Combined make to order and make to stock in the food processing industry

(1)

the food processing industry

A CASE STUDY

Master thesis, MSc Supply Chain Management University of Groningen, Faculty of Economics and Business

PUBLIC AND CENSORED REPORT

June 21, 2013 BEREND FLEURKE student number: 1616978 e-mail: B.K.Fleurke@student.rug.nl Saffierstraat 224 9743 LP Groningen +31 624623587 First supervisor Prof. D.P. Van Donk

Second supervisor Dr. X. Zhu

Supervisor case company P. D.

(2)

ABSTRACT

Food processing industries have to deal with growing logistical demands, growing variety of products and intense competition. Hence, food processing companies are trying to move a part of their traditional make to stock (MTS) production to follow a make to order (MTO) strategy. For a company, using a MTS strategy for some products and a MTO strategy for others might be much more effective than using either strategy exclusively. Therefore, firms could employ a hybrid approach, a combined MTO–MTS system, holding inventory of some products and producing others to order. Only a small amount of papers has been explicitly dealing with this combined problem and these papers have been rather limited in exploring all issues relevant for combined MTO-MTS situation. This paper explores how process, product and market characteristics affect the MTO-MTS decision in the food processing industry. Additionally, the novelty of this research is that this research investigates how market requirements affect the production and planning process. This paper explores this for a medium-sized multi-product food processing company in the Netherlands. The results section presents an overview of the process, product and market characteristics which influences the MTO-MTS decision. Additionally, it seems that market requirements influence the production and planning process significantly.

(4)

1. INTRODUCTION

For many years it was a common policy for food processing companies to produce in large batches and to limit the number of set-ups to keep production costs low (Van Donk, 2001). However, the last decade showed a number of changes. Actually, food processing industries have to deliver a greater variety of products, have to meet higher logistical demands while keeping costs as low as possible (Van Donk, 2001; Soman, 2005). Akkerman et al. (2010) emphasize that in the food processing industry, a typical production plant produces a multitude of intermediate products in an even wider range of packages. Often, the output variety is based on a relatively small number of raw materials. In general, such a divergent product structure is typical for the process industries (Fransoo & Rutten 1994; Meulenberg & Viaene, 1998; Costa & Jongen, 2006). Differences in products can be associated with customer-specific products, either in packaging form, size, print, labelling, or product recipe. Moreover, Akkerman (2007) illustrates several important characteristics of the current food processing industry – i.e., the focus on production efficiency, lead time reductions, the importance of shelf life, the need for flexibility, and an increasing concern for sustainable production.

This paper explores one of the possible ways to cope with these changing market demands: relocate the decoupling point and move a part of production to follow a make to order strategy (Van Donk, 2001). Olhager (2003) distinguishes make to stock (MTS), assemble to order (ATO), make to order (MTO) and engineer-to-order (ETO) production strategies. These different manufacturing situations are all related to different positions of the customer order decoupling point (CODP). This CODP is also labelled as the order penetration point (OPP). Hoekstra & Romme (1992, p. 66) defined the customer order decoupling point as: “the point that indicates how deeply the customer order penetrates into the goods flow.” The CODP divides the manufacturing stages that are forecast-driven from those that are customer driven (Olhager, 2003). For a company, using a MTS strategy for some products and a MTO strategy for others might be much more effective than using either strategy exclusively. Therefore, firms are beginning to employ a hybrid approach, a combined MTO–MTS system, holding inventory of some products and producing others to order (Akkerman et al., 2010; Van Donk, 2001; Soman, 2005; Perona et al., 2009; Kaminsky & Kaya, 2009; Rafiei et al., 2013). A combination of MTO and MTS products and their interaction with the limited shared capacity opens interesting possibilities as well as problems for production planning. For example, MTS products might be manufactured to fill capacity in periods of low demand for MTO items. However, it is difficult for researchers to answer questions such as how much inventory should be kept or how due dates should be set in a combined MTO-MTS production situation (Soman, 2005; Kerkkänen, 2007; Rafiei et al., 2013). The question whether a certain product will be made to stock or made to order is the principal issue in designing and managing the production planning and control function. This MTO-MTS decision is more strategically oriented and is complicated due to various factors involved (Soman, 2005). The solution of this complicated MTO-MTS decision needs to consider the trade-offs between process, product and market characteristics (Olhager, 2003).

So, this paper is focussing on food processing industries, where combined MTO-MTS production is quite common. Food processing industries differ from discrete parts industry not only on the basis of kind of products, but also on market characteristics, the production process, and the production control. For example, limited shelf life of products adds another dimension to the combined MTO-MTS problem (Soman et al., 2004). Therefore, combined MTO-MTS production in food processing industries is an interesting and relevant research subject.

1.1 Research objective & research questions

(5)

Moreover, Van Donk (2000) emphasizes that combining make to order and make to stock is a difficult planning task, which need further research. Next to that and to the best of my knowledge, no research is done about how market requirements of the food processing industry affect the production and planning process. The issues described above lead to the following research question:

How do process, product and market characteristics affect the MTO-MTS decision in the food processing industry and how do market requirements affect the production and planning process?

The scientific contribution of this paper is that this, to the best of my knowledge, is the first paper which addresses how market characteristics affect the production and planning process of a company in the food processing industry. Over the last years, the logistical and inventory costs of the case company increased. This increase was an effect of the aim for efficient and long production runs within the production process of the company. Further, more and more different products have ended up on stock in the external warehouse. The reason for this is that the company has been confronted with customers who increasingly demand for more customer specific products and faster deliveries. The practical contribution for the company is that this study presents a list of variables (dedicated for the case company) which influence the MTO-MTS decision. Additionally, the paper shows that market requirements influence the production and planning process of the company significantly.

(6)

2. LITERATURE REVIEW

2.1 The food processing industry

The food processing industry can be considered as a part of the process industries, which is defined as firms that ‘add value by mixing, separating, forming and/or chemical reactions by either batch or continuous mode (Dennis & Meredith, 2000, p. 283). In general, the production process can be divided into two stages: processing of raw materials into intermediate products and packaging of food products (see Figure 2.1).

Van Donk (2001); van Dam (1995); Ten Kate (1994); Van Donk & van Dam (1996); van Wezel & Van Donk (1996); van Dam et al. (1998) and van Wezel (2001) distinguished a dozen of general characteristics in the food processing industry. Table 2.1 shows these characteristics, which can be separated into plant, product and production process characteristics.

