000 -40000 -30000 -20000 -10000 0 10000 20000 30000 40000 50000 2011 2012 2013 2014 2015 Num be r of p all ets Year

Net pallet stock discrepancy per year

(9)

8

In fact, this gives a distorted picture because discrepancy in both ways is undesirable. If the absolute value of both differences is taken, the accumulated inventory inaccuracy becomes even worse (Figure 1.3). In 2015, an absolute discrepancy of almost 90.000 (-28.000 + 60.000 = +32.000 net) items was observed according to financial documents. On the total of 370.000 newly bought pallets this gives an inventory accuracy error of 24%, with a total value of €600.000.

Figure 1.3: Accumulated pallet stock discrepancy

Gumrukcu (2008) suggests to measure the percentage of packaging materials that are correctly recorded in the ERP-system (OIA) by the formula in Equation 1.1.

Equation 1.1: Calculation of overall inventory accuracy

𝑂𝑣𝑒𝑟𝑎𝑙𝑙 𝑖𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = 𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑎𝑐𝑐𝑢𝑟𝑎𝑡𝑒 𝑆𝐾𝑈 𝑟𝑒𝑐𝑜𝑟𝑑𝑠

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑆𝐾𝑈′𝑠 𝑡ℎ𝑎𝑡 𝑎𝑟𝑒 𝑐ℎ𝑒𝑐𝑘𝑒𝑑 ∗ 100%

For Avebe the OIA could not be calculated yet because most of the materials are not checked. However, the high error ratio gives Avebe enough motive to cooperate in an investigation to find out where and why these inventory inaccuracies exist. Apart from the discrepancies in pallet inventory records, there is another reason why solving these errors is of interest. Namely in the ERP-system a value is assigned to all inventory stock keeping units (SKU’s), expressed in money. If inventory quantities are not correct this has also influence on the inventory values. In that case at the end of the year not only inventory quantities have to be corrected, but also pallets values have to be written off representing a sudden financial loss. The occurrence of inventory errors in this case, makes it very suitable for this research. In return, possible solutions can be brought to decrease their IRI. In chapter 4 the case will further be analyzed.

1.3 Research question

In this research it will be investigated how pallet flows within Avebe currently look like and which actions are or are not accompanied by an administrative activity. Since pallets are a form of packaging existing inventory control strategies will be checked if they are still relevant if used in

€ -€ 100.000 € 200.000 € 300.000 € 400.000 € 500.000 € 600.000 € 700.000 0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000 2011 2012 2013 2014 2015 Num be r of p all ets Year

Pallet stock discrepancy per year

(10)

9 packaging inventory. Subsequently an advice is written on how to improve the inventory management of pallets. This includes physical flows, information flows and the responsibilily for each action that should be taken. This altogether must form a business procedure on which anyone can fall back if processes are threatened to go wrong. In the end this must give Avebe more control over their pallet flows and therefore it improves their packaging inventory system.

Hence the research question for this research is: “How can packaging material flows be controlled so that there will be less differences in stock between the ERP-system records and the physically available stock of packaging materials?”

1.4 Scope, limitations and preconditions

Avebe has several production- and storage locations in The Netherlands. This also includes the locations of external parties, the Logistic Service Providers. In Table 1.1 the locations and their type are displayed.

Table 1.1: Production- and storage locations of Avebe in the Netherlands

Name Place Type

Avebe Foxhol Foxhol Production location

Avebe GNV Gasselternijveen Production location

Avebe TAK Ter Apelkanaal Production location

Teuben-TAK Ter Apelkanaal Logistic Service Provider

Teuben Ter Apel Logistic Service Provider

Navoloodsen Ter Apel Logistic Service Provider

Houtunie Veendam Logistic Service Provider

Given the limited available time for this thesis it is not possible to investigate each individual location. A focus will be placed on the problems that have the highest potential profitability when they are removed. After an introductory research of pallet flows at these locations the scope of the project has become the production location in Ter Apelkanaal, in short: Avebe TAK. The reason for this is that at this location the pallet flow is the most complicated and contains several actions. Some of these actions are also executed at other locations but Avebe TAK covers all of them.

(11)

10

Theoretical framework

This section provides a review of background theories that were found in literature and are related to inventory management. It starts with general information about packaging (2.1). Then inventory- and information management principles are discussed (2.2-2.3), followed by an extensive explanation of IRI related to normal stocks with its symptoms, error types, compensation methods and solutions (2.4-2.5). In 2.6 the focus returns to packaging, where the little existent knowledge in relation to IRI is discussed. The last theoretical concept can be found in 2.7.2, where detail levels of solution implementation are shown. These theories will be used to get more knowledge about what have been studied by others so far. It is also used to develop a research framework (2.8) that will be used for treatment of the research problem stated in 1.3. The research design (Chapter 3) includes a methodology section (3.1) where the approach of this research will be explained.

2.1 Packaging

Almost all products that leave a factory in their way to the customer have some sort of packaging. Packaging can be defined as “the technology of enclosing or protecting products for distribution, storage, sale, and use” (Soroka 1999). This research emphasizes on packaging used for production, transport and warehousing. Since the pallet is one of the most important packaging materials used at the studied case, the pallet is further elaborated on here.

