Conclusion - RESEARCH SCOPE - Eindhoven University of Technology MASTER Redesign of the control

3. RESEARCH SCOPE

3.5 Conclusion

This research project concentrates on increasing on time and complete performance of factories, as the analysis of section 3.1 showed that they are underperforming in delivering material on time and complete. In section 3.2, it is concluded that factory performance is mostly affected by overdue PIs. The research focusses on SPO data produced by three Vanderlande factories and a single subcontractor. The analysis of the second tier suppliers, executed in section 3.4, revealed that there is a gap between the target level aimed by factories and the actual Delivery to request performance of suppliers. Filtering second tier suppliers on only those who supply SPO material gave similar results. Vanderlande factories and second tier suppliers are using local stock points to increase performance, but this stock is not proactively controlled. Proactive inventory control can increase the material availability and thus performance. Chapter 4 elaborates on research questions defined to create proactive inventory control.

Delivery to request second tier suppliers VIS

Orderlines Delivery to request

15 4. RESEARCH DESIGN

In Chapter 3 it is concluded that factories are underperforming in delivering material on time and complete. A cause of underperformance are PIs delivered later than requested. Analysis of second tier suppliers showed that they are underperforming in delivering PIs at the request date. Local stock points are supposed to increase performance, however target levels are not met. Chapter 4 shows the problem statement, research question, scope and project approach set in order to increase factory performance.

4.1 Problem statement

The research project analyses a two-echelon supply chain involving second tier suppliers, factories and a subcontractor. Vanderlande factories and subcontractors have to receive PIs within four weeks, due to order placement six weeks before order finish date, and two weeks required for production. VIM and VIS are creating stock as a precaution against material unavailability. Decisions for the stock levels are made locally and are not proactively controlled. Furthermore, second tier SPO suppliers are keeping stock to deal with demand uncertainties. Still, actual on time and complete performance is lower than the target level. Based on this information, the main problem for this research project is defined as:

Vanderlande factories and their suppliers are underperforming in delivering material on time and complete for Posisorter projects, whilst keeping local inventory levels that are not proactively controlled.

4.2 Research questions

To reach the preferred service level, second tier suppliers, factories, and subcontractor have to improve control of their inventory levels. Furthermore, considering the system as a whole while using central decision making might increase performance even more.

4.2.1 Main research question

The following research question is determined:

What is the effect of proactive inventory control for the Posisorter equipment on performance of the factories and suppliers?

4.2.2 Sub-questions

This question is be divided into six sub-questions, answered during the research project:

1. What are measurable performance indicators for the factories and suppliers?

2. What is the performance of the current approach for uncontrolled stock points in the factories and at Vanderlande suppliers?

3. How do the factories and suppliers perform if they use local stock points and make local decisions?

4. How do the factories and suppliers perform if they use a central stock point at the supplier, no stock points at factories and make central decisions?

5. How do the factories and suppliers perform if they use central and local stock points and make central decisions?

6. What is the difference in performance between proactively and non-proactively controlled stock points?

Sub-question 1 and 2 are answered in Chapter 3, sub-question 3 to 5 are discussed in Chapter 5, and sub-question 7 is answered in Chapter 8.

4.3 Project scope

Shrivastava (1987) stated that a more detailed project increases feasibility, but simultaneously decreases scientific relevance. Consequently, this research project focusses on a two-echelon supply chain incorporation four factories (VIM, VIS, VIA, and a subcontractor) and second tier suppliers.

Furthermore, solely data concerning the SPO projects is used for the analysis. The goal is to design an inventory model for all PIs required in SPO projects. This scope results in a project detailed enough to ensure a certain level of feasibility, and scientifically relevant as literature concerning local and central inventory control policies can be used.

4.4 Project approach

The regulative cycle of Van Strien (1986) is used during the master thesis project. According to Van Aken et al. (1997), the regulative cycle consists of five process steps: problem definition, analysis and diagnosis, plan of action, intervention, evaluation. Figure 8 illustrates the steps and the activities executed for the research project. The master thesis project only concerns the first three steps, though the conclusion and recommendations of the master thesis report can discuss useful insights for the intervention step.