Some studies (Van Dam, 1995; Van Donk, 2001) emphasize that the combination of MTO-MTS is quite common in food processing industries. A number of studies (e.g. Meulenberg & Viaene, 1998; Nakhla, 1995; Van Donk, 2001; Chang, 2012) show the increasing need for flexibility, due to growing logistical demands as the result of the change in market conditions for food processing companies. Other changes are a tendency towards more diversity and the growth of unique products for certain customers. The background and causes of these changes can be summarized under three main themes: Firstly, consumers’ wishes seems to change in an ever growing rate, causing an increase in packaging sizes, the number of products as well as in the number of new products introduced. Due to this large variety of end products, it is inefficient to produce and stock all end products. Therefore, a common practise used to mitigate the effect of product variety on the operational performance is to produce some semi-finished recipes in intermediate storage silos (Kilic et al. 2013). Secondly, many retailers are restructuring their supply chain

(7)

both in a physical and information flow sense. The aims are reduction in inventories, faster replenishment and shortening of cycle times. As a consequence, the logistical performance of food processing needs to be improved. Thirdly, the above mentioned changes have to be realized in a market which can be characterized by low margins in retailing and mergers and acquisitions in retail-chains. Both lead to a downward pressure on prices paid to producers (Van Donk, 1998; Soman, 2005). As a reaction, companies seek to produce some products to order and others to stock, which results in a reduction in the average batch size and a reconsideration of the way the production processes are planned and controlled (Van Kampen, 2012).

2.2 Make to order and make to stock issues

Olhager & Wikner (1998) and Olhager (2003) distinguished between make to stock (MTS), assemble-to-order (ATO), make to assemble-to-order (MTO) and engineer to assemble-to-order (ETO) production strategies. The differences between the strategies relates to the stage in the manufacturing value chain, where a particular product is linked to a customer order. Hoekstra & Romme (1992, p. 66) defined the customer order decoupling point (CODP) as: “the point that indicates how deeply the customer order penetrates into the goods flow.” This CODP is also labelled as the order penetration point (OPP). The position of the OPP for the production strategies is visualized in Figure 2.2. The dotted lines depict the production activities that are forecast-driven, whereas the straight lines depict customer-order-driven activities

Another possibility to delay product differentiation is the concept of postponement, whereby some of the activities in the supply chain are not performed until customer orders are received (Van Hoek, 2001). This is similar to the make to order strategy described above. Additionally, Van Hoek (1999) emphasize that the application of postponement in the food industry is fairly low, since food specific characteristics like: perishability and short lead times, limit the applicability of postponement. However, van Hoek (2001) emphasizes a strong focus on product standardisation in the food industry. The problem here is that standardisation is often difficult to achieve, as customers nowadays demand more and more product variety. Although, the divergence in the product structure does not have to happen at the first production stage, but could be placed later in the production process (Akkerman et al., 2010).

2.3 Combined MTO-MTS and the Customer order decoupling point

(8)

to delivery of the goods to the customers (Hoekstra & Romme, 1992, p. 2). During designing integral control, it is important to find a balance between: production and procurement costs, storage and distribution costs, and service to the customers. The result of this balancing establishes the decoupling point for a certain product-market combination.

Van Donk (2001) differentiates product & market characteristics, process and stock characteristics to determine the position of the decoupling point. Next, Olhager (2003) presented an extension of Van Donk’s proposed model (Figure 2.3). As we can see from this model, many factors can affect the position of the OPP. These factors are interrelated to some extent. Market characteristics affect product characteristics. For example, when customers claim short delivery lead times (market characteristics), it would make sense to use modular products (product characteristics) to adequately respond to customer demand.

Next to that, it is more profitable to produce standard products upon MTS strategy, since there is no customization for these products. MTO strategy is suitable for orders customized by various customers. The product range and customisation opportunities interact with market expectations and result in a delivery lead time that customers require with respect to the product offering. The product structure can be interpreted in terms of lead time with respect to the operations that need to be performed at each level. The relationship between production and delivery lead times is a major determinant of the position of the OPP (Olhager, 2003).

Table 2.2 shows the market characteristics and process characteristics of the food processing industry and their effects on the decoupling point.

Olhager (2003) summarizes some negative effects of forward and backward shifting of the decoupling point. When the CODP is shifted forward, negative consequences are the dependability of forecasts, a reduction in product customisation and an increase in work in process (due to more items being forecast-driven). When the CODP is shifted backwards, negative consequences are longer delivery lead times, reduced delivery reliability and reduced manufacturing efficiency (due to reduced possibilities in process optimisation).

Figure 2.3: Factors affecting the positioning of the order penetration point (Olhager, 2003, p. 323)

(9)

The papers of Van Donk (2001) and Olhager (2003) recognise that the CODP choice involves a trade-off between delivery time and inventory costs. This trade-off can be viewed as a problem of minimising the costs while meeting market requirements and satisfying process constraints (Soman et al., 2004). Regarding delivery time, the major factor is the production to delivery lead time ratio (P / D ratio). Christopher (1998) uses the term lead-time gap for the relation between production and delivery lead times. The costs are mainly affected by the relative demand volatility (RDV). The RDV is defined as the coefficient of variation, i.e. the ratio of standard deviation of demand and the average demand (Soman 2005, p. 30; Olhager, 2003, p. 326). Olhager (2003) developed a diagram (Figure 2.4) where these factors can be located. If the P / D ratio is greater than one, MTO strategy is not possible and MTS is the most suitable choice. If the ratio is less than one, MTO is possible.

However, it may also be possible to produce to stock to gain economies of scale. This is expressed through the RDV, such that a low RDV indicates that some recipes can be produced to stock. If the RDV is high it is not reasonable to use a MTS policy, since this would mean carrying excessive safety stock inventory. RDV has also been prescribed by D’Alessandro & Baveja (2000) for MTO-MTS decisions. They use RDV and average demand volume to categorise products into MTO and MTS. Products with high volume and low variability are MTS products. Products with low volume and high variability are MTO products (Soman, 2005).

