Capacity-based product classification in
hybrid production systems
Master Thesis
Maarten Christoffel Marie Vennix S3027163
Master thesis, MSc Technology & Operations Management University of Groningen, Faculty of Economics and Business
January 2018 Supervisor: dr. O.A. Kilic Co-Assessor: dr. D.P. Van Donk
Abstract
Purpose – In a hybrid production system, some products are produced with a Make-To-Order (MTO)
strategy, while other products are produced with a Make-To-Stock (MTS) strategy. Hence, a key decision in these systems is which products to produce with an MTO or MTS strategy. Commonly used approaches asses the products’ MTO/MTS suitability on a product-by-basis. However, it is well established that MTO production requires more capacity than MTS production. As a result of assigning products to either strategies one-by-one, the MTO/MTS classifications yielded by these approaches do not take the resulting capacity requirements of the entire product portfolio into account. The purpose of this research is to develop a novel MTO/MTS classification approach, where the allocation of products to an MTO or MTS strategy respects the production system’s optimal levels of MTO and MTS production. Through applying this approach the a case in the food processing industry, we explore the system’s performance under different levels of MTO and MTS production.
Method – This work employs a simulation study at a case in the food processing industry to
demonstrate the workings of the proposed MTO/MTS classification. As a result, a production system’s performance under different levels of MTO/MTS production is observed in a real-life setting.
Findings – It was found that the performance of a hybrid production system is very dependent on the
Table of content
1. Introduction ... 1
2. Background ... 3
2.1. MTO/MTS and capacity requirements ... 3
2.2. MTO/MTS classification approaches ... 3
2.3. A novel MTO/MTS classification approach ... 5
3. Operationalization of classification approach ... 6
3.1. MTO/MTS suitability metric ... 7
3.2. Determination of cut-off value ... 8
4. Experimental setup ... 8
4.1. Case description ... 9
4.2. Simulation design ... 12
4.3. Experimental design ... 16
4.3.1. Single-metric MTO/MTS suitability ... 16
4.3.2. System performance under different levels of MTO/MTS production... 17
4.3.3. Changing levels of MTO/MTS production to cope with a capacity change ... 18
4.3.4. Operationalization of output measures ... 19
5. Results ... 20
5.1. Single-metric MTO/MTS suitability ... 20
5.2. System performance under different levels of MTO/MTS production ... 22
5.3. Changing levels of MTO/MTS production to cope with a capacity change ... 24
6. Discussion ... 26
6.1. Single-metric MTO/MTS suitability assessment ... 26
6.2. System performance under different levels of MTO/MTS production ... 26
6.3. Changing levels of MTO/MTS to cope with a capacity change ... 27
6.4. The application of the proposed MTO/MTS classification approach ... 28
7. Conclusion ... 28
7.1. Implications and further research ... 29
7.2. Limitations... 29
1
1. Introduction
Production systems are often characterized as either Make-To-Order (MTO) or Make-To-Stock (MTS). However, as companies produce a wider range of different products, the adoption of a hybrid production system, where both strategies are combined, has become more common (Claassen et al., 2016). The benefits of hybrid systems are that to minimize holding costs and limit product obsolescence, for some products an MTO strategy is used. However, when all products are MTO, demand peaks cannot be smoothened causing demand to be missed during busy periods and a low utilization during periods of low demand. When some products, preferably those with longer shelf lives and predictable demand, are produced with an MTS strategy, then these products do not require production capacity in periods of high demand as they can be delivered from stock, while capacity is utilized in periods of low demand to replenish their inventories. These hybrid systems have received a lot of attention, and the benefits of this production strategy are well established (Van Donk, 2001; Soman, Van Donk & Gaalman, 2004; Nagib et al., 2016).
