Design of a no-rejection time-based
policy for the hybrid MTO – MTS
food processing industry
MASTER THESIS December 2013
by:
FLORENTINE CATZ
Dual Masters in Operations Management University of Groningen
Newcastle University
(Leading-) Supervisor: J. Riezebos (Vice-) Supervisor: Y. Yang
Abstract
In the hybrid make-to-order (MTO) – make-to-stock (MTS) food processing industry, planning and scheduling problems have received much attention by scholars. However, in literature it is assumed that MTO orders may be rejected, while in practice this might not always be possible. If uncertainties in demand cannot be absorbed by the MTO order acceptance/rejection decision, increasing and varying customer lead-times (LT) may appear. This dissertation contributes to the scientific and practical relevance by aiming to design a no-rejection time-based policy for hybrid MTO – MTS systems in the food processing industry with as main constraint that MTO orders may not be rejected.
Based on the findings from both practice and theory, it is in this dissertation suggested to divide all stock keeping units (SKUs) in the no-rejection time-based policy over three categories, namely MTS, MTO fast response and MTO slow response. The customer LT of the MTO fast response category is short and fixed and for the MTO slow response category long and variable. The proposed no-rejection time-based policy includes capacity distribution prioritization, maximum order quantity determination, and the MTS sub-categorization. The policy minimizes the negative affects regarding customer LTs and MTS services levels. This dissertation presents a mathematical model to analyse the interaction between the capacity availability, capacity requirements per category, MTO customer LTs and MTS service levels. In this dissertation the performance of the no-rejection time-based policy is tested based on historical data of a food manufacturer that is unable to reject MTO orders. The results have shown that with the no-rejection time-based policy both the fast response and slow response customer LT are achieved without affecting the MTS service levels This dissertation presents, for the first time, a time-based policy for hybrid MTO – MTS systems that always accepts orders.
Acknowledgements
I am very excited to hand in the final result of my Master Thesis, which is also the last part of my degree of Master of Science in Technology and Operations Management (University of Groningen) and Master of Science in Operations and Supply Chain Management (Newcastle University Business School). The work presented in this dissertation was carried out at Unilever - Sourcing Unit Oss in the Netherlands.
The final result never would have been accomplished without the help others, and I owe them my thanks. First of all, my (leading-) supervisor of the University of Groningen, Jan Riezebos, for his great communication, clear feedback and patience. Second, my supervisor at Sourcing Unit Oss, Jeroen Zwitserloot, for giving me the opportunity to perform my Master Thesis at Sourcing Unit Oss and his guidance and enthusiasm throughout the project. Next, I want to thank the planning department of Sourcing Unit Oss for creating a pleasant environment to work in, and especially, Martijn Keur, for supporting me when performing my research. Also, I would thank my (vice-) supervisor from Newcastle University, Ying Yang, for her time to read my Master Thesis and her great commitment when studying at Newcastle University. Finally, I would like to thank my family and friends, simply for being around during the whole trajectory.
Table of content
Abstract ... 2 Acknowledgements ... 3 Table of content ... 4 1 Introduction ... 5 2 Theoretical background ... 72.1 Hybrid MTO – MTS systems ... 7
2.2 Policies at medium-term plan level ... 8
2.3 Medium-term plan parameters ... 9
3 Research context ... 11
3.1 Methodology ... 11
3.2 Case company ... 11
3.3 Data collection and analysis ... 12
4 Findings explorative phase ... 14
4.1 Short and reliable customer LTs ... 14
4.2 Customer LT and its factors ... 14
4.2.1 Supply LT ... 15
4.2.2 Waiting and processing time ... 15
4.2.3 Transportation time ... 17
5 No-rejection time-based policy ... 18
5.1 Analysis of the parameters ... 18
5.2 Process steps ... 19
5.3 Mathematical model ... 22
6 Validation ... 27
6.1 Performance no-rejection time-based policy ... 27
6.2 Determining maximum FR order quantity ... 30
7 Discussion ... 33
7.1 Findings explorative phase ... 33
7.2 No-rejection time-based policy ... 33
7.3 Validation ... 34
8 Conclusions, limitations and further research ... 35
9 Appendices ... 36
1 Introduction
The production planning of the food processing industry can be considered complex as many food manufacturers operate in high-mix-low-volume production systems to meet diversified customer’s demands (Matsumoto, Kashima, & Ishii, 2011). Moreover, it is indicated that the food processing industry is part of ‘very competitive supply chains and has to cater an
increase number of stock keeping units (SKUs) of varying logistical demands like specific
features, special packaging and short due dates’ (Soman, van Donk, & Gaalman, 2004, p. 10). Furthermore, consumer behaviour is irregular, resulting in orders that include small deliveries within a short and dependable time window (Soman et al., 2004). Additionally, food processing characteristics, such as perishable goods, shared resources and limited capacity of machine and labour, increases the complexity of the production planning (Akkerman & van Donk, 2008). As a result, a dilemma in the food processing industry emerges as the
manufacturers need to respondto the market but simultaneously control manufacturing costs
(van Donk, Akkerman, & van der Vaart, 2008).
Traditionally, a make-to-stock (MTS) system is most common for the food processing industry (van Donk, Soman, & Gaalman, 2005). The main advantage of MTS systems is that a short customer lead-time (LT) can be achieved as finished goods inventory is present (Hemmati & Rabbani, 2009). However, due to increased competition in the food processing industry and a growing number of SKUs, the industry has to produce part of their SKUs as make-to-order (MTO) (Soman et al., 2004). MTO systems not only increase customers satisfaction, it also helps manufacturers to eliminate finished goods inventory (Iravani, Liu, & Simchi-levi, 2009). In hybrid MTO – MTS systems, a portion of the production system operates as a MTS system and the remaining part operates as a MTO system. A proper combination of MTO and MTS can exploit both the advantages of both low inventory and short customer LTs (Hemmati & Rabbani, 2009). However, MTO and MTS systems have
contrasting production planning policies, which increasesthe complexity (Kaminsky & Kaya,
2009; Rafiei & Rabbani, 2012; Soman, van Donk, & Gaalman, 2006). The production planning of MTO systems is often order-focused and based on actual demand, whereas the production planning of MTS systems is product-focused and based on forecasts (Kaminsky & Kaya, 2009).
These production planning and scheduling problems in the hybrid MTO – MTS food processing industry have received much research attention (Kalantari, Rabbani, & Ebadian,
2011; Rafiei & Rabbani, 2012; Soman et al., 2004). Some scholars made an attempt to
MTS systems by proposing a model to cope with MTO order acceptance/rejection policy, MTO customer due date setting, lot sizing of MTS SKUs and determining required capacity during the planning horizon. However, these studies are based on the assumption that MTO order rejection is possible, while in practice, this may not be allowed. For instance, in case long-term agreements exist with customers who require a MTO policy (MTO-customers). As managers occasionally fail to fulfil all MTO orders due to insufficient planning capacity (Wu & Chiang, 2009), this dissertation seeks to develop a no-rejection policy to support food manufacturers that are not able to reject MTO orders.