The combined MTO-MTS decision gets more complex as the amount of product variation grows and the market conditions become more dominant. Choosing between MTS and MTO is closely connected with the problem of scheduling the production. Scheduling production of multiple products on a single facility that incurs significant change-over costs or times is one of the classic problems in production research (Kerkkänen, 2007). The insufficiency of mathematical approaches in solving the MTO-MTS decisions has been noticed by Soman et al., 2004. A main reason for this is the difficulty of modelling the problem accurately. Hence, the methods for analysing the MTO-MTS problem are often more like frameworks. An advantage of combined MTO-MTS production is that MTS products can be used to fill capacity in periods of low demand. A challenge in combined MTO-MTS production is to decide which products should be made to order and which to stock (Soman et al., 2004; Soman et al., 2006; Olhager, 2010).

2.4 Hierarchical production planning structure

Literature and concepts discussed in this literature review helps to understand the complex trade-offs involved in combined MTO-MTS decisions. This section provides guidelines for making these decisions. It is already stated that combined make to order and make to stock on a process is a difficult planning task, which need further research. Soman et al. (2004) discussed a variety of production management issues in the context of food processing companies, where combined MTO-MTS production is quite common. They finalize their paper with a comprehensive hierarchical planning framework that covers the important production management decisions to serve as a starting point for evaluation and further research on the planning system for MTO-MTS situations (Figure 2.5).

(10)

The advantage of having such a hierarchical approach is the partition of a global problem into smaller manageable component sub-problems. The framework provides an overview of the generalized hierarchy of decisions involved in the dynamic, combined MTO-MTS production situations. The first level is about the MTO-MTS decision. Similarities in process, product and market characteristics are used to form product families. Additionally, the information needed for locating the decoupling will be used to decide on MTO-MTS partitioning (Soman et al., 2004, p. 232). The authors name the second level: capacity co-ordination. At this level the demand and capacity is balanced. This level specifies the target inventory levels for MTS product in each planning period and sets policies for order acceptance and due-dates setting for the MTO orders. The third level is about scheduling and control decisions. The production orders are sequenced and scheduled. The time horizon for this level is typically from a day to a week and is worked out with as much detail as is needed.

The suggested conceptual framework is an attempt to structure the production planning decisions in a combined MTO-MTS production situation and can be used as a starting point for designing or redesigning the planning and scheduling hierarchy structure for a particular situation (Soman et al., 2004, p. 233).

2.5 Conceptual model

The research question consists out of two main questions. The first question is about how process, product and market characteristics affect the MTO-MTS decision. The second question is about how market requirements influence the production and planning process. First, it is important to determine which characteristics influence the complex combined MTO-MTS decision. It seems from the literature review that process, product and market characteristics influence the combined MTO-MTS decision. The concepts of these characteristics are presented in the conceptual model (Figure 2.6). Secondly, the framework of Soman et al. (2004) will be used in order to explore how market requirements influence the production and planning process. As we can see, this conceptual model is an extension of the framework of Soman et al. (2004) which is discussed in the literature review.

(11)

(12)

3. METHODOLOGY

3.1 Case study

In this paper a case study approach is used to answer the research question, since it allows studying a phenomenon in its real life context (Yin, 2003). The advantages of this approach are the rich data which ensues from the semi-structured interviews, observations, informal conversations, attendance at meetings and the extensive analysis of ERP data. This rich data, in contrast to data gathered through surveys, is more likely to accurately reflect the contextual factors influencing the case company. Therefore, a case study shows a greater fit with the explorative nature of this research (Voss et al. 2002). Furthermore, Van Donk (2001) conducted research and stated that more cases are needed to elaborate about the consequences of changing the position of the customer order decoupling point. Leonard-Barton (1990, p. 241) defines a case study as: “a history of a past or current phenomenon, drawn from multiple sources of evidence. It can include data from direct observation and systematic interviewing as well as from public and private archives. In fact, any fact relevant to the stream of events describing the phenomenon is a potential datum in a case study, since context is important”. Some challenges of case study research are: it is time consuming, it needs skilled interviewers and care is needed in drawing generalizable conclusions from a limited set of cases and in ensuring rigorous research. Despite this, the results of a case research can have a high impact and can lead to new and creative insights, development of new theory and have high validity with practitioners, the ultimate users of research (Karlsson, 2009). The purpose of this case study is exploration, since this paper tries to explore how process, product and market characteristics affect the MTO-MTS decision of the case company.

Karlsson (2009) stated that the unit of analysis refers to the level of data aggregation during subsequent analysis. It is the major entity that will be studied in this paper. Case studies permit to collect data from many perspectives and time periods. Consequently, Yin (2003) emphasizes that the unit of analysis must clearly define the outset of the study. The unit of analysis in this paper are the process, product and market characteristics as well as the planning process of the case company.

3.2 Case selection

The case concerns a medium-sized multi-product food processing company in the Netherlands. Over the last years, the logistical and inventory costs of the company increased significantly. On the one hand this increase was due to an increase in the total volume of end-products at the warehouse. This increase was an effect of the aim for efficient and long production runs within the production process. On the other hand, more and more different products have ended up on stock in the external warehouse. The reason for this is, inter alia, that the company has been confronted with customers who increasingly demand for more customer specific products and faster delivery. The company needs to quote short and reliable lead times to their customers to remain competitive in the market.

In the past, a couple of students explored this company in more detail (see van Honschoten, 1998; Rozier, 2001; Dijkstra, 2007; Roosma, 2011). However, the case company is still looking for a policy or approach that will help them to decide how to deal with the complex MTO-MTS decision. Additionally, due to increasing customer requirements, the company experiences difficulties with the planning process. Hence, this particular company is selected to explore their production and planning process in more detail. The advantage of using one case is that there is a great opportunity for depth of observations. However, one case limits to the generalization of the conclusions or frameworks developed (Karlsson, 2009).

3.3 Data collection

(13)

interviews. This protocol serves both as a prompt for the interview and a checklist to make sure that all topics have been covered (Voss, et al., 2002, p. 205). Additionally, the funnel model will be used during the interviews. The funnel approach is so named because it starts off the module with a very broad question and then progressively narrows down the scope of the questions until in the end it comes to some very specific points (Oppenheim, 1992, p. 110). Interviews will be recorded, since the exactness of what people have said could be important (Karlsson, 2009). Additionally, recording the data makes it easier to structure all the obtained information. Questions will be asked to multiple respondents (see Appendix B) to increase the reliability of the outcomes and to allow for triangulation. Finally, the transcripts of the interviews will be sent back to the interviewees to further validate findings and to verify interpretations. Next, unstructured non participant observations of the production process will be performed. Non participant observation is a term that is used to describe a situation in which the observer observes but does not participate in what is going on in the setting. Unstructured observation does not entail the use of an observation schedule (Bryman & Bell, 2003). The aim of these observations is to get a more thorough understanding of the production process. Informal conversations and attendance at meetings are the last form of primary data used throughout this research. Meetings, like a meeting of the Operational Team, will be attended in order to get an understanding of the current business situation.