Figure 8, Regulative cycle by Van Strien (1975)

During the research, it has been important to find the right balance between rigour and relevance. In explanatory research the aim is to generate objective generic knowledge based on theories, whereas Field Problem Solving (FPS) has the aim to generate specific knowledge, and solve a specific business performance problem. The goal of the thesis is to make a contribution to both: improve the performance of the factories at Vanderlande and add something to existing literature. The model of Van Strien (1986) provided guidelines during the project, to ensure a both rigour and relevant research is conducted.

17 5. CONCEPTUAL INVENTORY CONTROL MODEL

In Chapter 4, it is concluded that proactive inventory control can increase factory performance. This chapter first discusses KPIs, in order to answer sub-question 1, next it elaborates on three conceptual inventory control models, in order to answer sub-question 3 to 5, and the chapter finalises with an analysis of the demand used as input for the models.

5.1 KPIs

Radke and Tseng (2012) emphasize that an ETO company should make a trade-off between low inventory budgets, high service level, and short delivery lead-times when determining the inventory levels. This trade-off can be translated into three KPIs, applicable in the supply chain of Vanderlande (Table 3). The high service level is similar to both the on time and complete performance of the second tier suppliers and the material availability at factories, resulting in the first two KPIs. The factories do not have an inventory budget, but Vanderlande strives to minimize its total costs, being the third KPI. Lead-time between the second tier supplier and the factory is assumed deterministic and is not translated into a KPI. The first KPI is indirect, as it directly affects the suppliers but indirectly affects Vanderlande, whereas the other two KPIs directly affect Vanderlande.

Table 3, Key Performance Indicators set for Inventory Control

KPI Definition

On time and complete performance

The percentage of order lines delivered on time and complete by a second tier supplier.

Material availability The percentage of order lines fulfilled from stock for the factory.

Costs The total costs of the inventory of the two-echelon supply chain.

An organisation optimises its inventory by finding an optimal balance between the inventory level, resulting in performance, and inventory costs. If material for orders cannot be picked from stock, this results in back-orders or penalty costs. Unfortunately, the penalty costs cannot be computed for Vanderlande, as those vary per project. Nevertheless, customer satisfaction is of great importance and the cost of losing customer goodwill can be immense. Instead of computing penalty costs, a service level is introduced, to ensure a high performance level (Bijvank, 2013). For Vanderlande this means that 98%

of the order lines should be delivered on time and complete. The objective function of the inventory model is therefore to minimize the costs subject to the service level constraint.

5.2 Conceptual model

The structure of an inventory model that is applicable, is a distribution system that consists of a single supplier and multiple buyers. Within Vanderlande, the second tier supplier is the supplier and all factories and the subcontractor are the buyers. Per sub-question the conceptual model is discussed.

5.2.1 Local inventory and local control

The first situation is a model considered by sub-question 3. In a locally controlled situation, the distribution system of Vanderlande would look like Figure 9. Demand arrives at the buyers and proceeds

to the supplier. The supplier then supplies the buyers. Every location has its own inventory model, which is controlled in a decentralised perspective.

Figure 9, Situation 1: Local decision making and local stock points

5.2.2 Local inventory and central control

The second distribution model is illustrated in Figure 10, and refers to sub-question 4. A decentral distribution model with five decision makers that locally control the inventory, can be organized differently by shifting from five decision makers into one decision maker, changing it into a central distribution model. In a centralised perspective, the decision maker has information of all locations and the goal is to optimise the system as a whole, instead of optimising the local stock points.

Figure 10, Situation 2: Central decision making and local stock points

5.2.3 Central inventory and central control

A special type of a central distribution system is one with zero inventory levels at the buyers. This is the third situation, reflected by sub-question 5 and displayed in Figure 11. In this situation, solely suppliers keep stock.

Figure 11, Situation 3: Central decision making and central stock

5.2.4 Assumptions and input parameters

A number of assumptions are made for the three conceptual models. Moreover, in order to optimise the decision variables of the models, input parameters are defined. These assumptions and input parameters are discussed in Appendix B.