In the hybrid system described above, one of the key decisions is which products to produce with an MTO strategy, and which products to produce with an MTS strategy (Soman, Van Donk & Gaalman, 2004). In current academic literature, this MTO/MTS decision is often made by evaluating the products’ MTO/MTS suitability on a product-by-product basis. Characteristics relevant for the MTO/MTS decision are identified, and products are assigned to MTO or MTS production based on these characteristics. The cut-off values of these characteristic, determined to distinguish products as MTO or MTS, are often determined in an ad-hoc manner, are case-specific, or remain unspecified in academic literature (Bachetti et al., 2013). An example of this is a case study by Van Kampen & Van Donk (2014), where the authors conducted a case study to study how often an MTO/MTS classification should be revised over time. Their case company assigns all products to MTO if they have an unreliable payment history, or when their forecast accuracy is below 80%.
2 been largely overlooked in previous work (Soman, Van Donk & Gaalman, 2004). Since that time of writing, the author of this work has not identified papers that explicitly address the observation made by these authors. Furthermore, both the above mentioned literature review by Bachetti et al. (2013) and the findings of Van Kampen & Van Donk (2014) indicate that the MTO/MTS partition is often still made on a product-by-product basis and uses arbitrarily determined cut-off values. Hence, the resulting capacity requirements are a direct result of these cut-off values, and the levels of MTO and MTS production the classification yields are not explicitly considered.
Thus, previous research has produced insights on which characteristics should be used to determine a single product’s MTO/MTS suitability. However, when an MTO/MTS classification is made on a product-by-product basis, capacity requirements might increase, such that the manufacturing system’s performance suffers as a result. Current classification approaches do not take the capacity requirements a product classification yields into consideration. Therefore, the aim of this paper is to develop a novel approach to the MTO/MTS decision problem, which bridges the gap between what is known about the characteristics to be employed in the decision making, and the effects of the levels of MTO and MTS production in a manufacturing system. By analyzing a manufacturing system’s performance under different levels of MTO and MTS production, we aim to demonstrate the importance of considering these levels explicitly when a product classification is made.
The MTO/MTS decision approach developed in this work is employed at a case company in the food processing industry to demonstrate its workings. The setting of the food processing industry is chosen as hybrid systems are often employed in that industry, due to industry-specific characteristics such as, for instance, varying product shelf lives in a company’s product portfolio. Furthermore, the case company’s specific production setting allows us to demonstrate the advantages of incorporating capacity considerations in the MTO/MTS classification decision. The manner in which this will be done is described later in this work.
3
2. Background
This background section will review the academic literature in order to synthesize the topics of capacity allocation to MTO and MTS products, and the commonly employed MTO/MTS classification approaches. First, the relation between MTO production and capacity requirements is further introduced, after which commonly used MTO/MTS classification techniques are reviewed. Then, integrating these topics, a novel MTO/MTS classification approach is proposed.
2.1.
MTO/MTS and capacity requirements
In an MTS system, production is based on forecasted demand levels, and the finished goods are stocked. This buffer of finished goods decouples the production system from fluctuations in market demand. Hence, as demand variations are smoothened by the inventory buffer, the main focus of an MTS system is the efficient use of available capacity. The number of setups can be minimized by producing in large batches, reaping the benefits of Economies of Scale. On the contrary, in an MTO system, capacity is used as a buffer to be able to meet fluctuating market demand. As no inventory is held, any market variance is directly imposed upon the production system, and the manufacturing system should have the capacity to be able to follow fluctuating demand levels. Hence, customer demand dictates the production run lengths and the resulting number of setups. Although this decreases the production system’s efficiency, the system does not requires stocks to be held, limiting holding costs and the risk of product deterioration. Therefore, when combining these two strategies in a hybrid production system, a capacity trade-off should be made, balancing the higher efficiencies of MTS products with the volatile and less efficient properties of MTO products. Soman, Van Donk and Gaalman (2004) concluded that the interaction effects between products assigned to MTO and MTS production and available capacity are captivating, yet not understood well. Subsequently, they called for more research on this topic. The following subsection will review the literature on how the MTO/MTS decision is made in current academic literature, to determine if and how the aforementioned call to research was heeded.
2.2.
MTO/MTS classification approaches
4 classification. Multiple characteristics may be selected, each with their own cut-off value. Hence, based on the number and nature of the characteristics employed (qualitative/quantitative), a technique should be selected in order to reconcile multiple characteristics. This subsection continues by reviewing commonly used characteristics, the determination of their cut-off values, and the techniques used to reconcile multiple characteristics in developing an MTO/MTS classification.