(12)

11 Packaging involves three levels: primary, secondary or tertiary (Hellström and Saghir 2007; Dang and Chu 2015). Primary packaging has direct contact with the product, and the secondary packaging contains several primary packages, such as a box. Several primary or secondary assembly packages on a pallet or a roll container is called tertiary packaging. Products on pallets can thus be classified as tertiary packaging, see also Figure 2.1.

Figure 2.1: Three levels of packaging (Mistry, 2013)

2.2 Inventory management principles

Even using the simplest inventory models can cause IRI. Before telling how this can happen, a short explanation of inventory replenishment systems is outlined. When material inventories are used they should also be replenished somehow. Two questions are always subject of decision then: when to order and how much to order. Roughly there are two main types of inventory replenishment policies. The first is a Fixed-Order Quantity System, or (R,Q) model. R stands for the reorder point; at this time an order quantity of Q items will be ordered. The reorder point is a predetermined level of inventory items. When the inventory level reaches the reorder point, an order of Q items will be placed. The second replenishment type is a Fixed-Time period system (s,S). In this model inventory levels will be checked in fixed time intervals. The system wants to maintain a maximum inventory level of S. The inventory will therefore be restocked till level S by ordering a quantity of S-s items. The ordered quantity thus depends on the inventory level s at the moment of checking. In Figure 2.2 both systems are visualized as a function of time.

Figure 2.2: Left: Fixed-Order Quantity system. Right: Fixed-Time Period system

(13)

12

What these inventory policies have in common is that the inventories in the system are always assumed to be perfectly recorded. Inventory inaccuracy can make automatic ordering unreliable. Raman et al. (2001) said that IRI may undermine decision support tools such as automated replenishment and automated demand forecasting systems that do not account for the inventory uncertainties. Systems will order when unnecessary or fail to order when they should. In the worst case, a state called “freezing” is reached (DeHoratius et al. 2008). This scenario, in which a business has no items on the shelf (so no sales and/or production), but have a sufficient positive inventory record (so no replenishment), results in a lasting physical stock-out. Bensoussan et al. (2007) add that if an inventory manager does not observe sales, or does only observe the inventory level in the records, this frozen state will not be discovered.

2.3 Inventory information management

Many companies have automated their inventory management processes and now rely on information systems when making critical decisions. To support organizational control of these processes ERP-systems are used. Since the 1990s the use of ERP-systems in organizations experienced a rapid growth (Maas et al. 2014). They have the ability to standardize and integrate processes. ERP-systems stimulate organizational structure through forcing users to follow prescribed procedures (Maas et al. 2014). Since an ERP-system is almost always a packaged software solution instead of a custom-made system, it is essential to find a match between the ERP-system and the organization’s business processes (Luo and Strong 2004). This can be done by customizing the ERP software (technical customization) and/or the business processes (process customization). If these two do not match, errors will most likely occur in the management of inventory.

Information management of items can be managed at different levels: product type level, product batch level and product individual level (Kärkkäinen 2001). This classification is often used for Activity-Based Costing purposes. The costs associated with one of the three categories can easily be assigned to products that are part of these categories. But this classification of activities can easily be used for other forms of item management, such as inventory tracking. Factors that can influence the use of one type of product level management is the value of the item, the criticality of the item to the functionality of the process it belongs to. Also the length of the products lifetime, complexity of the system the item is attached to and external requirements can be factors that influence product individual level management (Kärkkäinen 2001). This can be made clear through a car example. Managing a product type level threatens all cars of the same brand and type as one. Managing at product batch level can mean that all cars of the same type produced on one day are considered as one. When managing information at an individual level, all individual products can have different characteristics. The product often gets a unique product identifier (e.g. serial or license plate number). The unique characteristics do not need to be specifically created by the manufacturing process, but can also be created later on. For example, all cars of a batch are exactly the same after the manufacturing process, but gets unique at the moment that a license plate is assigned.

(14)

13

2.4 Inventory inaccuracy

Inventory inaccuracy is a main issue in businesses dealing with physical assets (Fleisch and Tellkamp 2005). It was explained by Iglehart and Morey (1972) as an error in the inventory stock record whereby the stock record quantity is not in agreement with the physical stock quantity. This was later confirmed by DeHoratius et al. (2008), Rekik (2011) and many others. Inventory inaccuracy occurs when the inventory record (what is available according to the IS) does not match the physical inventory (what is really available). This is known as a substantial problem in the retailing industry. DeHoratius and Raman (2008) found inaccuracies in 65% of the nearly 370.000 inventory records examined across 37 retail stores. And that while the study was conducted at a large public retailer with highly modern operations including electronic sales scanning and automated replenishment. As direct impact of these inaccuracies, lost sales is the most called effect in this industry. Inventory record errors can have several causes, often through human actions (Nachtmann 2010). The causes that are often mentioned are transaction errors, misplacement errors, shrinkage, product quality errors and supply processes. Before these causes will be explained in detail, first is started with the symptoms to which inventory inaccuracy can be recognized.

General inventory inaccuracy symptoms

Inventory inaccuracy is a well-known appearance in operations management literature. Some typical symptoms of inventory accuracy problems are (Miller 1997):

 Lots of inventory errors; faulty location, material, quantity.