The MTO order acceptance/rejection decision can be seen as an option to absorb uncertainties in demand in hybrid MTO – MTS systems. In case of constrained capacity, more MTO orders are rejected, whereas in case of idle capacity more MTO orders are accepted. It is expected that in case MTO order rejection is not allowed uncertainties are absorbed, by deciding on customer LTs as these increase if capacity is constrained. In this dissertation, policies that focus on customer LT are called time-based policies. While time-based policies have not been presented for hybrid MTO – MTS systems, they have for MTO systems. For instance Ebadian, Rabbani, Torabi, & Jolai (2009) developed a model to manage delivery dates of arriving MTO orders in order to reach short and reliable customer LTs in a MTO system. Other scholars, focused on customer LTs by proposing a model for MTO systems that support decision makers when they verify the feasibility of due dates required by potential customers (Corti, Pozzetti, & Zorzin, 2006).
This dissertation aims to contribute to the scientific and practical relevance by designing a no-rejection time-based policy for the hybrid MTO – MTS food processing industry. The research question that follows:
Which time-based policy supports the hybrid MTO – MTS food processing industry with as main constraint that MTO orders may not be rejected?
The sub-questions addressed in this dissertation are:
1. Which parameters influence customer LT according to theory? 2. Which parameters influence customer LT according practice?
3. Which parameters, process steps and mathematical model support the policy? 4. How effective is the policy if tested on historical data?
2 Theoretical background
This section describes the theoretical background of this dissertation. The first section reviews in short studies about hybrid MTO – MTS systems. The second section discusses the policies at the medium-term plan level within hybrid MTO – MTS systems. The last section answers the first sub-question by providing an overview of medium-term plan parameters that influence customer LT according to theory. It is important to note that throughout this dissertation the term MTO is defined as SKUs that are not held in stock, which is a widely used definition in the hybrid MTO – MTS literature (Soman et al., 2006).
2.1 Hybrid MTO – MTS systems
The first scholar that mentioned a MTO – MTS system was Williams (1984). Williams proposed a method for analysing one-stage systems that includes both the stochastic nature of the real world and the interaction between SKUs and capacity. Williams demonstrated that situations occur where many different SKUs in an inventory system compete for manufacturing capacity. In addition, he criticised the fact that many authors so far assumed that demand and manufacturing time for SKUs are known. However, this assumption is not always justified. This reveals topics for further research, such as which goods should be stocked, what special business should be accepted, what is the effect on the stock system and how can this effect be buffered (Williams, 1984).
Two decades later, Soman et al. (2004) presented a state-of-the-art literature review on hybrid MTO – MTS systems in the food processing industry. They stated that many food manufacturers are shifting towards a MTO system, due to the increasing variety of SKUs with varying logistical demands and production characteristics. In their research a comprehensive hierarchical production-planning (HPP) framework is proposed, which is displayed in Figure 2-1.
This framework covers the important production planning decisions and serves as a starting point for further research on hybrid MTO – MTS systems (Soman et al., 2004). It distinguishes three levels depending on the frequency in which decisions are made in, namely low (process state), medium (medium-term plan state) and high (production plan state). In subsequent research, scholars studied all three levels of the HPP framework. Because the MTO order acceptance/rejection decision is assigned to the medium-term plan level of the HPP framework this research focuses solely on the medium-term plan level. The literature related to this level of the HPP framework is reviewed below.
2.2 Policies at medium-term plan level
According to Rafiei & Rabbani (2012), the medium-term plan level is the most important level of the HPP framework. Soman et al. (2004) included the following medium-term decision subjects in their HPP framework: MTO acceptance/rejection policy, due date policies for MTO SKUs, lot sizes for MTS SKUs, and monthly production volumes. When they introduced the HPP framework they assumed that MTO order rejection is possible. Subsequent scholars also stated this assumption and according to the authors’ knowledge, no scholar even questioned it.
Kalantari et al. (2011) presented a decision support model for the MTO order acceptance/rejection policy in hybrid MTO – MTS systems. In this model the MTO order acceptance/rejection decision is based on customer priority, capacity and inventory calculation, prices, and delivery dates of MTO orders. Although Kalantari et al. (2011) mentioned that the optimal situation is to accept all arriving MTO orders at the company, they argued that constraints in production resources and material supplies forces the company to choose the best set of MTO orders. Rafiei & Rabbani (2012) addressed the medium-term plan level by proposing a model that includes all medium-term decision subjects, including MTO order acceptance/rejection policy, MTO order due date setting, lot-sizing of MTS SKUs and determining required capacity during the planning horizon. In this model only the most desirable MTO orders are accepted. More recently, Rafiei et al. (2013) proposed a bi-level hierarchical planning structure for hybrid MTO – MTS systems. They include both medium-term plan state (tactical) and process state (operational) in their hierarchical planning structure by developing a systematic and integrated approach towards tactical and operational issues. The feasibility of MTO order completion is analysed based on the available time of the planning period until the due date of the MTO order, and on the available capacity. In this approach the MTO orders are accepted if it is profitable to increase capacity. In case profit contribution of the MTO order is greater than the cost of capacity increase, the MTO order is stated as profitable and is thus accepted; otherwise, it is rejected (Rafiei et al., 2013).
case of idle capacity. In this dissertation, it has been proposed that when MTO order rejection is not allowed, uncertainties may be absorbed by deciding on customer LTs. The next section investigates which parameters influence customer LT in hybrid MTO – MTS systems.
2.3 Medium-term plan parameters
This section answers the research question ‘which parameters influence customer LT according to theory’. Two aspects of customer LT can be considered, namely the mean and the variance of customer LT. This section focuses on the variance of customer LT as absorbing uncertainties in demand especially influence the variance of customer LT. This sub-question is answered by considering studies regarding hybrid MTO – MTS systems (Kalantari et al., 2011; Rafiei et al., 2013) and MTO systems (Corti et al., 2006; Ebadian et al., 2009) as the MTO order acceptance/rejection decision is applicable to both systems. Kalantari et al. (2011) presented guidelines in their research to negotiate with customers about due dates in hybrid MTO – MTS systems. In the model of Rafiei et al. (2013) due dates are negotiable if capacity increase is not profitable. Corti et al. (2006) proposed a time-based policy for MTO systems by developing an integrated model that supports decision makers in the process of verifying due date feasibility. Ebadian et al. (2009) addressed MTO systems by adjusting the capacity in their model to their promised delivery dates. Their aim is to achieve short and reliable due dates. Ebadian et al. (2009) presented a time-based policy for MTO systems and acknowledge the importance of capacity modification as outsourcing activities are considered in their model. They mentioned customer LT, but focused on decreasing variance in customer LT by modifying capacity. Therefore, a first parameter that influence customer LT is capacity modification.