5.2.5 Demand estimation procedure

One of the input parameters is the demand. A data analysis is executed to estimate the demand input parameter. Demand for the first four weeks is deterministic and demand of the weeks after the fourth week is stochastic. The stochastic part of the demand can be approximated with ratios derived from historic or forecasted SPO projects. Due to the project environment of Vanderlande, the forecasted SPO data provides a better fit than historic data, although it is unknown whether information about PIs can be derived from the SPO Future demand. The stochastic part is determined by the following steps:

1. Per component, define all sub-components;

2. Define all possible types per sub-component;

3. Define all instances within a specific type;

4. Attach a BOM to each instance;

5. Partition the instances and types on their BOM: create variants with equal BOMs. A variant can thus contain one instance or multiple instances of one type.

6. Determine what characteristics identify the variant.

7. Create an overview from the purchase items and variants. This results in a table with vertically all variants within a component, horizontally all items and in every cell the quantity of that item in a variant. Table 4 provides example of this overview.

8. Use historic data to determine the average ratio of each variant per week, compared to other variants. Use the forecast data to determine the weekly expected order quantities.

9. Use the ratios to approximate the demand per item per week. The purchase item orders are assumed to arrive with a Binomial distribution, based on their occurrence in variants.

Next, the purchase item demand has to be allocated to a stock point. The items can be stocked at the buyers or at the sellers. There are decision rules to determine the demand at each stock point:

1. Component level: Component (B) is solely produced by VIM. Component (C) is produced by VIA, by a subcontractor and by VIM. Component (F) is produced by VIM, VIS, and VIA.

2. VIM is supposed to be the flexible factory, therefore the demand is first allocated to other factories.

3. VIA has highest priority as most projects are sold to American companies. It can produce 80 exits per week. The subcontractor can produce 120 exits per week and has second priority. All other exits are produced in Veghel.

4. VIA can produce 100 meter of component (F) and has highest priority. This needs to be a variant with a width of 1400 mm, an angle of 20°, and made out of steel. VIS can produce 200 meter and these should also be made out of steel. VIM can produce 200 meter per week.

Table 4, Example demand analysis

According to Chopra and Meindl (2013) a company must understand the uncertainty within its products to identify the extend of unpredictability of demand. By conducting this analysis, one can conclude how many variants with a unique BOM there are within a component, what characterises the variants and what ratios should be attached to the variants, to estimate the future PI demand distribution. All sub-components and types are illustrated in Figure 12, Figure 13, and Figure 14. The figures show two types of split ups: an AND-split up, meaning that both branches lead to a required sub-component, and an undefined split up, meaning that a decision has to be made on which sub-component is required.

5.3.1 Variants

The three components that are manufactured by the factories and subcontractor require different ways to compute the demand in components per week. This is shown in Table 5.

Table 5, Components and their measurements

Component Way of measuring

B presort/shoe-merge Number of projects

C, sorting Number of exits

F, transportation Meters

Component (B)

Step 1 is to define all sub-components, step 2 to define the types per sub-component, and step 3 to define instances. Component (B) consists of two sub-components (Figure 12). The first sub-component has two types: a Pre-sort or a Shoe merge. The second sub-component is the Bed section, which has no types. The Pre-sort type has twelve instances, the Shoe merge 24 instances, and the Bed section six instances.

Figure 12, Sub-components and types of Component (B)

In step 4 a BOM is attached to each instance, and in step 5 the BOMs the instances are partitioned on their BOM. All instances of the Pre-sort type have a unique BOM, resulting in twelve variants. The instances of the Shoe merge can be partitioned into four variants, and the six instances of the Bed section have a similar BOM, resulting in one variant. The variants of component (B) are illustrated in Figure 28 (Appendix H). The BOM of a Pre-sort consists of on average twenty PIs, the Shoe-merge sub-component consists of two to three PIs, and the BOM of a Bed section of a single PI.

Step 6 is to determine the characteristics of each variants are these are required to compute ratios based on the SPO Future demand. The twelve variants of the Pre-sort are characterised by:

 Left or right sorting;

 A width of 900, 1000, 1100, 1200, 1300, 1400 mm.

The variants of the Shoe merge also have two characteristics:

 Left or right sorting;

 A solid or molded merge block.

The Bed section characteristics are not needed, as it only has one variant.