 No trust in the records; numerous calls to check on availability on floor.  The system fails to order when it should.

 Unwanted financial reporting surprises.

 Lost materials due to unknown causes (Kang and Gershwin 2004).  High backorder penalties or lost sales (Kök and Shang 2006).  Ineffective inventory ordering decisions (Kök and Shang 2006).

DeHoratius et al. (2008) suggest that even small discrepancies can result in substantial lost sales, missed service level targets, and suboptimal business performance. If someone wants to find out if his business is affected by IRI or just wants to check the magnitude of the problem, literature prescribes several inventory- and stock keeping quantitative indicators. IRI, measured in absolute difference between the recorded and actual inventory quantity, is associated with factors such as an item’s annual selling quantity, its cost, and the frequency of an audit (defined as the number of days since the previous physical audit). IRI is also associated with factors such as inventory density, defined as the total number of units found in a storage area, and product variety, defined as the number of different material types within a warehouse (DeHoratius and Raman 2008).

General inventory inaccuracy causes and error types

(15)

14

cause activates a certain error. To make the flood of different causes more structured, four main error types that generate inventory inaccuracy were classified (Bensoussan et al. 2007; Rekik 2011):

 Transaction errors.  Misplacement errors.

 Shrinkage, spoilage and theft.  Product quality and yield.

Even though ERP and inventory management software made collecting and storing data easier and cheaper (Luo and Strong 2004), data collected about flows of SKU’s often contains errors. During inventory transactions unintentional mistakes can happen on a regular basis. Think of inventory receiving, counting or checking out. An example that is given by Bensoussan et al. (2007) describes a supermarket with two different tastes of soup. The customer buys one soup of each taste but the cashier scans only one and doubles the quantity. An error was created. Furthermore, in process industries where units are not discrete, it may be difficult to do exact measurements in liters, kilo’s or other units. In the case of packaging, errors can occur when units are sold separately from a box or pallet where the whole box or pallet with a half load is left behind. The tertiary packaging material cannot be split in the records so it can’t be booked correctly.

If some of the inventory is misplaced, it is not available for use till it is located. Misplacement is more likely when multiple storage locations are used or exact item locations are dynamic instead of fixed. Misplacement can be seen in two ways, physical misplacement and electronic misplacement. Physical misplacement occurs when items are stored at the wrong shelf or storage location. Electronic misplacement occurs when items are placed wrong in the information system.

Items can naturally lose their properties while they are held in inventory. Examples are drugs, chemicals and food products which have an expiration date. Some products can have a limited lifetime (seasonal products) or their quality is going down in time (fruits). Also damaging of items by customers or employees, such as tearing a package or scratching the item while trying cause shrinkage. Damaging of items lead to a situation where the inventory consists of both good and defective products. Defective products cannot be used or sold as new, and should therefore not be counted in inventory records anymore. Or even worse, people can steal items from the shop floor. As long the shrinkage, spoilage or theft is not detected and processed in the information system, the inventory record is not accurate. Without noticing the information system still counts on the old inventory levels.

(16)

15

General reactions to inventory inaccuracy

To get inventory inaccuracy in control, multiple compensation methods could be used. DeHoratius et al. (2008) mention that there are three main ways that can be used to respond to inventory inaccuracy. These ways agree with those described by Uçkun and Canan (2008). The methods will be explained here shortly:

 Prevention

Through the implementation of process improvement the causes of inventory inaccuracy are reduced or even eliminated. If processes are controlled so well that discrepancies cannot occur anymore, there will be no more IRI. In practice this is unfortunately a utopia. It should always be combined with another compensation method, or the minor inaccuracies that remain must be taken for granted.

 Correction

By auditing policies such as periodical review of inventory existing inventory discrepancies are identified and corrected. Most companies have already implemented a cycle-count program (Kök and Shang 2006).

 Integration

This way is not directly about reducing inventory inaccuracy, but more like living with it so that the negative consequences are not felt. Since IRI cannot be banned for 100%, ways are found to make inventory planning and decision tools robust enough to account for the presence of inventory inaccuracy.

Concrete solutions that could be applied as part of these compensation methods are discussed in the next section, paragraph 2.5.

2.5 Solutions to inventory inaccuracy provided by literature

Solution methods for decreasing IRI that are discussed in literature yet will be presented in this section.

Product identification and tracking

A technique that has often been studied in literature and helps in preventing inventory inaccuracy is product identification and tracking. Product identification is important for multiple reasons. First it is used for checking if you have the right product in front of you. Second, if machines can read the identity of the passed item, processes can be automated and gains in efficiency and error reduction can be obtained (Kärkkäinen et al. 2001). Third, product identification helps tracking items anywhere throughout the supply chain (Uçkun 2008).