Customer LTs increases as capacity is constrained. In case capacity fluctuates through capacity modification activities, uncertainties in demand can be absorbed. In this way, variances in customer LTs can decrease. Kalantari et al. (2011) assumed that capacity modification is possible through overtime shifts or outsourcing activities. Rafiei et al. (2013) allowed capacity modification, as an increase in capacity was profitable. Corti et al. (2006) adopted a capacity driven approach to compare the capacity requested by both potential customer and the already confirmed MTO orders, with the actual level of available capacity. The model assumes limited capacity and calculates how many of hours of outsourcing would be requested in case of under capacity.
An overview of the findings is presented in Table 2-1. The answer to sub-question one regarding the parameters that influence customer LT in a no-rejection time-based policy according to theory is: (1) capacity modification and (2) the ROP. These are taken into account when developing the no-rejection time-based policy in section 5.
Assumptions Parameters Authors MTO order
rejection Customer LT Capacity modification ROP Kalantari et al. (2011) Hybrid MTO – MTS systems Allowed. Based on customer and MTO order priority
Delivery dates are negotiable with customer.
Overtime shifts or outsourcing.
MTS processing is triggered by demand used the next period.
Rafiei et al. (2013) Hybrid MTO – MTS systems Allowed. Based on customer and capacity availability.
Due dates are changed if capacity increase is not profitable.
Capacity increases if profitable.
MTS inventory levels are defined each planning period by balancing demand and capacity.
Ebadian et al. (2009) MTO systems Allowed. Based on importance of customer.
Due dates are fixed and short (the aim)
Outsourcing. - Corti et al. (2006) MTO systems Allowed. Based on probability values.
Decision support model to decide on feasibility of due dates Capacity is limited. Hours of outsourcing are calculated. -
Table 2-‐1: Medium-‐term planning parameters used by scholars (based on: Corti et al., 2006; Ebadian et al., 2009; Kalantari et al., 2011; Rafiei et al., 2013).
3 Research context
This research aims to develop a no-rejection time-based policy for the food processing industry. As the literature review indicates that the output variables are not yet explored in literature it is proposed to conduct a design study. A design approach helps to solve a frequently occurring practical problem and may also contribute to science (Karlsson, 2009). This section describes the methodology, the data collection & analysis, and the problem formulation.
3.1 Methodology
In this dissertation, design science is defined as follows ‘design science is research that seeks to explore new solution alternatives to solve problems, to explain this explorative process, and to improve the problem-solving process’ (Holmström & Ketokivi, 2009, p.67). Therefore, this research uses the regulative cycle of van Strien (1997) to answer the research question. The regulative cycle consists of the following phases: design problem, diagnosis/analysis, design solution, implementation and validation (see Figure 3-1).
Figure 3-‐1: Regulative Cycle (van Strien, 1997)
In this dissertation, a single case company is selected as a new phenomenon will be investigated and detailed data and insights should be extracted (Eisenhardt & Graebner, 2007). Therefore, a case company needs to be selected where data can be gathered and analysed. The focal company should be suitable for our purposes. As the problem space addresses the food processing industry, a selection criterion is that the case company is a food manufacturer. Additionally, an advantage of selecting a company from the food processing industry is that the outcomes can be related to general production strategy findings (Olhager, 2003) as well as to industry specific findings (van Donk, 2001; van Kampen, Akkerman, & van Donk, 2012). A second criterion is that the case company not rejects MTO orders.
3.2 Case company
‘internal’ because they are part of the same multinational as the food manufacturer. Currently, there are twelve internal customers at the food manufacturer, each serving one country and all requesting MTS services. Each internal customer shares information about stock levels of finished goods and forecasts per SKU with the manufacturer. The external customer is called ‘external’ because an external party is being served that is not part of the multinational. The food manufacturer has one external customer who serves the United Kingdom (UK) market. In comparison with the internal customer, the external customer requests a MTO service. The capacity required for the MTO and MTS production is equally distributed. At the manufacturer, no MTO/MTS SKU combinations exist and therefore only pure MTO and pure MTS SKUs are considered in this research. Currently, circa two hundred SKUs are produced at the food manufacturer. Furthermore, the multinational has its own supply chain centre. Due to centralization of the food manufacturer the supply chain centre determines parameters for the food manufacturer, including the ROP.
In Figure 3-2 the research scope of this project is presented. The research scope is an extension of the typical two-stage food process defined by Soman et al. (2004). At the case company and also at many food manufacturers (Akkerman, 2007) the intermediate stock point as depicted can only be stored temporarily, due to limited capacity and instability, and perishability constraints.
3.3 Data collection and analysis
The research project was performed at the food manufacturer during a period of five months. This section is described by distinguishing three periods, namely the explorative phase (design problem and diagnosis phase), the design solution phase, and the implementation and validation phase of the regulative cycle van Strien (1997).
Explorative phase (sub-question two)
During the explorative phase, qualitative and quantitative methods were used to gain insight in the challenges and opportunities when achieving a no-MTO order rejection policy. Most useful information was gathered through semi-structured interviews with: the internal customer from the Netherlands, the external customer from the UK, and with the planner, purchaser and supply chain manager of the food manufacturer. A semi-structured approach was chosen in order to avoid social desirability bias. All interviews were tape-recorded and stored in a central case database. After insights were gained through qualitative methods, quantitative methods were used to analyse aspects more in-depth.
Design solution phase (sub-question 3 & 4)
In the design solution phase, triangulation of the gathered and analysed data from the explorative phase was achieved by presenting the results to the planner, the supply chain manager, and the external customer during second interviews. In these interviews, the parameters and the practical contribution of the model were discussed. New insights were gained to improve the applicability of the no-rejection time-based policy and the decision support model. As a result, the form of the decision model is a nonlinear programming tool in Microsoft Excel, mainly chosen because of the familiar interface of Microsoft Excel, which makes the use of the tool even more suitable for the purposes of this study.