The analysis of the ERP system and content analysis of documents are the main secondary data sources. The ERP software will be analysed to get a thorough understanding in product, production and demand characteristics. Next, some relevant documents about the principles of the company and the production and market characteristics will be analysed.

The second part of the main research question is about how changing market requirements influence the production and planning process. To answer this question, some concepts of the framework of Soman et al. (2004) will be explored (see Figure 2.5). Semi-structured interviews with the planner form the basis of the data collection to answer the second part of the research question. Additionally, the company started with keeping up planning data of the past couple of weeks.

Obviously, it is important that case research is conducted well, so that the results are both rigorous and relevant (Voss, et al., 2002). Because a number of methods are used to collect data, a deeper understanding of the issue will be obtained and findings could be triangulated (Karlsson, 2009; Cooper & Schindler, 2006). Internal validity is established since the process, product and market characteristics described in the literature review forms the basis for this research. Additionally, the concepts of the framework of Soman et al. (2004) will be used in order to answer the second part of the research question. External validity means knowing whether a study’s findings can be generalized beyond the immediate case study (Voss, et al., 2002). Apparently, the outcomes of this study could be generalized to companies within the food processing industry because these companies have to deal with the same market characteristics. However, the characteristics of the production process are dedicated to the case company and this makes it more difficult to generalize the findings of this study. Finally, reliability of data will also be increased if multiple sources of data on the same phenomenon are used (Voss, et al., 2002).

3.4 Data analysis

Obviously, the conceptual model forms the guide for this study (Figure 2.6). Some of the variables mentioned in the conceptual model are operationalized below.

Process characteristics:

Lead time

The time it takes from baking a particular recipe. Changeover times

The influence of changeover times on the production process. Controllability

(14)

Product characteristics:

Product perishability

The risk that products start to perish. Product customization

The number of unique products that can be produced from one particular recipe. Product specificity

The number of customers that ask a particular product.

Market characteristics:

Delivery reliability and delivery time

The importance of delivery reliability and delivery time according to the market. Demand variability

Order patrons in demand. A demand variability analysis (RDV) as suggested in D’Alessandro & Baveja (2000) will be followed.

Demand volume

Order frequencies and average order sizes per article.

During the semi-structured interviews, the above mentioned concepts will be discussed in detail. When there is a clear understanding of these characteristics, it is time to discuss how these variables influence the combined MTO-MTS decisions. Analysis of ERP data forms also a main source of information. This software package will help to determine several factors described above. The period from January 2012 till March 2013 is chosen for the demand analyse, since such a relative long period gives the opportunity to take seasonality into consideration. Furthermore, the more recent trend of increasing demand in mixes (mixes are often responsible for low volume orders) is included.

(15)

4. RESULTS

4.1 Process characteristics

The production process (Figure 4.1) has three steps: processing, granulation and packaging. An additional fourth step is mixing (between granulation and packaging). The processing stage starts with mixing the raw materials according to recipe. Next, dough slices are created and a machine cut the slices into small breads. Then, the breads need to ascend and subsequently the breads are baked into an oven. Thereafter, the breads are moved to the cooling tower, where these are crumbled, cooled and dried. From this cooling tower, the intermediate product could be stored in one of the 11 silos. The intermediate storage locations of the case company increase the flexibility of the company. The flexibility of the intermediate storage locations depends on the function of the particular silo. First, some of the storage capacity can be used to buffer in time against stoppages and changeovers within the bakery department. Besides, the silos can be used for anticipation inventory. Doing this, inventory is held to cope with expected future demand. By using the intermediate storage locations, the company can postpone the decision which end products to make.

At the granulation process, different pairs of roles granulate the raw crumbs into the desired granulation size. After this process, the granulated crumbs are sifted to determine the upper and lower granulation size. For example, a common granulation size is 0,3 - 1,6 millimetre, the crumbs below 0,3 mm are by-products since they are too small. Next, the crumbs with the desired size move to one or two of the two temporarily storage bunkers or the products are stored in big bags. The capacity of these two bunkers is 7-8 Tons, depending on the granulation size. Nowadays, the company keeps many big bags in stock, due to increasing demand for different mixes. After the optional mixing department, the products are packaged and transported to an external warehouse. There is a possibility to re-pallet the finished goods. This means that, for example, the end products on a wooden euro pallet can be changed to a plastic euro pallet.

4.1.1 Production times, minimal batch size and changeover times

The bakery department is the bottleneck. The throughput times for each batch in the bakery department are approximately two hours and for the granulation department about half an hour. Appendix C (confidential) shows the production speed per hour for each recipe. As we can see, the production speed in Tons/hour is more or less 3 Tons per hour. The average production speed for the granulation department is something like 3 to 5 Tons per hour, depending on the granulation size. The minimal batch size of the bakery department is 6 Tons, this is equivalent to production time of 2 hours.

The company differentiates different colours (product families) of breadcrumbs. These colours are: white, brown, yellow, orange/red, green/purple and brown. On average, changeover between product families takes about 6.5 hours. One exception is a set-up from a brown recipe to a white recipe, in that case the changeover time is approximately 1 hour. When the bakery changes to another recipe within the same colour family, changeover times are approximately 1 hour. The recipes within the colour families have slightly different colours (dark – light colours). The planner tries to minimise possibilities of

Bakery Granulation Mixing Packaging & Labeling Raw material storage Intermediate storage External warehouse 11 silos Mixing Temporarily storage 2 bunkers 2 silos

(16)

contamination, therefore sequencing the colours from light to darker colours is favourable. The changeover times mainly consist of cleaning the machines.