In many applications, automatic identification is used to increase process effectiveness. Automatic identification is used to automate processes to save time and reduce errors (Kärkkäinen et al. 2001). Automatic data handling applications are commonly used especially in warehouses and retail outlets. In these situations, this technology makes inventory bookkeeping more easily and accurate. There are more than a few alternatives by which items can be identified:

 Bar-code technologies

(17)

16

Bar-code technologies and RFID are entitled the most common for inventory automating processes (Kärkkäinen et al. 2001; Uçkun 2008). Furthermore, there are a lot of other technologies by which items can be recognized, all with their own precision and advantages. Some examples are camera vision recognition, magnetic ink, smart cards, Bluetooth and GPS. These are mostly expensive solutions. Because the one-time-use policy that is used for packaging, these solutions are not further considered. RFID technology is more advanced than a barcode. RFID has two advantages compared to a bar-code: there is no need for line of sight and RFID tags have unique codes. However, this solution is known for its higher investment costs (Uçkun 2008). De Kok et al. (2008) found that the break-even prices are highly correlated with the value of the items that can be lost in the records (the higher the value, the sooner break-even).

Cycle counting

The objective of cycle counting is to ensure that production and inventory control systems use records that are accurate (Wilson 1995). After a cycle count the inventory records that contain errors are corrected. Cycle counting is thus a form of correction. Although applying cycle counting can decrease IRI, it can also lead to additional cost (Gumrukcu 2008). This must be taken into account when considering implementation of cycle counting. The effectiveness of this method varies on the characteristics of the item types. Gumrucku (2008) examined two general kinds of item types which commonly exist: high-demand-low-cost items and low-demand-high-cost items. Cycle counting is particularly essential for low-demand-cost items. Gumrukcu (2008) also states that also for high-demand-low-cost item types, cycle counting can be an unnecessary effort.

An opportunity of cycle counting is that it can also have a learning part. If cycle counting is performed periodically and at one time it appears that there is a big difference in accuracy, it is easier to count back to the moment where this difference probably arose. If the cause could be identified, it is possible to fix this cause and prevent it from happening again. The downside of this method is that miscounts can occur. Often it is assumed that an inventory count completely eliminates any discrepancy between the recorded and physical stock. However, in practice large errors remain in stock records because of inaccuracies in the counting procedure (Iglehart and Morey 1972). Furthermore, transactions occurring during the count make it extremely difficult and costly to ensure complete accuracy.

Cycle counting can be implemented in different ways. At specific times one can take a sample survey in which some records are compared against actual stocks. This is a cost efficient method, where or example only A-class items are checked. Another way is to count the whole inventory, and compare this against inventory records. These two can also be combined, where a sample check is taken weekly and a full comparison once a quarter or once a year.

Bayesian updating of inventory records

(18)

17 Maintaining a Bayesian inventory record consists of multiple parameters. The first is the inventory level (It-1), the amount of stock after the last period. Next to this sale demands (St) are fulfilled and

replenishments (Rt) arrive. Lastly an invisible demand-factor (Ut) arrives that represents the stock

losses. The Bayesian inventory record level (It) is updated according the formula in Equation 2.1.

Equation 2.1: Bayesian inventory level quantity update

𝐼_𝑡 = 𝐼_𝑡−1− 𝑆_𝑡+ 𝑅_𝑡− 𝑈_𝑡

Invisible demand (Ut) is allowed to be negative here. Subsequently the inventory manager makes a

replenishment order decision with order quantity Rt+1+L in which L stands for the lead time. The order,

that can serve both a Fixed-order quantity or a Fixed-time period policy, must be extended by the uncertainty quantity (Ut). This is also the hardest part of this approach. It is influenced by aspects

including inventory level, total number of demands since last count and elapsed time since last count (Iglehart and Morey 1972). The total IRI quantity is the addition of the errors that arose during normal operating in the period since the last inventory count plus any residual error from the last count.

Maintaining a Bayesian inventory record is rather complex but simulation studies of DeHoratius et al. (2008) show that it is a legit alternative for other compensation methods. It still remains a challenge to develop effective replenishment and audit policies based on it. Therefore, organizations may be just satisfied with choosing to buffer inaccuracy uncertainty with some additional inventory (Iglehart and Morey 1972). This can be seen as extra safety stock meant to moderate the invisible demand-factor Ut. Bayesian updating must however always be seen as an addition to cycle counting.

2.6 Packaging inventory management

Till now IRI was particularly discussed related to inventory management in general. So now the question is, how could this all be applied to packaging? Can the same methods be used to deal with inventory inaccuracy of packaging? Packaging can significantly affect the efficiency of logistic systems and activities (Dang and Chu 2015). But is it different to known inventory management principles or does it have a lot in common? The aim here is to understand where and how packaging decisions might impact the inventory system and inventory inaccuracy. Yet, Rekik (2011) found that literature on inventory models has rarely differentiated between inventory records and physical inventory, especially not for packaging materials. This is in line with DeHoratius and Raman (2008) who said that most traditional inventory models do not take inventory inaccuracy into account.

(19)

18

2.7 General solution implementation approaches

Additionally, two general methods for the implementation of the developed solutions are handed in this paragraph. These two will be combined to make sure that the proposed solutions can be structurally implemented in the right part of the company.

Standardization

Standardization is a tool from the lean philosophy that ensures that improvements will be sustained. Without standardization, all improvements will roll back after some time. Standardizing work adds discipline to the culture, an element that is sometimes neglected but very crucial for establishing smooth work procedures. The idea of standardized work is that the current best practice is documented and that this is the baseline for continuous improvement (Filius and Heidema 2014). If the quality of work is decreasing, processes should go back to the standards. Standardized work contains instructions for the practical execution of tasks.