Implementation phase & validation phase (sub-question 5)
4 Findings explorative phase
In this section the sub question ‘which parameters influence customer LT in hybrid MTO – MTS systems according to practice?’ is answered. Contradictory to the theoretical findings that were addressed in Section 2, it is found that order rejection is not allowed in practical circumstances. Additionally, customer LTs at the case company consists of large variances as the customer LTs absorb the uncertainties in demand in case MTO order rejection is not allowed. This section will first address the need for short and reliable customer LT based on the interview findings. Second, it presents and explains the factors of customer LT and its parameters in more detail (see Figure 4-1).
4.1 Short and reliable customer LTs
It is stated by Soman, van Donk, & Gaalman (2007) that MTO customers generally prefer short and reliable customer LTs. This can be acknowledged based on the interview findings
with the MTO customer on the 9th of August 2013 which indicated that: (1) the MTO
customer prefers to have reliable customer LTs and (2) the MTO customer prefers to have a short customer LT for promotional driven MTO orders as they require a more market responsive supply chain as supported by Wong, Stentoft Arlbjørn, Hvolby, & Johansen (2006). For promotional driven MTO orders a customer LT of 1 or 2 weeks is desired, whereas for not promotional driven MTO orders the requested customer LT varies between 4 and 12 weeks. It is moreover indicated by the MTO customer that whenever a promotion opportunity arises, specific promotional driven SKUs are chosen to fulfil this opportunity.
Furthermore, interview findings (15th of August 2013) showed that promotional driven MTO
orders arrive on average once a month. When this happens the production planning is often rescheduled resulting in convincing employees to achieve shorter customer LT as demanded by the MTO customers. As ingredients and packaging materials are often not in stock, the suppliers need to be encouraged to deliver their supplies sooner. However, currently it occurs that SKUs are used to fulfil this promotion opportunity for which ingredients and packaging materials are in stock. For these SKUs a shorter customer LT can be achieved as the supply LT is inapplicable. At last, it can be concluded that a lot of changes in the production planning can affect operators their satisfaction. Now the need for a short and reliable customer LT in practical circumstances is indicated based on the interview findings more detailed information about customer LT and its factors is presented in the next part.
4.2 Customer LT and its factors
LT. The parameters that influence the variance of customer LT can decrease customer LT by absorbing uncertainties in demand. These are capacity modification, MTS capacity requirement, MTO order quantity, and MTO capacity reservation.
Figure 4-‐1: Customer LT and its factors
4.2.1 Supply LT
In this dissertation the supply LT of a SKU equals the longest LT of all the ingredients and packaging materials needed for that SKU. The supply LT depends on the inventory costs of both ingredients and packaging materials, and the perishability of ingredients. Also, inventory storage constraints may limit the possibilities to store ingredients and packaging materials. The overall supply LT for a SKU decreases by keeping the ingredients and packaging materials in stock that have the longest supplier LT. However, some SKUs include ingredients with low perishability. Stocking ingredients with low perishability is risky, as there is a chance that no MTO order is placed resulting in obsolete ingredients. At the case company, ingredients with high and low perishability exist and this needs to be taken into account.
4.2.2 Waiting and processing time
The waiting time is the time between the moment that ingredients and packaging materials are available for processing until the moment the MTO order is processed. The processing time is the manufacturing time needed to process the MTO order. Several parameters that influence the waiting and processing time will be discussed next.
adapting work schedules affects operators’ satisfaction (Robbins & Judge, 2010). If the maximum capacity is reached, the food manufacturer may decide to increase capacity by outsourcing activities. While it is possible at the case company to outsource, this is not preferred as it increases production costs.
A second parameter that influence customer LT is processing sequence per CT. Starting with MTO processing results in shorter customer LTs, as in this way the time between MTO order placement and earliest MTO processing time decreases. At the case company no sequence dependent setups exist between MTO SKUs and MTS SKUs. Therefore, it is possible to separate the MTO and MTS processing. In the current policy the planner also separates MTO and MTS processing. Another reason to start with MTO processing is to achieve a reliable customer LTs, as in this way a fixed starting point for MTO processing is used.
The cycle time (CT) period is the third parameter that influences customer LT. At the case company the CT period is one week with averagely 96 processing hours per CT period. The customer LTs of MTO orders vary between 1 and 12 weeks. Based on interviews and observations it is found that MTO orders are not planned the same CT period as MTO orders arrive, because then the production planning is already made and rescheduling is not desired. Therefore, a shorter CT period decreases customer LTs, at it decreases the time between MTO order placement and the processing time. Notice that before a MTO order is processed, first the supplies need to be delivered (supply LT). The MTO order can be planned before supplies have arrived by taken into account the supply LT.
The fourth parameter that influences customer LT is MTS capacity requirement. In hybrid MTO – MTS systems both the MTO and MTS category fight for the same capacity. Naturally, the capacity availability for the MTO category depends on the capacity needed for the MTS category and vice versa. MTS capacity requirement is influenced by the ROP. It is known that when the inventory level of MTS SKUs reaches the ROP, MTO orders will be released (Silver et al., 1998). At the case company the ROPs is defined in weeks. If for example the ROP is 3 weeks, the inventory level equals the sum of MTS demand forecasts of upcoming 3 weeks. Therefore, this ROP only includes the buffer time. The purchaser of the food manufacturer uses different ROPs to order the supplies. At the case company the supply chain centre of the multinational determines the ROPs per MTS SKU. This limits the opportunities to absorb uncertainties by deciding on the ROPs. However, uncertainties can still partly be absorbed by the MTS sub-categorization as the planner can decide to plan some MTS SKUs before they reach the ROP. In case capacity is constrained at a certain CT period, the planner first plans the SKUs of the MTS category that are less than or equal to the ROP that CT period. In case of idle capacity the planner also plans SKUs of the MTS category that have not yet reached the ROP that CT period. Therefore, every CT period the MTS category can be divided over two sub-categories:
At the case company, this way of absorbing uncertainties belongs to the current policy. In the current policy capacity is distributed according to the following priorities: (1) MTS≤ROP, (2) MTO and (3) MTS>ROP.
Another parameter that influences customer LT is the MTO order quantity as more processing time is needed for large MTO order quantities and vice versa. At the case company MTO orders of 45.000 units (40 processing hours) and 6.000 units (5.33 processing hours) arrive for a single SKU. Recently, even a MTO order of 162.000 units (144 processing hours) arrived for this SKU. This SKU is very promotional driven and has therefore varying MTO order quantities. Determining a maximum MTO order quantity is a way to absorb uncertainties in demand as in this way the maximum processing time needed is limited. While the case company determines minimal MTO order quantities to have long runs and low set-up times, it does not define maximum MTO order quantities. When the maximum MTO order quantity is determined also the corresponding MTO arrival rate should be determined.