The characteristics employed depend on the aim and context of the inventory classification. For the specific aim of MTO/MTS classification, the most commonly used characteristics are weekly demand volume and the coefficient of demand variation (D’Alessandro and Baveja, 2000; Huiskonen, Niemi and Pirttilä, 2003; Beemsterboer, Land and Teunter, 2017) and forecast accuracy (Van Kampen and Van Donk, 2014). The characteristics mentioned are employed often in classification schemes. However, additional characteristics are employed when they are deemed to be of high importance to the MTO/MTS decision in a specific context. For instance, product shelf life is an important MTO/MTS characteristic in the food processing industry (Van Donk, 2001), and therefore is often employed when making an MTO/MTS classification in that context. As the characteristics to be employed in an MTO/MTS classification is not the main aim of this paper, we refer to Bacchetti et. al (2013) and Van Kampen, Akkerman & Van Donk (2012) for a broader overview of commonly used characteristics. However, for the purpose of this paper, we conclude that multiple characteristics are of relevance to MTO/MTS classification, which can be jointly employed.
After the characteristics to be employed in the MTO/MTS classification are selected, their cut-off values are determined. The cut-off values of these characteristics are often case-specific. For instance, demand for products across a portfolio often follows the Pareto-principle where 20% of the products constitute 80% of total demand volume. Therefore, the cut-off value between MTO/MTS is often set so that this 20% of products are produced with an MTS strategy, while the other 80% is produced with an MTO strategy (Cavalieri et al., 2008; Bachetti et al., 2013).
5 production strategies are determined per quadrant. A second commonly observed decision technique is that of the decision tree, where the classification is performed by assessing the products’ MTO/MTS suitability per characteristic in a stepwise fashion (Van Donk, 2014). Also here, the cut-off values are determined per characteristic.
It becomes clear from the above that both in the current MTO/MTS classification literature and in practice, often multiple characteristics are employed to base the MTO/MTS decision on. The products are assigned to either MTO or MTS production based on their value on these characteristics one by one. In following that approach, the portfolio’s required capacity as a result of the level of MTO/MTS production, as described in subsection 2.1, is neglected. Hence, we conclude that after the call for research by Soman, Van Donk & Gaalman (2004), the contemporary classification literature still approaches the MTO/MTS decision on a product-by-product basis, instead of also respecting the capacity constraints of the production system.
2.3.
A novel MTO/MTS classification approach
From the above, we concluded that MTO/MTS classifications are often made on a product-by-product basis, and ignore the capacity requirements of the resulting classification. Therefore, in this section a novel MTO/MTS classification approach will be proposed. We submit that although certain characteristics are valid measures of assessing a product’s MTO/MTS suitability, the question of how much capacity to allocate to MTO and MTS production should not be directly connected to subjectively drawn cut-off values of these characteristics. Hence, we propose a classification approach where the products’ MTO/MTS suitability is determined based on certain characteristics, but where decision on the production capacity devoted to MTO and MTS production is made based on the optimal capacity load between MTO and MTS production. Thus, MTO/MTS suitability is determined by the appropriate characteristics, but we submit that the cut-off value should be determined by identifying the optimal MTO/MTS capacity load.
6 Figure 1: Framework for MTO/MTS classification
In Figure 1 above, the bubbles represent the individual products, of which the size represents a product’s capacity load. This capacity load is needed to be taken into account, as moving a single product from an MTO to an MTS strategy may have a small or large impact on the capacity requirements, based on a product’s demand volume and production speed. Hence, the proposed MTO/MTS classification approach requires the answering of two separate questions. First, how to score the products’ MTO/MTS suitability on a single dimension, and secondly the question of what the optimal levels of MTO and MTS production in a system are. Methods for operationalizing these questions are developed in the next section.
3. Operationalization of classification approach
In the previous section, a novel MTO/MTS classification approach was presented. In order to demonstrate its workings, the framework will be applied at a case company in the food processing industry. By conducting this research at a case company, the appliance of the MTO/MTS classification approach is demonstrated in a real-life context.