1. What is the procedure?

2. Who should execute the procedure? 3. When should someone do it? 4. How should someone do it? 5. What are the expected results?

This method of standardization can be used to document standardized processes and procedures. It is applicable for all three compensations methods discussed in section 2.4.3 and for the business system described in section 2.7.2.

Solution implementation levels

The various possible solutions that help in decreasing inventory inaccuracy are often meant to be applied in a specific part of the business. Sometimes whole business processes have to be changed while at another moment only a few changes in job execution is needed. Therefore, the four layers of a business system are here discussed. Understanding the business system is also usable for improving inventory management. It can give some guidance in the implementation of final solutions.

(20)

19

Figure 2.3: Four layers of a business system

(21)

20

2.8 Research framework

In this paragraph sub-questions of this research will be presented. The sub-questions are formed based on initial knowledge from a first global study on the case and the relevant literature that was found before. After that a conceptual model is developed to show how theoretical concepts are probably related to inventory inaccuracy.

Sub-questions

To get the main question answered it will be split up in parts. By answering each sub-question a part of the main question will be covered. A list of underlying questions is given below, partly inspired by the theoretical background section.

1. How is the pallet flow within Avebe TAK currently organized? 2. What physical actions are executed related to empty pallets?

3. What administrative actions are executed related to empty pallets and who is responsible for them?

4. How is the inventory of new and used pallets currently managed? 5. In what ways can packaging flows be physically managed?

6. In what ways can packaging flows be managed within an ERP-system and what is the expected accuracy?

7. Which method is the recommended for Avebe to manage pallet flows to fit physical and information flows?

8. How can this method be applied to other locations / organizations with similar problems?

Conceptual models

This research contains an investigation into the root causes of inventory inaccuracy of pallets within Avebe TAK and the design of a method to reduce these. The contributing factors found in literature are expected to have effect on this phenomenon and are drawn in a conceptual model (Figure 2.4). These factors will be investigated further whether they have influence on the problem. It will also be checked if there are possibly more factors in this case study that contribute to inventory inaccuracy.

(22)

21 It is expected that all main factors, colored in orange, lead to a discrepancy in pallet inventory. A discrepancy leads subsequently to inventory inaccuracy. By handling these factors according to the business procedures, the inventory inaccuracy can be controlled and decreased.

The solution concepts from literature that likely contribute to the decrease of IRI of packaging can also be modeled. This is shown in Figure 2.5. There are three compensation methods distinguished. For each method a solution is discussed. Solutions are then standardized by determining their properties and parameters which are used during implementation.

Compensation methods Solutions Business System Level Prevention Correction Integration Identification & tracking Cycle counting Bayesian updating Policy Process Procedure Work instructions Standardization In cr ea si ng le ve l o f de ta il What Who When How Results Start

Figure 2.5: Solution concepts

2.9 Chapter summary

(23)

22

Research design

This chapter describes the followed research method. It contains approaches found in literature that were used to conduct the research. Also a sub-question answering plan is attached and some sources of data that were used.

3.1 Methodology

Avebe has started an improvement project which is part of their Six Sigma quality program, which strives for quality near perfection. In this specific project about pallet inventory inaccuracy a team is investigating what possible problem-solving methods exist that can be applied. To begin with, the Six Sigma quality program makes use of measurement based sub-methods under which DMAIC (Six Sigma Institute 2015). The five phases of a DMAIC project are shown in Figure 3.1. It is an improvement system for existing processes that score below target and are candidates for improvement. The DMAIC method in Six Sigma is often described as a suitable approach for problem solving (De Mast 2012).

Figure 3.1: Five phases of Six Sigma’s DMAIC

In the umbrella project, of which this thesis research is part of, the DMAIC approach is already chosen. This thesis project will be performed during the define, measure and analyze phases of the overall project. The focus of this sub-project is on these first three phases only and does not contain implementation. After the third phase an advice is given how to structure the processes of pallet flows so that inventory differences will be minimized. The company then decides when and how the implementation will be realized. This will be executed in the ‘improve’ phase of the pallet project according to the given advice. The ‘control’ phase is also the responsibility of the company itself, but also here an advice for approach will be given. An analysis of Six Sigma shows that it has similarities to other quality programs like Total Quality Management or Lean Manufacturing, but that it is different in scope and complexity (Walters 2005). Six Sigma contains comparable ideas and philosophies of other quality programs, but with a practical point of view. Six Sigma’s DMAIC is therefore confirmed by the researcher as a valid approach for this case study. A short explanation of how each of the first three phases will be approached is given now.

(24)

23

Figure 3.2: Fishbone diagram (Duffy et al. 2012)

Building the diagram always starts with the final effect or ‘the problem’ attached to a horizontal line. In this case the problem would be the inventory record inaccuracy of empty pallets. Connected to the problem are six main cause-categories. These are intended to create a first focus and giving inspiration while brainstorming about possible causes. Subsequently the generated causes are connected to the main categories they are related to. In practice a set of cause-categories can be chosen depending on the industry the fishbone diagram is made for. The set that is often used in manufacturing industries is the ‘6M’ set. The six M’s stand for machine, method, materials, measurement, man power and mother nature (environment) (Fishbone (Ishikawa) Diagram 2016). Additional causes may be found with some internal observation. The conceptual model already showed some possible error types of inventory inaccuracy, which can be assign to the causes.