A last parameter that influences customer LT is MTO capacity reservation. MTO capacity reservation is a way to absorb uncertainties. In case the time between MTO order placement and the requested delivery date is short, a company can decide to reserve capacity. If customer LTs of MTO orders are long, these can be planned based on the actual processing time needed for these MTO orders. MTO orders that correspond with a short customer LT are often promotional driven MTO orders. If capacity is not reserved for these orders, the planner has to reschedule the production planning to achieve the corresponding customer LTs. It is risky to reserve capacity for the promotional driven MTO orders, as they have often very varying MTO order quantities.
4.2.3 Transportation time
The transportation time is the time needed to transport SKUs from the food manufacturer to the customer. The transportation time depends on the geographical distance between the locations and the type of transport (Riezebos, 2006). Moreover, it is assumed that the transportation time is constant as it is stated by Riezebos (2006) that delivery has become a dependable process in many situations.
To conclude, practical findings indicate that a quick response to the market is desired for promotional driven MTO orders. These MTO orders often arrive in varying quantities and for these orders a short customer LT is generally requested. The customer LT consists of the factors supply LT, waiting time and processing time, and transportation time. The parameters that influence the waiting and processing time are: (1) capacity modification, (2) processing sequence, (3) CT period, (4) MTS capacity requirement, (5) (maximum) MTO order quantity, and (6) MTO capacity reservation. These parameters serve as input parameters for the no-rejection time-based policy proposed in the next section.
5 No-rejection time-based policy
Based on the findings from practice and theory, this section answers sub-question three by proposing a no-rejection time-based policy. First, the parameters that can absorb uncertainties in demand and decrease variance in customer LT are analysed. Second, the process steps of the no-rejection time-based policy are presented followed by a mathematical model that assesses the relation between capacity availability, capacity requirements, customer LTs and MTS service levels per CT period.
5.1 Analysis of the parameters
In a no-rejection time-based policy reliable customer LTs can be achieved if the parameters that influence customer LT can (fully) absorb uncertainties in demand. Practical findings have shown that the variance of customer LT can decrease through the parameter ROP, MTS sub-categorization, capacity modification, maximum MTO order quantity, and MTO capacity reservation (see Figure 4-1).
The first parameter that influences the variance of customer LT is the ROP. Both theoretical and practical findings have shown this. However, deciding on the ROP does not influence the variance of customer LT at the case company as the supply chain centre determines the ROPs. At other food manufacturers the ROP may influences the variance of customer LT and therefore this presents some interesting research opportunities The MTS sub-categorization also influences the MTS capacity requirements. This parameter is already included in the current policy and will also be used in the no-rejection time-based policy. The second parameter is capacity modification. At the case company it is only allowed to adapt work schedules a couple of times a year for two reasons. Firstly it effects operators’ satisfaction and secondly because outsourcing activities are limited as it increases production costs. This limits the possibilities to modify capacity. Therefore, the starting point for this dissertation is to develop a no-rejection time-based policy in which capacity is limited and cannot be modified. The third parameter is the maximum MTO order quantity. As mentioned previously, by determining the maximum MTO order quantity the variance of customer LT can be absorbed. The last parameter is MTO capacity reservation. At the case company the requested customer LT of the MTO customer are long enough to plan the MTO quantities that are ordered. In case a promotional driven MTO order arrives, a fast response is requested and in this case the production plan is often rescheduled. As the occurrence of MTO orders is highly uncertain, capacity reservation for the promotional driven orders is rather risky (van Donk, Soman, & Gaalman, 2003) and is therefore left out of scope.
are included in the no-rejection time-based policy to absorb uncertainties, both parameters cannot fully absorb uncertainties in demand.
An option left to fully absorb uncertainties in demand is having varying customer LT. Practical findings have shown that the MTO customer prefers to have short customer LTs for promotional driven MTO orders. Additionally, it was found that some SKUs are more promotional driven than others. As a short customer LT is desired for these SKUs, uncertainties in demand should not be absorbed with promotional driven SKUs. Therefore, this dissertation proposes to divide the MTO SKUs over two categories, including a MTO fast response category (FR) and a MTO slow response category (SR). The FR category
corresponds with a fixed and short customer LT, whereas SR category corresponds with
several variable and longer customer LTs. This division is proposed in a way that uncertainties in demand are merely absorbed by the SR category and the FR category is not affected. These categories combined with the MTS sub-categorization results in the following categories: • FRcategory • SRcategory • MTS category o MTS≤ROP sub-category o MTS>ROP sub-category
In the no-rejection time-based policy the MTS category is divided over a MTS ≤ ROP and MTS > ROP sub-category. Note that the MTO, FR and SR categorization is SKU dependent, whereas the MTS≤ROP and MTS>ROP sub-categorization depends on the inventory level per MTS SKU per CT period.
5.2 Process steps
Absorbing uncertainties in demand can result in increasing and varying customer LTs for the SR category. Also, the service levels of the MTS SKUs can be affected. To minimize these negative effects and subsequently achieve a short customer LT for the FR category, the process steps for the no-rejection time-based policy are defined.
Figure 5-‐1: Current production planning
Because the FR customer LT is short and MTO orders are highly uncertain, capacity reservation is risky. Therefore this category cannot be planned in advance. Rescheduling the current production planning is necessary when a FR order arrives (see Figure 5-2). Capacity is in this dissertation assumed to be limited and therefore priorities should be determined to distribute capacity over the (sub-) categories when a FR order arrives. The processing time needed per category should be recalculated according to the following capacity distribution priorities:
Priority 1: FRcategory
Priority 2: MTS ≤ ROP sub-category
Priority 3: SR category
Priority 4: MTS > ROP sub-category
The FR category gets the highest priority in the no-rejection time-based policy, because a fixed customer LT corresponds with this category. The MTS≤ROP sub-category gets second
priority in order to not affect MTS service levels. The SRcategory gets priority three so that
the corresponding customer LTs can be achieved. The MTS>ROP sub-category gets the lowest priority as this sub-category increases finished goods inventory and thus inventory costs. The following processing sequence should be used when developing the production planning: (1) FR, (2) SR, and (3) MTS. The FR category is processed at the beginning of the CT period in order to achieve a short and reliable customer LT. The (sub-) categories for which not enough capacity is available should be rescheduled to next CT period.