4.1.2 Controllability of the process

Weather conditions are of significant importance for controlling the dryer. Different temperatures and wetness influences the moisture percentage of the breadcrumbs. Imagine a rainy day. The outside air, which is needed to cool the breadcrumbs, contains significant more moisture. As a consequence, the just dried breadcrumbs get wet again. Moreover, the planner indicates that lead times per recipe can fluctuate significantly. This is due to the fact that the production process is dependent on different quality issues like the heat of the water and weather conditions. For instance, due to perfect circumstances, 3.1 Tons per hour can be produced per hour of recipe 1. But one day later, 2.9 Tons per hour can be produced, due to differences in heat of the water, different weather conditions or fluctuating quality of raw materials.

4.1.3 Process limitations

Obviously, there are two types of storage limitations: capacity constraints and time constraints. The maximum capacity of the 2 raw material silos is 50 Tons. 2 of the 11 intermediate silos have a capacity of 23 Tons and the other 9 silos have a capacity of 18 Tons. The time constraint has to deal with the perishability of the products. Furthermore, the granulation department can fill the 2 temporarily silos, while the mixing department uses the packaging line. However, the two departments cannot use the packaging line simultaneously, but it is possible for the granulation department to fill big bags in the same time. Hence, the mixing department limits the flexibility of the granulating and packaging department. Another significant important limitation is that customers do not accept delivery of products with also an older charge number. So article number 20002: White 01 N 0316 HE25 (recipe 1, with a granulation size of 03-1.6 millimetre on a wooden Euro pallet in packages of 25 kilogram) has a unique charge number, while article number 20002: White 01 N 0316 KE25 contains exactly the same product but is stored on a plastic pallet and therefore has another charge number. The planner indicates that it happens quite a lot that there are enough finished goods available, however the products are stored on different pallet types with older charge numbers.

4.2 Product characteristics

Obviously, the product flow of the company is divergent. A small number of raw materials could be processed to 49 different recipes. From these recipes multiple granulations can be produced. Then, it is possible to mix different crumbs. Subsequently, there are 3 different sorts of plastic packages. Additionally, the packaging department can bundle the products in packages of 3-25 kg bags, 3 kg buckets or big bags. Moreover, these bags can be printed with a customer specific label or with a standard label of the company. Next, the packaged product can be stored on 8 different pallet types. Finally, there are some possibilities to protect the whole pallet with carton or extra plastic. Apparently, product convergence emerges especially at the packaging stage. There are 620 stock keeping units (SKUs) after the packaging stage. During the period of measurement, 18.5% of all SKUs are complete customer specific. This means that these articles are dedicated to a specific customer.

4.2.1 Perishability

(17)

more limited. The company has to guarantee their customers a shelf life of four months at least. Therefore, slow moving products have a risk of becoming obsolete.

4.2.2 Product customization

In this case, commonality means the number of unique products that can be produced from one specific recipe. Obviously, the degree of commonality differs per recipe. There are differences in the number of granulation sizes and the number of end articles. Figure 4.2 gives an overview of the commonality per recipe.

These differences are relevant for the decision which recipes to store in one of the 11 intermediate silos. Storing the recipes which consist out of relative many granulations and end articles increases the flexibility. Since the recipes can be used for several different end articles.

4.2.3 Product specificity

In this case, specificity means the number of customers who ask for a particular article. Obviously, the decoupling point for ‘popular’ articles (with many customers) could be shifted downstream, because demand for that particular article is not dependent on just one customer. There are seven categories differentiated in Figure 4.3. As we can see, 61% of the 364 articles come from just one particular customer, hence these articles are customer specific.

(18)

4.3 Market characteristics

The company differentiates itself by trying to be a preferred business partner through focussing on customer value adding. Additionally, customers indicate that they appreciate the extensive collaboration, flexibility, reliability and the broad assortment. Obviously, products should meet customer requirements. Next to that, flexibility and fast deliveries are of significant importance in the food processing industry. During the period January 2012 till March

2013, the company differentiates approximately 300 customers. Total demand during that period was about 21,500 Tons. The classical Pareto distribution holds for the demand analysis of the company. 20% (60) of the customers are responsible for 85% of total demand. Next to that, the primary 5 customers are accountable for 35% of total demand. Consequently, a relative small number of customers are responsible for a high percentage of total demand. It seems that these customers are managed in a special way. The case company tries to collaborate

extensively with some of these customers by developing new unique recipes/mixes with special attributes to distinguish them from competitors. As we can see from Figure 4.4, total demand of batters and mixes increased slightly. However, it is expected that demand for batters and mixes will increase significantly in 2013 and that this will hold for the coming years. The gross margin of these mixes is almost 2 times more than the gross margin on breadcrumbs. It seems that most of the batters/mixes consist out of low volume orders. 61% 23% 4% 3% 2% 2% 5% 1 customer 2-3 customers 4-5 customers 6-7 customers 8-9 customers 10-11 customers 12 or more customers 0 5000 10000 15000 20000 25000 2010 2011 2012 2013 V o lu m e i n To n s Year Batters / Mixes Breadcrumbs

(19)

4.3.1 Predictability of demand

Most customers order, more or less, in a regular way and for most of the customers is demand rather predictable. On the other hand, there are customers with more unpredictable demand. For these customers, it is difficult to forecast demand. Imagine if a large retailer promotes schnitzels in a particular week, customer demand will increase significantly. As a consequence, short term demand for the case company will increase significantly. Next to that, fishery and weather conditions influences demand variability as well. Many fishes are layered with breadcrumbs. So, when the fishing is unexpected high, demand for breadcrumbs will increase accordingly.

Currently, the company is experimenting with a software package where customers can adjust or enter their expected demand. Additionally, the company uses software which gives a signal if customer demand deviates from the usual demand pattern. When such a signal arises, an employee calls the particular customer in order to verify the customers demand.

4.3.2 Demand variability

The company baked 49 unique recipes during the period of measurement. Recipe 1 is far most responsible for total demand (Appendix D: confidential). Similar to the total demand analysis, the Pareto distribution holds also for the demand analysis on recipe level. 20% of the recipes (10) are responsible for approximately 74% of total demand. Figure 4.5 shows a plot of average demand of the recipes (except recipe 1) per delivery date on the x-axis and demand variability (coefficient of variation) on the y-axis.