Since problems are almost always the result of multiple sequential causes, there is a need to systematically find these underlying causes. In Six Sigma there is a technique available to identify the root cause of a problem. It is called the ‘5 whys’ technique. This technique determines the root cause of a problem by repeating the question “Why?” five times (Duffy et al. 2012). Why does this problem exist? Why does it happen? Each cause on the Fishbone diagram is further analyzed to determine if a more fundamental cause can be found. The last answer is considered to be the root cause. Serrat (2009) described that for finding the root cause of a problem the ‘why’ question has to be asked for an average of five times, but in some cases this point is reached earlier. After this it is clear how the pallets are flowing through the processes and what the associated problems are. The define phase is thereby finished.

The next phase exists of measuring of the flows and other parameters such as inventory levels. Another tool named ‘root cause analysis’ will be used to quantify every single root cause. A value will be assigned to indicators that measure processes, actions, flows and inventories. The goal is to find out what the size of the problems is, how important various causes are and how many times it leads to errors. This information is likely to be retrieved from the ERP-system and validated by employees of the company. Also effects will be analyzed and are given a weight of importance.

(25)

24

The sub-questions that were defined in section 2.8.1 are going to be answered with help of a few data sources. Primary sources of data will be observation of what is done in practice at the physical locations. Also meetings, discussions with operators, inventory managers and other employees are a primary data source. Another primary source is the companies’ ERP-system, where exact numbers can be found of pallet flows per moment in time, if recorded. Which source is going to be used depends on the main fishbone category of the problem. If we for example look for causes or solutions to a ‘material’ or ‘machine’ related topic we could ask the logistic employees who work with this equipment every day or take a tour through the facility. If the related category is ‘measurement’ we could look into the ERP-system for mutations and analyze these. Table 3.1 gives an overview of the sub-questions and the data sources that are selected to answer this question. The method of how these data sources will be used is shown in column three of this table.

Table 3.1: Sub-question answering plan

Sub-question Data sources How to obtain?

1 Employees with several functions Observations, guided tour at locations

2 Logistic employees, information system Conversations, exploring SAP

3 Logistic employees, Information system Conversations, exploring SAP

4 Logistic employees, Information system Conversations, exploring SAP

5 Literature Search terms:

- By-products - ERP stock - Intern transport - Inventory balance - Inventory control - Inventory inaccuracy - Inventory management - Inventory matching - Inventory policies - Inventory reliability - Inventory valuation - Material management - Packaging management - Pallets - Production packaging - Supply management - Traceability

And some combinations of these terms. These terms are used in Google Scholar and Web of Science to find relevant articles in relevant journals (e.g. Journal

of Operations Management and

Information Systems Research). Also backward and forward searches are performed.

6 Literature, ICT-department For search terms see previous question.

For aimed practical solutions meetings were held with ICT-employees.

7 Project team Potential solutions will be presented to

(26)

25

8 Background section in combination with

conclusions

General and case specific characteristics are known at the end of the research. From this generalizability can be evaluated.

An extensive search was performed to find literature about cases with sort-like problems. However there are a few examples of handling storage and production of inventory inaccuracy in general, no examples were found that gave a direct business procedure for handling packaging inventory inaccuracy. This again indicates that the development of procedures for preventing and compensating IRI can be of value, especially for non-core business products.

After that we have found potential solutions for the major problems then the recommended solutions will be rated in an ‘Effort / Impact matrix’ (Figure 3.3). The location of the solution in this matrix indicates which solution implementation will generate the most added value for the business and for what price.

Figure 3.3: Effort / impact matrix

Lastly, a regularly planned team meeting for pallet project ‘Woodstock’ was also attended by the researcher. This team was used to brainstorm and talk about analyzations, results and possible solutions related to the pallets flows at all locations of Avebe. The associated roles in the team can be found in Table 3.2.

Table 3.2: Associated roles in project team 'Woodstock'

Role in project

Project lead ‘Woodstock’ Administrator

ICT employee Financial employee

(27)

26

Case study

This chapter describes the case study that was performed in order to study the research topic. First some general information is given about the case. Hereafter the pallet inventory system is analyzed and some observations are made related to logistic activities. Then a problem analysis is performed where critical processes and root causes will be identified. The case study will end with some conditions that must be fulfilled by the solution.

4.1 About Avebe U.A.

The cooperative Avebe U.A. is a worldwide operating potato starch manufacturer located in the province of Groningen, Netherlands and was founded in 1919. It produces starch products based on potato starch and potato protein that is used in consumer food, animal feeds, paper, construction work (glue) and textile. Avebe is now a cooperation with over 2500 members and worldwide the largest producer of potato starch and potato starch derivatives (Avebe U.A. 2015). Avebe has several production sites in Europe: three in the Netherlands, two in Germany, one in Sweden and one in Turkey. Avebe’s products are sold to customers all over the world. At the production sites starch can be packed in three ways: in bulk (no packaging needed), in sacks of around 20-25 kilos, or in big bags (with weights up to 1250kgs). Sacks and big bags are filled with powder at the production lines and are then placed on a pallet to make handling easier. In The Netherlands alone there are about 15 different pallets in use.