7
3.1.
MTO/MTS suitability metric
As described in the background of this paper, we require a single metric based on multiple characteristics with which to assess products’ MTO/MTS suitability. We identified no such metric in previous research, as characteristics were often evaluated in a step-by-step manner. However, in previous research, MTO/MTS suitability metrics based on a single characteristic have been used. Therefore, we will develop a single metric based on multiple characteristics, and will compare its resulting MTO/MTS classification against classifications based on previously utilized single-characteristic metrics. The performance of the suitability metrics will be evaluated by means of simulating the case company’s system performance under multiple levels of MTO and MTS production. Simulation is chosen as a means of analyzing the performance of these suitability metrics, as its modelling flexibility allows us to evaluate a large number of MTO/MTS configurations. Furthermore, using simulation will allow us to explore their performance in a real-life setting. Following Van Kampen, Akkerman & Van Donk (2012), the characteristics to be used will be selected based on the context of the MTO/MTS classification, and are therefore selected along with the case company and its specifics in the next chapter. However, the manner in which these characteristics will be combined into a single metric is part of the operationalization of the classification approach, and is described below.
In order to combine multiple characteristics into a single metric, we will use a simple, intuitive MTO/MTS suitability formula. Characteristics that are positively related to MTS suitability are put in the numerator of the formula, while characteristics that positively relate to MTO suitability are put in the denominator. In that manner, low scores indicate that a product is well suited for MTO production, while high scores indicate that a product can best be produced with an MTS strategy.
8 Having described how the products’ MTO/MTS suitability will be assessed, this section now turns to the question of MTO/MTS capacity allocation.
3.2.
Determination of cut-off value
After the products are plotted along a single dimension based on their MTO/MTS suitability, the question arises on where to draw the line to distinguish between MTO and MTS products. As described in the background, this cut-off value will be determined based on the capacity load the products constitute. This capacity load is defined as the products’ total demand divided by its processing speed. See Formula 1 below for the manner in which this capacity-based cut-off value is expressed.
𝐶𝑢𝑡 𝑜𝑓𝑓 𝑣𝑎𝑙𝑢𝑒 = ∑𝐷𝑖 𝑆𝑖∗𝑀𝑇𝑂𝑖 ∑𝐷𝑖 𝑆𝑖 (1)
Where D is a products total demand, S denotes a product’s processing speed and MTO is set to 1 if a product is produced with an MTO strategy, and 0 for an MTS strategy. The index i iterates over all products in the production system’s product portfolio.
In order to determine the optimal cut-off value, this paper will use the aforementioned simulation as a tool. This tool is applied as no analytical insights for the question of a system’s optimal levels of MTO of MTS production were developed in earlier work. Furthermore, simulation enables us to explore the system’s performance under different levels of MTO and MTS production.
In the next section, the case company is introduced. After describing the specifics of the company’s production setting, the manner in which the above mentioned simulation model is constructed will be described. Then, based on the case specifics, we define the experiments that will be conducted in pursuit of the aim of this paper.
4. Experimental setup
9
4.1.
Case description
The company selected to serve as a case is that of a company in the food processing industry. This industry was chosen as the industry-specific characteristics make that the MTO/MTS classification question is very relevant there (see, e.g. Van Donk, 2001). As described in the introduction, the company has a specific two-stage production system. In this case description, will show that this two-stage manufacturing system is an ideal setting to demonstrate the workings of the developed MTO/MTS classification approach. The company produces discrete food products, which can either be only packaged, or both packaged and assembled with other products to yield assortment-boxes. The company delivers these products to supermarket distribution centers and operates across national borders. Through conducting interviews and studying flowcharts, product data and demand data, the required information to perform this research was obtained. The interviews revealed that the company uses a hybrid production system in both the packaging and the assembly stage, and that the choice between MTO or MTS is an important consideration for the company. Hence, an MTO/MTS classification is made for both stages.