Component (C)

Step 1 to 5 are also executed for component (C). Component (C) consists of three sub-components: a Switchframe, Fillings plate and a Safety wedge (Figure 13). There are twelve types of Switchframes, and each type has 24 instances. The Fillings plate sub-component has three types, and each type contains 180 to 1200 instances. There is only one Safety wedge type, and 6000 instances within this type.

Step 5 is the partition step, resulting in 24 variants of the Switchframe sub-component. A Switchframe sub-component variant can have a BOM consisting of 14 to 28 purchase items. The instances of the three types of Fillings plates and the Safety wedge do not lead to variation within the BOM. It thus suffices to attach one BOM to a Fillings plate and one to a Safety wedge, resulting in one variant per sub-component. The 24 variants of the Switchframe, the single variant for the Safety wedge, and the single variant for the Fillings plate are illustrated in Figure 29 (Appendix H).

Figure 13, Sub-components and types of component (C)

In step 6 the characteristics of the variants are identified. The Switchframe variants are characterised by:

 Dual or single sorting;

 Left or right sorting;

 Angle of 20 or 30 degrees;

 Missing pin detection or no missing pin detection;

 Molded merge block or solid merge block.

The other two components do not need characteristics, as partition led to one variant per sub-component.

Component (F)

Component (F) consists of one sub-component: Carrier. There are four types of Carriers (Figure 14), and each type contains 24 or 12 instances. By partition, 24 variants are defined. The variants are displayed in Figure 30 (Appendix H). The BOM of a variant of component (F) consists of on average 30 purchase items.

Figure 14, Sub-component groups of Component (F)

The execution of step 6 results in the following characteristics to identify the variants:

 Plastic or steel;

 Angle of 20 or 30 degrees;

 The width of 900, 1000, 1100, 1200, 1300, 1400 mm.

5.3.2 Ratios

The goal of step 7 is to create an overview of purchase items and variants and the goal of step 8 to determine ratios, based on both historic and SPO Future demand data. A ratio is the relative size of a variant compared to other variants. The computations and results are given in Appendix I. The intention was to approximate PI demand with the ratios, and use this as input for the inventory model. However, during the computations, it was concluded that most characteristics required to define a variant are described in the SPO Future demand. These are the following characteristics:

 Number of projects;

 Steel or plastic carriers;

 Expected shipping date for manufacturing;

 Number of SPO meters;

 Width of the SPO;

 Number of exits;

 Sorting angle of 20 or 30 degrees;

 Dual or single sided sorter.

There are three characteristics required for the variants that are not defined in the SPO Future demand:

 Left or right sorting;

 Missing pin detection or no missing pin detection;

 Molded merge block or solid merge block.

5.4 Conclusion

In section 5.3, a demand analysis is executed to define input for the conceptual models discussed in section 5.2. Due to the project environment, the goal was to first use SPO Future demand to compute ratios of variants that require information defined in the conceptual design phase, and then use historic data for all ratios of variants that require information defined in the detailed design phase, as this information is not available in the SPO Future demand. However, the demand analysis proved that SPO Future demand can be used to compute most ratios. Therefore the majority of the information required to compute PI demand is available after sales engineering, and thus before engineering. As a result, it seems worthwhile to adjust the control structure. An intermediate question is formulated:

To what extent can Vanderlande order its purchase items for the SPO projects at the end of the conceptual design phase?

Chapter 6 analyses the purchase items individually, in order to answer this question.

24 6. PURCHASE ITEM ANALYSIS

In Chapter 5, an intermediate question is defined, as the results of a demand analysis of SPO projects revealed that most information about characteristics of the SPO, required to determine PI demand, is available after the conceptual design phase, accomplished by sales engineering. This phase is followed up by the detailed design phase, which takes approximately two months. Splitting a SPEC into OFs and POs takes approximately one week. If Vanderlande is able to order PIs before the start of the detailed design phase, this results in around two extra months for the second tier suppliers to produce and deliver material. The characteristics required to determine a BOM are described in Chapter 5. In this

In document Eindhoven University of Technology MASTER Redesign of the control structure and inventory control at Vanderlande Gubbels, P.M. (pagina 31-0)