The coefficient of variation (CoV) scales the standard deviation by the average demand per delivery date and therefore let us compares variability among recipes of vastly different demand volumes. A relative low CoV means more predictable demand and relative high CoV means more unpredictable demand. The raw data used for this Figure could be found in Appendix D (confidential). Recipe 1 is excluded, since this recipe has an average demand of more than 22,000 Tons and including this recipe will result in a more vague Figure. As we can see from Figure 4.5, there is one (black) dot on the x-axis. This mark represents recipe 47 and it means that this recipe is just delivered 1 time in the period of measurement. Therefore the standard deviation for recipe 47 is zero and subsequently the CoV is also zero.

0,00 0,50 1,00 1,50 2,00 0 2.000 4.000 6.000 8.000 10.000 C o V

Average demand per delivery date in kg

(20)

4.3.3 Order sizes and order frequencies

From the recipes discussed before, many different granulations can be produced. All these different granulations and mixes have their own article number. Over the period of measurement (January 2012 – March 2013 = 65 weeks), there were about 11,750 orders, spread over 380 different end-products (on article level) with a total demand of approximately 20,500 Tons. There is weekly demand for some of the articles, however there are also articles who are just ordered once in the period of measurement. Figure 4.6 shows the number of weeks with demand (Y-axis) and the average demand in kg per demand week (X-axis).

Figure 4.6: Order patron of all articles

As we can see from Figure 4.6, there is weekly demand for 8 articles during the period of measurement because there are 8 (black) dots on the 65 week line. Obviously, it is easy to notice that most of the articles are ordered not regularly. These articles are shown on the lower side of Figure 4.6. Besides, these orders are responsible for a low average demand. When the raw data is analysed, it becomes clear that most of these articles are mixes, batters and customer specific breadcrumbs.

4.4 The influence of market requirements on production planning

Typical to the process of order confirming is the philosophy of first come first serviced. Daily, a rappel list (Appendix E, confidential) is discussed with the planner. The moment the order arrives, the amount of stock available and the due date of the order are considered by the customer service department for confirming the orders. Then, the planner checks how to deal with the subsequent orders (Appendix F, confidential). When Operations is not able to fulfil demand according to customer requirements, customer

0 10 20 30 40 50 60 70 0 5.000 10.000 15.000 20.000 25.000 # o f We e k s w ith d e m an d

(21)

service will contact the specific customer and propose a new delivery date. The company develops every week an overview of the planned production and an overview of the produced production. Appendix G (confidential) shows an example of the actual production in a particular week and explains how to read the overview.

The planner indicates that rush orders influences the planning process significantly. It happens regularly that customers require a particular article within a very short time frame. When the particular article and recipe are not on stock, difficulties arise for the planner. Perhaps, the planner can adjust the current planning by baking more of an already planned recipe. However, if the rush orders consist out of a recipe which is not planned during the coming days, the planner should adjust the planning and changeover times will increase significantly. Another difficultly is the influence of breakdowns and interruptions on the production planning. The planner emphasizes that unforeseen interruptions occur regularly.

Currently, the planning is almost purely done by intuiting and experience of the planner. After analysing the planning of the past couple of weeks, it becomes clear that batch sizes vary heavily from recipe to recipe (Appendix H, confidential). A reason for this is that the case company does not make use of forecasting techniques. Therefore, the planner has limit knowledge about future demand. Additionally, it seems that patterns in past demand are not structurally recorded.

After every week, it is possible to analyse the differences between the planned, produced and optimal production per week. Significant differences in these production numbers are observable. Figure 4.7 (confidential) gives an overview of the optimal, planned and produced production per week.

Planned production means how much the planner planned to produce in a particular week. Approximately, 2 days before the actual production, the planner plans a production schedule. To calculate the optimal production per week, the planner uses the planned production per week and sums up how many hours could be used more for production to fulfil a particular week. Then, a lead time of 3 Tons per hour is used to determine the (optimal) additional production per week. Hence, the calculation of the optimal production per week is not very precise because the planner assumes a lead time of 3 Tons per hour and as said before, this lead time can fluctuate significantly. Additionally, the planner does not take changeover times into consideration when the optimal production is calculated. So, the planner assumes one (large) batch that could be produced more to calculate the optimal production.

As we can see from Figure 4.7, the planned production in week 16 is higher than the actual production. This is due to a broken conveyer and disturbances

in supply (Appendix G, confidential). Moreover, in most weeks the actual total production per week is higher than planned production. The main reasons for this are the requirements of some customers. It Figure 4.7: Planned, actual and optimal production per week

(22)

happens occasionally that some customers require deliveries within 1 or 2 days. Due to this unexpected demand, the total number of recipes baked per week is significant higher than planned (Table 4.2). Evidently, these changes influence the average batch size per week.

Figure 4.8 (confidential) gives an overview of the planned, actual and optimal batch size per week. These are calculated by dividing the actual, planned and optimal production by the total number of recipes baked in a particular week.

Here follows an example to calculate the average optimal batch size for week 6. As we can see from Figure 4.7, planned production of week 6 is 302.5 Tons and the planned number of recipes baked is 11 (Table 4.1). Now, the planner calculates how much could be produced more

to fulfil week 6. It seems that optimally, an additional batch of 145.3 Tons could be produced. Hence, optimal production in week 6 is 302.5 + 145.3 = 447.8 Tons. This number should be divided by the planned number of recipes baked in week 6 and the additional optimal batch should be added. As a consequence, the average optimal batch size for week 6 is 447.8 / (11+1) = 37.3 Tons.

As we can see from Figure 4.8, the average actual batch size is, in the period of measurement, always lower than the average planned batch size. Again, the main cause is unexpected demand. Due to unexpected orders, the planner should adjust the planning and implement a new recipe or the volume of an already planned recipe should be adjusted. Because of this, the average batch in a particular week decreases. Table 4.3 presents the planned and actual changeover times per week. As we can see, the actual changeover times are almost always higher than the planned changeover times.

Figure 4.8: Planned, actual and optimal batch size per week

(23)

5. DISCUSSION

5.1 MTO-MTS decision

The main goal of this paper was to explore how process, product and market characteristics affect the MTO-MTS decision in the food processing industry. Additionally, the goal was to find out how market requirements affect the production and planning process.

Table 5.1 gives an overview of the characteristics which influences the MTO-MTS decision. The Table differentiates if a particular variable has a high influence (is present) or a low influence (is not present) on a company in the food processing industry and shows the preferred production strategy per variable. For example, if delivery requirements (like fast and reliable) are high, it is recommended to follow a MTS strategy. Additionally, if demand is highly predictable, it is recommended to follow a MTS strategy.