Figure 5-‐2: Rescheduled no-‐rejection time-‐based production planning
For the no-rejection time-based policy the MTO customer and manufacturer should agree on the FR – SR categorization. Therefore, a prerequisite for the no-rejection time-based policy is collaboration between the MTO customer and the manufacturer. This is in line with the no-rejection policy, because a long-term agreement entails the no-no-rejection policy but also enables collaboration between the customer and manufacturer. Also, a better value to the customer is delivered if customers and manufacturers co-operate (Kurnia & Johnston, 2001). For the FR category the corresponding customer LT should be determined. While different factors and parameters influence the customer LT, the FR customer LT should be assessed based on the supply LT and the CT period. The supply LT can decrease by stocking ingredients and packaging materials with the longest suppliers LT. Stocking ingredients with low perishability is risky, hence, only the SKUs that do not include low perishability ingredients should be put in the FR category. Furthermore, a short CT period should be defined to achieve the short FR customer LT. In case the requested customer LTs cannot be achieved, the MTO customer can still decide to categorize other SKUs in the FR category. Another option is to move these SKUs from a MTO to a MTS policy (Soman, 2005). In this case, the manufacturer will deal with a fast replenishment policy rather than a fast response policy. The amount of ingredients and packaging materials to stock depends on the FR order quantities. FR order quantities should be defined in order to determine how much ingredients and packaging materials to stock. This dissertation merely focuses on the quantity of the FR category as a whole (FR order quantity), rather than determining maximum order quantities per SKU.
However, if total processing time needed (input) exceeds total processing time available (output) per CT period for a certain period, the customer LTs will keep increasing and service levels will keep decreasing over time. Therefore, the maximum FR order quantity should be determined.
As many parameters play a role in the no-rejection-time-based policy a mathematical model is developed to analyse the interaction between the parameters based on input parameters. An estimation of the maximum FR order quantity can be derived with the results of the mathematical model by defining several scenarios.
5.3 Mathematical model
This section presents the Nonlinear Programming Excel tool for the no-rejection time-based policy that assesses the relationship between capacity availability, capacity requirements per category, customer LTs, and MTS service levels. In case not enough capacity is available for a (part of the) (sub-) category, this (sub-) category is rescheduled to next CT period. The Nonlinear Programming Excel tool calculates how much processing time in hours is rescheduled per (sub-) category. The model does not include the time between MTO order placement and processing time, and transportation time.
The assumptions of the no-rejection time-based policy are: • MTO order rejection is not allowed
• SKUs are classified over the FR, SRand MTS category
• One fixed customer LT is assumed the FRcategory
• One variable customer LT is assumed the SR category • One fixed ROP is assumed for the MTS category
• Capacity availability in hours per CT is fixed and known
• Processing times needed in hours per SKU per CT period are known • Inventory levels per CT period are known
• Inventory levels are accurate
• Capacity is distributed according to the following priorities:
§ Priority 1: FRcategory
§ Priority 2: MTS ≤ ROP sub-category § Priority 3: SR category
§ Priority 4: MTS > ROP sub-category
It is assumed that the following parameters are known:
: Length of CT period (unit time)
LTFR: Customer LT of the FRcategory (unit time)
LTSR: Customer LT of the SRcategory (unit time)
: ROP of MTS category (unit time)
It is assumed that for each CT period t the next parameters are known:
: Capacity availability per CT period t (unit time)
PTit: Processing time needed in hours per SKU i per CT period t (unit time)
: Percentage of MTS category that is less than or equal to the ROP at CT
period t (0 ≤ Xt ≤1)
The subsets ,FR, and SR are elements of the SKUs i=1,…,N:
i ∈ MTS i ∈ FR i ∈ SR
The subsequent equations are used to calculate the processing time needed in hours (capacity requirements) per (sub-) category per CT period t.
PTMTSt= PTit
i∈MTS
∑
: MTS capacity requiredPTFRt= PTit
i∈FR
∑
: FR capacity required (FR order quantity) PTSRt= PTit
i∈SR
∑
: SR capacity requiredPTTOTt= PTMTSt+ PTFRt+ PTSRt : Total processing time needed (input)
The demand in hours for the MTS≤ROP and MTS>ROP sub-categories at CT period t are calculated with the following equations:
The processing time needed per category at CT period t depends on the rescheduled processing hours of this category at CT period t-1. The processing times at CT period t plus the rescheduled processing hours at CT period t-1 per category are indicated with
RPTFRt, RPTMTS≤ROPt, RPTSRt andRPTMTS>ROPt.
The objective function of the Nonlinear Programming tool for the no-rejection time-based policy is to maximize:
Zt= at10000 + bt1000 + ct100 + dt10
The objective function (Zt) includes the capacity distribution priorities and ensures that
negative effect regarding MTS service levels and customer LTs are minimized. The largest number (10000) is assigned to the decision variable corresponding with the category that has the highest priority (FR). The second large number (1000) is assigned to the decision variable corresponding with the category that has the second priority (MTS≤ROP), et cetera.
Subject to:
≥ atRPTFRt+ btRPTMTS≤ROPt+ ctRPTSRt+ dtRPTMTS>ROPt
≥ 0 ≤ 1 The decision variables are:
: Percentage of capacity distributed to RPTFRt at period t.
: Percentage of capacity distributed toRPTMTS≤ROPt at period t.
: Percentage of capacity distributed to RPTSRt at period t.
: Percentage of capacity distributed to RPTMTS>ROPt at period t.
As the processing time needed per category at CT period t depends on the rescheduled processing hours per category at CT period t-1 the following equations are used:
RPTFRt= PTFRt+ (1− at−1)RPTFRt−1
RPTMTS≤ROPt= PTMTS≤ROPt+ (1− bt−1)RPTMTS≤ROPt−1
RPTSRt= PTSRt+ (1− ct−1)RPTSRt−1
RPTMTS>ROPt= PTMTS>ROPt+ (1− dt−1)RPTMTS>ROPt−1
With the results of the model (at, bt, ct,, dt) the capacity distributed (CD) to the categories per CT period t can be calculated. These represent the output per category and are calculated with the following equations:
CDFRt= atRPTFRt
CDMTS≤ROPt= btRPTMTS≤ROPt
CDSRt= ctRPTSRt
CDMTS>ROPt= dtRPTMTS>ROPt
If a category is (partly) rescheduled to next CT period the response time increases. In this dissertation the extra response time in case a category is (partly) rescheduled to next CT period is called rescheduled time (RT). The rescheduled time depends on the rescheduled time of the previous period. As the mathematical model does not show which SKUs of a category are rescheduled, this dissertation supposes to calculate the rescheduled time per CT period t with the following equations:
RTFRt= (1− at)CT + RTFRt−1
RTMTS≤ROPt= (1− bt)CT + RTMTS≤ROPt−1
RTSRt= (1− ct)CT + RTSRt−1
RTMTS>ROPt= (1− dt)CT + RTMTS>ROPt−1
If the rescheduled time of the MTS≤ROP sub-category equals one CT period for a certain period, service levels may be affected. Therefore, this category should be minimized. For instance, at the case company the ROP included 3 weeks of buffer time. If the MTS≤ROP sub-category is rescheduled at three subsequent CT periods, the service levels of MTS SKUs are affected. This dissertation assumes that in case the FR and SR category are rescheduled, the rescheduled customer LT of the FR and SR category can be calculated by adding rescheduled times to customer. This following equations are used:
RLTFRt = LTFR+ RTFRt
RLTSRt = LTSR+ RTSRt
6 Validation
This section is divided in two parts. The first validates the no-rejection time-based policy by comparing the current and the no-rejection time-based policy based on historical data extracted from the case company. The historical data is used as input data for the mathematical model. The second part shows how the mathematical can be used to determine the maximum FR order quantity. Determining the maximum FR order quantity is important as it is expected that the FR order quantity will increase with the no-rejection time-based policy.