Production processes are arranged in such a way that all finished packaged products leave the production line on a disposable pallet at first. After production logistic activities are taking place before the products are going to the customer. These actions are sometimes executed at Avebe’s own production locations but can also be outsourced to a few logistic service providers (LSP’s). At these parties only logistic activities and storage takes place, no production. For some of the logistic actions pallets become physically available and should subsequently be recorded as stock in the ERP-system. In total about 370.000 new pallets are purchased over a year.

4.2 Analysis of Avebe’s pallet inventory system

(28)

27

Figure 4.1: High level logistics process of pallets at Avebe

In Figure 4.2 a high level view of the empty pallet flow of Avebe is shown. Production, logistic activities as well collecting occurs at multiple locations. At Avebe TAK, there are multiple production inventories containing empty pallets, meant for the different production lines. Pallets can be moved sometimes between these locations. Next to the pallets that are used in production, extra empty pallets are needed for some logistic actions. These logistic actions can also yield empty pallets which are going to a pallet inventory for used empty pallets. Used empty pallets will be collected and dispatched.

Figure 4.2: A high level empty pallet flow of Avebe TAK

Avebe, which is historically an organization with a cost-leadership strategy, uses an open pallet system. This means that most pallets that leave the factory with finished products are not retrieved by the selling company after the product has been delivered. They are sold as part of the final product. The literature section also discussed two levels of information management relevant for packaging (2.3). The type that is now used within Avebe can be affirmed as the product type level. This is seen by the fact that at Avebe, the only differentiation between packaging is made based on material type. One particular packaging unit cannot be distinguished from other items of the same type. The only thing that is measured in the ERP-system is the available quantity of each material and sometimes the use. Also no form of product identification (2.5.1) is used for recognition or tracking of packaging1_.

1_{For finished goods they make use of barcode scanners attached to forklifts. Only loaded pallets have a}

(29)

28

Logistic activities

There are several logistic activities that have influence on the empty pallet inventories. The activities that were found are explained down here. Except of the last one, these actions are sometimes called logistic production.

 Loose loading

Loose loading is an action that occurs if shipping of products is done without the underlying pallets. This will be for example done when a sea container has to be loaded with as much products as possible or when a customer explicitly asks for it.

 Change of pallet

Sometimes the pallet under finished products has to be changed for another one. This happens for example when customers want a pallet under their products that cannot be placed under the products in the production process directly. It can also happen that products were made-to-stock on a certain pallet (single-use), while the customer wants another pallet type. If a pallet is swapped by a logistic action, also recyclable pallets can be used (Figure 4.3).

 Change number of sacks per layer

Sometimes the change of pallet results in a change of the number of sacks per layer. This happens when the new pallet is significant smaller than the original pallet. Then the number of sacks per layer has to be changed because sacks will otherwise stick out outside the new pallet. This results in that the sacks have to be stacked four on a layer instead of five.

 Change number of sacks per pallet

When height limitations are encountered or there are other reasons to make changes in pallet height, the number of sacks per pallets is changed. This can be more or less sacks than original. The above mentioned change of sacks per layer can be a reason to produce a pallet with less sacks, but also for fitting in a container or to fulfill customers’ specific wishes.  Pallets used for storage

Big bags are less suitable to stack on top of each other because the big bags are not as stable as a pile of sacks. Therefore, there is a maximum of two big bags that can be stacked up. It is also the case that when a big bag with a pallet is stacked on top of another pallet that the bottom one becomes dimpled. Then there is a chance that the forks of a forklift damage the bottom bag. To prevent for this an extra pallet is placed between both big bags for storage purposes. This pallet becomes available again when the big bags are unstacked.

 Pallets from third parties

Once in a while, third party materials are delivered on pallets. Think of raw materials such as empty bags or chemical. These pallets must also be disposed. They are thus also brought to the waste collection depot. These pallets are not recorded in SAP at all.

(30)

29

Identifying inventory record inaccuracy symptoms

After some internal observation in the warehouses of Avebe TAK it appeared that processes in SAP do not represent the processes that are actually taking place. In SAP some steps are skipped or cannot be recorded at all, because a lack of possibility in the system. It also happens that correctly implemented processes are edited manually to do something else. So many symptoms currently show that the pallet flows are not clearly visible in SAP at all. A list of the symptoms is given below. Five general symptoms of inventory inaccuracy (2.4.1) that were found are listed first, followed by two case specific findings:

 Lots of inventory errors; faulty location, material, quantity.

 No trust in the records; numerous calls to check on availability on floor.  Unwanted financial reporting surprises.

 Lost materials due to unknown causes.  Ineffective inventory ordering decisions.

 Physical processes are not recorded in information system.  High inventories that are not recorded in SAP.