It seems that the case company does not take these variables into consideration. Obviously, the company produces almost all orders to order. To enhance the efficiency of the bakery, the planner aims at long production runs and therefore additional stock occurs.

Van Donk (1998) emphasizes that applying general rules for the location of the decoupling point are difficult in practise. It is interesting to notice that the case company is dealing with this issue for more than 10 years. Somehow, the case company does not apply general rules to locate the decoupling point, but the planner determines whether a production will be produced to order or to stock. Some authors (Van Donk, 2001; Soman et al., 2004 and Soman, 2005) did similar research in the food processing industry, but it seems that there is still a significant gap between theory and practice. Hopefully, this thesis contributed to addressing this gap.

Apparently, the market characteristics of the food processing industry have a downstream effect (MTS) on the decoupling point. The case company experiences that customers require faster deliveries and this results in pressure on the planning and production process. Next to that, customers do not accept two subsequent deliveries with the same or older charge numbers. Therefore, customers prefer a make to order

Market perspective High Low

Delivery requirements (time and reliability) MTS MTO

Assortment MTO MTS

Demand mixes (small batches) MTO MTS

Demand in different sorts of appearances MTO MTS

Predictability of demand MTS MTO

Demand variability MTO MTS

Product perspective High Low

Perishability of the product MTO MTS

Recipe specificity MTO MTS

Product divergence MTO MTS

Process perspective High Low

Sequence dependent changeover times MTS MTO

(Variability in) processing times MTS MTO

(24)

policy with short response time. Other important market characteristics of the food processing industry are a broad assortment (in different sort of appearances) and an increase in demand of small orders.

There are two main areas of responsibility that affect the divergent product structure. These are the Commerce department, which has to decide which article and which appearance to put on the market and Product development that is responsible to translate functional product specifications into physical recipes (Hoekstra & Romme, 1992). The commerce department determines the commercial product structure and hence lays the basis for the product structure. Therefore it is essential for operations management to be involved in the establishment of commercial activities. Product development is responsible for the realization of the specifications as determined by the commerce department. The way in which product development gives shape to this constitutes another important characteristic of the product structure. New recipes have significant consequences for the operations department. For example: new ingredients, storage of these ingredients, product divergence and set-up times. Hence, Operations has the task of confronting Development with the consequences of such decisions.

There is an opportunity to categorise the recipes showed in Figure 4.5 into several groups: (i) high volume, low variability (ii) high volume, high variability (iii) low volume, low variety and (iv) low volume, high variety. However, it is difficult and subjective to determine the border of low variable recipes, high variable recipes and low volume and high volume recipes (Soman et al., 2004). Some areas of conflicts will arise and it is important that Commerce and Production decides jointly whether a recipe falls in a MTO or MTS category. This information is of significant importance for the choice to produce recipes to order or to (intermediate) stock. Classifying a particular product or recipe as MTO rather than MTS can have significant implications in terms of longer lead-times for customers, less inventory and more changeover time.

The recipes with a high volume and a low variability are candidates for MTS production. The recipes with a relative low volume and high variability should be produced on MTO basis. The recipes in the relative high volume and high variability may be produced on MTS basis (Soman, 2005). However, Chatfield (2013) emphasize that increased demand variance requires greater levels of safety stock to be carried to maintain desired service levels. This requires: investments, production capacity and storage capacity. As a consequence, it is recommended that closer ties should be sought with these customers in order to reduce their variability.

As we can see from Figure 4.2, a number of recipes are used for just one article only. These products should not be stored in the intermediate silos since these recipes are client specific as soon as the bakery starts with baking the particular recipe. The order sizes are of significant importance for these recipes, since there is limited opportunity to combine batches. Dedicating a silo for one of these recipes will reduce the available amount of storage capacity for the other recipes. On the other side, it would make sense to use the intermediate silos for the recipes which consists out of many granulation sizes and end products. Doing this, mix flexibility increases since the intermediate product can be used for different end products. Next to that, the decision about the granulations size, mixing and packaging features will be postponed. Furthermore, as we can see from Figure 4.3, the number of customers per article differs significantly. The decoupling point for dedicated articles (with just 1 customer) should be shifted upstream. On the other hand, it is recommended to shift the decoupling point downstream for articles with many customers.

5.2 Production planning

The second part of the research question is about how market requirements affect the production and planning process. The hierarchical planning framework, presented by Soman et al., 2004, could be used to link the theory of production planning in the food processing industry with practice.

(25)

These are extensively discussed throughout this paper. Currently, the company does not take these characteristics into consideration. The next step in the framework is about capacity co-ordination (medium term plan state). According to the framework, capacity co-ordination is influenced by customer demand forecasts. As said before, the case company does not make use of demand forecasting techniques. Therefore, it is complicated for the planner to come up with an adequate planning. The MTO order acceptance policy, the due date policies for MTO products and the lot sizes for MTS products should be coordinated in this phase. The findings of this thesis can help with setting a policy for these factors. The third level is about scheduling and control decisions (process state). Here, the production orders are sequenced and scheduled. It becomes clear that the planning process of the company starts at this final step. The weekly production schedules are discussed in the results section and it becomes clear that rush orders of some strategic customers have a significant impact on the production schedule. More recipes are produced than planned and, therefore, changeover times increase significantly and that leads to a decrease of the average batch sizes. The result is a more inefficient production process. Furthermore, this increase in changeovers and smaller batch sizes reduce the total productive capacity. Hence, Commerce should realize that there is a possibility that the proposed budget should be adjusted.

(26)

6. CONCLUSION AND FURTHER RESEARCH

6.1 Conclusion

A list of variables is presented to decide how to deal with the complex MTO-MTS decision. The general decoupling point concept has been adapted to the specific characteristics of the case company. The paper gives an overview of the influencing process, product and market characteristics of a food processing company. The result is a list of variables which should be taken into account to determine the most suitable location of the decoupling point for a particular product. Furthermore, this paper contributed to the knowledge about the decoupling point theory in the food processing industry.