6.1 Performance no-rejection time-based policy
In order to assess the performance of the no-rejection tine-based policy the current policy should be compared with the no-rejection time-based policy. To compare both policies, similar historical data per CT t is obtained over a period of 42 weeks. For the no-rejection time-based policy the processing time needed for the MTO category is divided over a FR and SR category depending on the FR – SR categorization. The historical data serve as input parameters for the mathematical model.
For the current policy notations, subsets and formulations should be defined for the MTO and MTS category. The following notations are used for the current policy:
: Customer LT of the MTOcategory
In the current policy the subsets and are elements of SKUs =1,..,N.
The subsequent formulations are used to calculate the capacity requirements (processing time needed) per category per CT period t:
= = =
For the current policy the following input parameters are assumed:
And for the no-rejection time-based policy the following:
= 1 wk.
LTFR= 1 wk.
LTSR= 3 wks.
= 3 wks.
The results of the mathematical model for the current policy are illustrated in Figure 6-1 and 6-2. These graphs present the cumulative processing time needed (input) and cumulative capacity distributed in hours (output) per CT period t.
Figure 6-‐1: Current policy: input/ output curves
Figure 6-‐2: No-‐rejection time-‐based policy: input/ output curves
The performance of the policies is measured by calculating the response time between the input and the output. The response time is derived from the figures by subtracting the CT period t of an input curve from the CT period t of an output curve. Figure 6-3 and Figure 6-4 present a close up of both policies. In these figures the response time of the MTO categories are indicated with arrows.
Figure 6-‐3: Close up of current policy: input/ output curves
Figure 6-‐4: Close up no-‐rejection time-‐based policy: input/output curves
The exact input and output data per category per CT period t is provided in Appendix 9-1. Figure 6-3 shows that in the current policy the customer LT of the MTO category is 3 weeks at CT period t=9. Figure 6-4 shows that in the no-rejection time-based policy the customer LT of the FR and SR category at CT period t=9 are 1 and 3 weeks, respectively. The exact
RLTFRt and RLTFRt) can be calculated with the equations given in section 5.3. The results of both the current policy and the no-rejection time-based policy are presented in Table 6-1 and Table 6-2. The tables show the frequency per RLT/RT intervals over a period of 42 weeks.
RLT/RT intervals (weeks) Frequency
RTMTS≤ROP 0 42
RLTMTO 3 42
RTMTS>ROP 0 37
0,2 4
0,3 1
Table 6-‐1: Current policy: frequency per LT or RT interval
LT or rescheduled time (RT) (weeks) Frequency
RLTFR 1 42 RTMTS≤ROP 0 42 RLTSR 3 42 RTMTS>ROP 0 32 0,1 1 0,3 2 0,4 2 0,6 3 0,7 1 0,9 1
Table 6-‐2: No-‐rejection time-‐based policy: frequency per LT or RT interval
Table 6-1 shows results of the mathematical model based on the current policy. In the current policy the RLTMTO and RTMTS≤ROP were 42 times, 3 and 0 weeks, respectively. The MTS>ROP sub-category was rescheduled 5 times. In the no-rejection time-based policy the customer RLTFR, RLTSR and RTMTS≤ROP were 42 times, 1, 3 and 0 weeks. The RTMTS>ROP was rescheduled 10 times.
The results have shown that based on historical data the no-rejection time-based policy will only rescheduled the MTS>ROP sub-category. Rescheduling this category is seen as a positive result, as this category increases finished goods inventory. Furthermore, it can be said that with the no-rejection time-based policy overall a shorter customer LT is achieved as the FR and SR customer LT of 1 and 3 weeks, respectively, are always achieved over a period of 42 weeks. Last, the MTS service levels did not decrease, as also this category was never rescheduled to next CT period.
6.2 Determining maximum FR order quantity
Determining the FR order quantity depends on the other input parameters. Therefore, historical data should serve as input parameters for the mathematical model. For the FR
category the following FR order quantities (PTFRt) arriving with rate λFR are determined:
Scenario 1: PTFRtis 30 hours and arrives at rate λFR=0,5. Scenario 2: PTFRtis 60 hours and arrives at rate λFR=0,5. Scenario 3: PTFRtis 90 hours and arrives at rate λFR=0,5.
Figure 6-5 presents the total input and total output of the sum of the categories for each scenario (1, 2 and 3). The response time (average of RT and RLT) between input and output for the scenarios 1, 2 and 3 at CT period 9 is indicated with the arrows in Figure 6-5.
Figure 6-‐5 No-‐rejection time-‐based policy: input/ output for scenarios 1,2 and 3
An important result is that the input and output curve diverge if the quantity (processing time needed) of the FR category increases from 30, to 60, to 90 processing hours. If input is larger than output for a certain period, the response time increases over time. The maximum FR
order quantity arriving with rate λFR should be determined at the point at which the input
curve is parallel to the output curve. To determine the maximum FR order quantity, the following scenarios are defined:
Scenario A: PTFRtis 40 hours and arrives at rate λFR=0,5. Scenario B: PTFRtis 50 hours and arrives at rate λFR=0,5. Scenario C: PTFRtis 60 hours and arrives at rate λFR=0,5. Scenario D: PTFRtis 70 hours and arrives at rate λFR=0,5.
Figure 6-6 shows the input for the four scenarios (A, B, C and D). It also presents the trend lines per input/output curve. The output curve is represented with the maximum output based on the maximum capacity per CT period t.
0 200 400 600 800 1000 1200 1400 1600 0 2 4 6 8 10 Inp u t/ o u tp u t ( hou rs ) CT periods t (weeks)
Figure 6-‐6 No-‐rejection time-‐based policy: input/ output curves for scenarios A, B, C and D.
A result of Figure 6-6 is that the input and output curve is most parallel for scenario B. Therefore, the maximum FR order quantity given in processing hours needed should be around 50 hours. Only an estimation of the FR order quantity can be given with the mathematical model, as it is determined with the defined scenarios.