So in conclusion five of seven general symptoms are in some extend observed in the production environment of Avebe TAK related to packaging. Symptoms related to failed orders or backorders could not be found. Possibly this has to do with the fact that pallets are ordered manually and that new pallets are almost always available due to high stocks. Two additional symptoms were found that both have to do with badly implemented processes. A more detailed report of the observations at the warehouse, which gives some extra information about underlying motives, can be found in Appendix F.

4.3 Inventory inaccuracy cause analysis

In an ideal situation, every physical action related to pallets should be going together with an administrative action. When pallets are used for production, a production order should be created in the system whereby the exact quantity of used materials is subtracted from the inventory. When pallets become available by any of the above mentioned logistic actions, these pallets should be added to the inventory records. When pallets are collected for transport to a pallet recycling company, pallets must be written off in the records. In this paragraph critical processes are pointed out and root causes will be indicated. After finding out the root causes they will be rated on impact so that the most important root cause(s) can be pointed out.

Finding critical processes

In our way to locate the main issues, critical processes are going to be discovered first. This gives insight in where the origins of inventory inaccuracy lay. Hellström and Sahir (2007) state that packaging engineers cannot make accurate decisions without identifying the processes the packaging will be put through. This requires identifying all business processes that are associated with packaging. To get insight in the flows of wooden pallets at the production location in Ter Apelkanaal, a process map has been made for both the physical and the information flows. To acquire basic and additional information a meeting was held with employees of in- and outbound going goods. Later the map was presented to and validated by the same people. After the confirmation of all flow directions were all the next step was to fill in the flow quantities.

(31)

30

information flows (Figure 4.5). In the first process map (Figure 4.4) flows of new, empty pallets are colored in black. These pallets are not always recorded in SAP. In green the pallets flows are displayed which are leaving the production location to the customer or to external warehouses. These pallet flows are recorded in SAP, but only as part of finished product (tertiary packaging), not as single packaging item. The flows in blue show the pallets that were primarily used in production but which are not part of finished product. Those are often pallets that became available by logistic actions.

Figure 4.4: Physical flows of wooden pallets at Avebe TAK – current state

(32)

31 If the two diagrams are compared is can be seen that a lot of physical flows are not linked to an administrative flow, implying that these flows are not recorded in SAP. The differences are especially seen in the outgoing empty pallet flows. These outgoing flows are all used wooden pallets. It was already said that a lot of inventory errors were found. In the lower picture, the parameter ‘inventory corrections’ shows that in the IS there were 16.000 corrections made in positive direction and 21.300 in negative direction. In Appendix A and B both maps can be found enlarged including their flow quantities in items for the financial year (FY) 20152_.

In Table 4.1 an analysis is shown of the most important flows (displayed in random order, sorted on implementation status). For all important processes which are shown in the second column, the flow quantities were explored. These can be found in the third column. Those are also the quantities that could be seen in Appendix A. The fourth column indicates if the process is currently implemented in SAP.

Table 4.1: Analysis of most critical logistic processes at Avebe TAK, quantities for FY2015

From this table the most critical process(es) must be chosen. These processes will receive the highest priority from now on because they affect IRI the most. In Figure 4.6, Table 4.1 is turned into a Pareto diagram for a better and clearer overview3_{. In the diagram processes that are already implemented}

in SAP are given the color green, processes that are not are given red and ambiguous processes are colored yellow.

2_{A Financial Year runs from 1 August the year before to 31 July in the FY.}

3_{An enlarged version of the same diagram can be found in appendix C.}

# Processes Amount in

pallets (output)

Implemented in SAP

1 Log actions (not in SAP) <500No

2 Loose loading 2.700No

3 Other production 1.800No

4 Pallets from material deliveries 4.000No

5 Pallets used for storage <1000No

6 Transport to pallet recycling company 41.000No

7 Receipt from external warehouse 145Not always

8 Sendings to external warehouse 6.100Not always

9 Change number of sacks per pallet/layer 2.900Yes

10 Change of pallet 3.000Yes

11 Receipt from supplier 165.700Yes

12 Regular production 160.000Yes

13 Regular sales 154.200Yes

(33)

32

Figure 4.6: Pareto diagram of process flow quantities for FY2015

The Pareto principle states that for many activities, roughly 80% of the effects are due to 20% of the causes (Grosfeld-Nir 2007). For the pallet flow, it can be seen that 88% of the flow quantities came from 3 out of 14 (21%) processes. The Pareto principle applies here. The process which has the highest quantity, receipts from the pallet supplier (11), is fully recorded in SAP. Of these received pallets 160.000 items were used for regular production (12). The most of them leave the company by regular sales (13). There are no reasons to doubt these numbers because they are realistic and no complaints were expressed about these first three processes. The fourth biggest process, transport to pallet recycling company (6), is colored red. This means that this process is not implemented in SAP and therefore it is a candidate for getting high priority in further investigation.

The other red colored processes have significant lower quantities and therefore less attractive to care about in the first step. Loose loading (2) has a ninth place here. For TAK this is not directly a process which gives the most benefit if solved, but from conversations it is known that for other locations this process is much more important. The ambiguous processes (7, 8) cover only a very tiny part and should not have first attention. So even though some processes here are not considered important enough to elaborate on directly, they should still be taken into account in the future because they might be important for other locations.