As a consequence of the significant increase in product variety and shorter lead-time requirements of the customers, the company is forced to shift to a hybrid MTO-MTS strategy. However, it seems that the case company produces almost all orders to order. For efficiency reasons, batch sizes are larger than necessary and this result in additional stock. The results show that the market requirements have a downstream effect on the decoupling point. Fast and reliable deliveries are of great importance in the food processing industry and there is a tendency towards more diversity and the growth of unique products for certain customers. Therefore, the production and planning process becomes more complex and difficulties arise in managing inventories since more different SKUs should be stored. The main managerial implication of this study is that it provides management some decisional factors for the MTO-MTS decision. However, these decisions are based on historical data and it becomes clear that the market is changing continuously. Hence, Operations and Commerce should extensively deliberate and collaborate by accepting customer requirements and how this influences the location of the decoupling point.

Additionally, it is investigated how market requirements affect the production and planning process. It becomes clear that rush orders are very common in the food processing industry. Consequences of this are more changeovers and smaller batch sizes and therefore a decrease in efficiency. The presented variables in this paper can help with dealing with these fluctuating market characteristics.

Although the paper describes a single case study, the presented variables can easily be related to other food producing companies. Many of the food processing companies have a two-stage production system with intermediate storage and a divergent product structure. Additionally, most of the food processing companies have to deal with the similar market characteristics described in this paper.

6.2 Limitations

This thesis obviously has certain limitations. First, the study did not take into account the holding costs and backorder costs. Obviously, these costs influence the MTO-MTS decision. Secondly, it seems that the ERP data was not always up to date and consistent. As a consequence some quantitative results could differ from reality but expected is that this will not influence the results significantly. Finally, with regard to the generalizability of the results, the planning software of the case company is dedicated to the company and the planning is purely done by the experiences and intuition of the planner. Therefore, the results about how market requirements influence the production and planning process are difficult to generalize.

6.3 Managerial recommendations

(27)

there is a thought of whenever a customer pays an additional price, the order should be accepted. However, it seems that it is not clear if the extra costs for these small (rush) orders are really allocated to the customer. Hence, the company should secure that the additional costs per order are allocated to the right customer. Another recommendation is to make some products customer specific at the external warehouse. Apparently, certain articles are customer specific because of a customized label. As a consequence, these articles are dedicated to a particular customer. The company should avoid having inventory of these products because of the high risk of perishability. Therefore, it is recommended that standardized (and efficient produced) packages should be stored in the external warehouse from where the products can be labelled. A last recommendation is to reduce the number of appearances. Doing this, it becomes easier to manage inventories. It seems that, due to hygiene regulations, plastic euro pallets become more popular in the food processing industry. Due to this trend, the case company can inform their customers to get deliveries on a plastic euro pallet. Furthermore, due to health and safety regulations, it is advisable to remove the 25 kg packages. As a consequence, holding and transportation costs will increase, since total volume per pallet decreases. However, it should be investigated to increase the number of 20 kg packages on a pallet. Most important is that Commerce maintains a policy to maintain or decrease the number of appearances.

6.4 Suggestions for further research

(28)

MANAGEMENT SUMMARY (IN DUTCH)

Kenmerkend voor de voedingsmiddelenindustrie zijn: weinig verwerkingstappen, een divergente goederenstroom en de beperkte houdbaarheid van producten. Jaren geleden streefden veel bedrijven in deze industrie naar efficiënte lange productieruns. Zo konden de relatief lange omsteltijden vermeden worden en werden producten op voorraad geproduceerd (MTS). Tegenwoordig is de voedingsmiddelenindustrie beïnvloed door de industriële trend van verhoogde productieverscheidenheid, snellere leveringen en kleinere ordergroottes.

Een mogelijke manier om op deze veranderingen in te spelen, is het verplaatsen van het klantorder- ontkoppelpunt (KOOP). Onder het ontkoppelpunt verstaan we het punt dat aangeeft hoe ver stroomopwaarts in een bedrijfskolom een klantenorder doordringt in het productieproces. Indien een bedrijf pas een product gaat maken nadat het besteld is spreken we van een make to order (MTO) strategie. Voordelen hiervan zijn minimalisering van voorraden en de mogelijkheid om producten exact af te stemmen op de wensen van de klant. Een belangrijk nadeel is dat klanten langer moeten wachten op het product.

Het kan voor een bedrijf efficiënter zijn om een gecombineerde MTO-MTS-strategie toe te passen. Producten die weinig besteld worden, zullen op order worden geproduceerd en producten die vaker worden besteld kan men beter op voorraad produceren. Er is echter weinig onderzoek gedaan naar gecombineerde MTO-MTS-strategieën. Ook is er weinig onderzoek gedaan naar hoe veranderingen in de marktvraag invloed hebben op het productie- en planningsproces van bedrijven. Deze thesis speelt hier op in en onderzoekt welke factoren invloed hebben op de complexe MTO-MTS-keuze. Daarnaast is er onderzoek gedaan naar hoe de dynamische markt van de voedingsmiddelenindustrie invloed heeft op productie- en planning.

Het bedrijf waar dit onderzoek is uitgevoerd ervaart dat bovenstaande veranderingen in de markt invloed hebben op het productie- en planningsproces. Net zoals veel bedrijven in de voedingsmiddelenindustrie kenmerkt het bedrijf zich door de mogelijkheid om tussentijds producten op te slaan. Het is van belang om te onderzoeken welke variabelen bepalen welke recepten men het best tussentijds kan opslaan. Interviews met werknemers vormen de voornaamste bron om data te verkrijgen, daarnaast worden ERP- data geanalyseerd om op een kwantitatieve manier informatie te verzamelen.

Uit het onderzoek blijkt dat proces-, producten marktkarakteristieken de locatie van het ontkoppelpunt bepalen en dus invloed hebben op de keuze om op order of op voorraad te produceren. Echter, het bedrijf produceert vrijwel alle orders op order. Batches worden groter gemaakt om toch efficiënt te kunnen produceren en zo ontstaat extra voorraad. Er is geen duidelijk beleid om producten op order te produceren of op voorraad. Daarnaast wordt duidelijk dat spoedorders een aanzienlijke invloed hebben op het productie- en planningsproces. Door de spoedorders worden meer recepten geproduceerd dan gepland. Hierdoor stijgen de omsteltijden en daalt de gemiddelde productierun. Daarom wordt er minder efficiënt geproduceerd. Verder wordt duidelijk dat er geen gebruik wordt gemaakt van vraagvoorspelling.