10 Figure 2: Case company's production setting
In the figure above, the production priorities per machine are denoted on the lines, which represent products exiting the stage. Hence, the first priority (P1) of the packaging stage is to process all components required by assembly. The second priority is to process all orders for MTO end-products (P2). Any remaining capacity will then be used to process products to replenish the stocks of MTS item (P3). The assembly stage has identical priorities, yet does not produce components and therefore only has two priorities: assembling orders for MTO assortment boxes (A1) and then assembling assortment boxes to replenish the stocks of MTS assortment boxes (A2). Both stage’s capacities for a day are utilized to meet demand in the sequence of MTO/MTS categories described. The within-category priorities are presented below in Table 1. Hence, MTO orders are processed from small to large to fulfill as many customer orders as possible, while MTS orders are processed starting with the item that has the largest difference from its base stock level.
Table 1: Production priorities of case company's production stages
Category Explanation Within-category priority
P1 Packaging for MTO assembly Follows assembly priority
P2 Packaging for MTS assembly Follows assembly priority
P3 Packaging for MTO EP Small Large
P4 Packaging for MTS EP Large Small
A1 Assembly for MTO EP Small Large
A2 Assembly for MTS EP Large Small
11 capacity requirements the assembly stage imposes on the packaging stage. Hence, only when the packaging stage has produced all components required by the assembly stage, it can utilize its capacity to produce end-products of its own. Recall that both stages use a hybrid production strategy, and that producing a high number of products with MTO increases capacity requirements. When the assembly stage allocates a large part of its capacity to MTO production, a higher capacity to produce its components is required from the packaging stage. Therefore, the packaging stage would require its capacity to be allocated to MTS production, in order to protect itself against increased capacity requirements imposed by the assembly stage. On the other hand, when the assembly stage processes a high amount of products with an MTS strategy, it requires less capacity from the packaging stage, and the packaging stage can afford to use more MTO production to decrease stock levels. Based on the above, we submit that determining the cut-off value between MTO and MTS production based on capacity requirements is high importance in this setting, as the packaging stage should theoretically shift its MTO/MTS distribution more towards MTS production, when the assembly stage increases its MTO production. The discussion above provides the reasoning for taking this specific company as a case, as described in the introduction of this subsection. We will continue by further introducing the case company’s production control policies.
MTS items are planned using an s,S-policy. Hence, stock levels for MTS products are inspected daily, and production orders to raise stock levels to the base-stock level (S) are issued when the current stock drops below the reorder-point (s). Base stock and reorder levels are calculated by the company using the formulas 2 and 3 below. These formulas are adopted from practice in order to analyze the system, a discussion on their optimality would fall outside of the scope of this research.
𝑅𝑒𝑜𝑟𝑑𝑒𝑟 𝑝𝑜𝑖𝑛𝑡 (𝑠) =1
2∗ [𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑜𝑣𝑒𝑟 𝑠ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒] (2)
𝐵𝑎𝑠𝑒 𝑠𝑡𝑜𝑐𝑘 (𝑆) = [3 𝑑𝑎𝑦𝑠 𝑜𝑓 𝑓𝑜𝑟𝑒𝑐𝑎𝑠𝑡] + 𝑠 − [𝑠𝑡𝑜𝑐𝑘 𝑎𝑡 ℎ𝑎𝑛𝑑] − [𝐹𝑜𝑟𝑒𝑐𝑎𝑠𝑡 𝑜𝑣𝑒𝑟 𝑙𝑒𝑎𝑑 𝑡𝑖𝑚𝑒] (3)
12 Table 2: Product characteristics of case company
Table 2 shows that, for every characteristic employed, the case company’s product portfolio is diverse in terms of MTO/MTS classification criteria. Combined with their employment of the two-stage assembly system without intermediate stocks, we submit this case is well-suited to study the effects of the two-stage MTO/MTS decision in.
4.2.
Simulation design
As described in Section 2, the suitability formula’s performance and the system’s performance under varying levels of MTO production will be analyzed by means of simulation. To that end, the case company’s production system was replicated in a simulation model. The main parameters of the simulation study are supplied by the case company and are summarized in Table 3.