This section has answered sub-question four by addressing the performance of the no-rejection time-based policy based on historical data. First, it showed that only the last category is rescheduled if the no-rejection time-based policy is applied to historical data at the case company. This means that the FR and SR customer LT of 1 and 3 weeks respectively, were always achieved over a period to 42 weeks. Also, MTS service levels remained the same. Furthermore, finished goods inventory decreases with the no-rejection time-based policy as only the MTS>ROP sub-category is rescheduled. The second part of this section have shown that determining a maximum FR quantity minimizes negative effects in a no-rejection time-based policy, regarding customer LTs and MTS service levels.
-‐200 0 200 400 600 800 1000 1200 1400 1600 0 2 4 6 8 10 Inp u t/ o u tp u t ( hou rs ) CT periods t (weeks) Maximum capacity Scenario A Scenario B Scenario C Scenario D Lineair (Maximum capacity)
7 Discussion
This dissertation questioned the MTO order acceptance/rejection decision that is stated in several studies for the complex hybrid MTO – MTS systems. It determines that order rejection is not allowed in practice if long-term agreements with MTO customers exist. Specifically, it argues that if uncertainties in demand are not absorbed by the MTO order acceptance/rejection decision, a rejection time-based policy should be applied. In a no-rejection time-based policy reliable customer LTs can be achieved if the parameters that influence customer LT can (fully) absorb uncertainties in demand. In the remainder of this section, the results are discussed.
7.1 Findings explorative phase
Theoretical findings have shown that the parameter ROP can absorb uncertainties in demand. In case the ROP is higher, uncertainties in demand can be absorbed and variance in customer LT may decrease. While several studies have mentioned this (Kalantari et al., 2011; Rafiei et al, 2013), the role of the ROP in hybrid MTO – MTS systems is not completely understand (Soman et al., 2004). In this dissertation the ROP is left out of scope, as at the case company they cannot decide on the ROP because the supply chain centre determines this. It is expected that in case the ROP is included in the no-rejection time-based policy variance in customer LTs decrease even more. Furthermore, a larger maximum FR order quantity may be determined while achieving the same customer LTs and service levels. However, a high ROP also results in increasing finished goods inventory. Investigating the role of the ROP the no-rejection time-based policy is therefore suggested as a further research topic.
7.2 No-rejection time-based policy
The no-rejection time-based policy suggests dividing all SKUs over a FR category, SR category, and MTS category. The policy argues that the FR customer LT is short and fixed and for the SR customer LT is long and variable. The SR customer LTs should absorb uncertainties in the no-rejection time-based policy. As MTO customers prefer having reliable customer LTs (Iravani et al., 2009), variance in customer LT is minimized in the no-rejection time-based policy.
It can be said that the focus of this dissertation is to minimize negative effects regarding customer LTs and MTS services levels, rather than finding the optimal situation. The reason for this is that the policy is constrained by the fact that order rejection is not allowed. The MTO order acceptance/rejection is a way to increase profitability as only profitable orders can be accepted. Also, is it a way to increase MTO customer satisfaction when only orders are accepted based on the importance of MTO customers. In case MTO order rejection is not allowed, profitability and customer satisfaction may decrease. Therefore, the challenge of the no-rejection time-based policy is that negative effects are minimized.
Another point may be that the FR-SR categorization has an advantage and disadvantage. On the one hand, the FR category of the no-rejection time-based policy satisfies MTO customers, because a fixed FR customer LT is offered. While on the other hand, the SR category decreases satisfaction as a variable customer LT is offered. However, the variance of the SR customer LT is minimized by the determination of the FR order quantity. Besides, the no-rejection time-based policy meets the needs of the MTO customer by offering a short customer LT for the promotional driven orders. Practical findings have shown that the no-rejection time-based policy also increases food manufacturer satisfaction. An explanation for this is that the food manufacturer expects that uncertainties in the SR category decrease with the no-rejection time-based policy.
There is a need to decrease uncertainties in demand through integration within the supply chain (van Donk & van der Vaart, 2005). In the no-rejection time-based policy these uncertainties decrease by determining the FR order quantity. However, in this way the no-rejection time-based policy only decreases uncertainties in the FR category. Uncertainties in the SR category will remain the same in the policy. However, it is expected that in the no-rejection time-based policy the MTO customer will only offer promotions from the FR category and that therefore, also the uncertainties in the SR category will decrease.
7.3 Validation
8 Conclusions, limitations and further
research
While most research is based on the assumption that order rejection is possible, MTO order rejection is not always allowed in practice. When uncertainties in demand cannot be absorbed by the MTO order acceptance/rejection decision, increasing and varying customer LTs appear. Therefore, this dissertation has proposed a no-rejection time-based policy for hybrid MTO – MTS systems to support the food manufacturers that are not able to reject orders. Returning to the sub-questions introduced at the beginning of this dissertation it is possible to state the following conclusions.
Extant theory describes the influence of the parameters ROP and capacity modification on the variance of customer LT. Additionally, practical findings have shown that the variance of customer LT is influenced by the parameters: (1) ROP, (2) MTS sub-categorization, (3) capacity modification, (4) maximum MTO order quantity, and (5) MTO capacity reservation. The parameters that are included in the no-rejection time-based policy are limited to: MTS sub-categorization and maximum MTO order quantity. To conclude, practical findings indicate that a quick response to the market is desired for promotional driven MTO orders. Based on the findings from both practice and theory, the no-rejection time-based policy suggests to divide all SKUs over three categories: MTS, FR and SR. The customer LT of the FR category is short and fixed and for the SR category longer and variable. As many parameters play a role in the no-rejection time-based policy a mathematical model is developed to present the interaction between the capacity availability, capacity requirements per category, customer LTs and MTS service levels.
When testing the policy with the mathematical model based on historical data extracted from the case company, the results show that the no-rejection time-based policy only affects the MTS>ROP sub-category. This means that the FR and SR customer LT are always achieved and MTS service levels do not decrease. Also, is shown that determining a maximum FR quantity prevent the no-rejection time-based policy of having negative effects regarding customer LTs and MTS service levels.
The findings of this study have four important implications for further practice. First, as the no-rejection time-based policy is developed based on practical findings of one single case company, caution must be applied, as the findings might not be transferable to all other food manufacturers. Second, the FR-SR categorization may depend on the type of order rather than SKU. If the FR-SR categorization is order depended, supply LT cannot decrease. Third, another implication is that only the maximum order quantity for the FR category is included, while also the amount of ingredients and packaging materials per SKU should be known. Last, in the mathematical model it is not taken into account that the inventory level of the MTS>ROP sub-category decreases if this is (partly) rescheduled, because this was too complex. This therefore needs to be seen as a limitation for this study.