Table 3: Parameters in simulation study
In Figure 3, a high-level flowchart of the simulation’s workings is presented. What follows is a detailed explanation of this model’s workings. In the simulation model, demand for the simulated day becomes known at the start of the day (t). Based on this demand, all MTS items are delivered from stock. When the
Area Characteristics Measure Range across
portfolio
Product High perishability Days 2 – 13 days
Market Varying demand uncertainty High variation in periodic demand Demand volume MAPE CoV Average weekly volume 0 – 1.89 0.11 – 2.04 0.27 – 1487.78 Parameters Values
Capacity 80% in all-MTS scenario
Setup times Period duration
New schedule generation time Demand levels
Forecasted demand levels Production speeds
12 minutes for packaging, 18 minutes for assembly 1 day
13 stock for a certain product is insufficient, demand is considered to be lost. Following the case company’s operations, components for assembly are packaged one day in advance. Therefore, the assembly planning is always made one day in advance (t+1), and the packaging planning is made and executed on t accordingly. Hence, the day starts with MTS products being delivered, after which the assembly planning of t-1 is executed. Then, based on forecasts for MTS items and today’s demand for MTO items, the assembly planning for t+1 is made. After this, the packaging planning is made based on today’s demand, stock levels for MTS items and the components requested by the assembly stage’s operations of t+1. In making a planning, capacity is allocated by first producing all components, then producing all MTO items and using remaining capacity for MTS items. For priority rules within these item categories, we refer to Table 2. At the end of the day, orders for MTO products are filled using the processed products of that day. Note how, as components need to be packaged one day in advance, the lead time for MTO assembly products is 1 day, while packaged MTO products are delivered at the end of the day on which they were ordered. As shelf life is a factor of interest, a product’s stock is modelled as a collection of batches, each with their own remaining shelf life. At the end of a day, all stocked batches are checked for deterioration and considered perished when their remaining shelf life has run out.
Certain assumptions and simplifications are made in order to capture the case company’s production system in a simulation model. The priority rules presented in Table 1 and the formulas used for the s,S-policy to control the stock levels of MTS items are rules-of-thumb employed by the company. In practice, situations may arise where the company decides to deviate from these rules. These instances are not considered in the simulation, and the described rules-of-thumb form the decision rules employed in the simulation model. Furthermore, the company’s packaging stage consists of multiple production lines. These were combined into a single line, in order to be able to research the effects of the levels of MTO and MTS production on a production system more precisely. Finally, as a simplification to the planning process, the aspect of the lead time a customer requires was ignored.
16
4.3.
Experimental design
Following from the operationalization of the classification approach, experiments will be conducted for two purposes. The first round of experiments will have the aim of developing and testing the performance of the aforementioned suitability formula. The second round of experiments will be conducted in order to explore the system’s performance under different levels of MTO and MTS production. These rounds of experiments will only consider the packaging stage of the case company. However, as described in subsection 4.1, it is expected that the assembly stage’s MTO/MTS distribution will have an impact on the packaging stage’s optimal MTO/MTS distribution. Recall that the interrelatedness between the packaging and assembly stages’ MTO/MTS distributions is expected to occur because the assembly stage’s MTO/MTS distribution interferes with the capacity the packaging stage has left to process the its own end-products. In order to demonstrate the workings of the proposed MTO/MTS classification approach, a third round of experiments will be conducted to show how the MTO/MTS can be altered in order to cope with a decrease in available capacity. This subsection will continue by operationalizing these experiments.
4.3.1. Single-metric MTO/MTS suitability
As described in Section 2, a single metric to assess the products’ MTO/MTS suitability will be developed by incrementally adding new characteristics to a suitability formula. If the newly added characteristic yields a better/worse system performance, the characteristic is maintained/excluded from the formula. The characteristics relevant to the MTO/MTS decision in the context of the case company were identified in Table 2. Therefore, these are to be used in the formulation of the classification formula.
17 Table 4: Design of Experiments: Suitability formula
Table 4 shows that the first characteristic to be added to the formula is the products’ shelf life. This characteristic was selected to be added first as initial experiments showed that the best performing single-characteristic classification was yielded when the suitability was assessed in terms of the products’ shelf life.
4.3.2. System performance under different levels of MTO/MTS production
After having developed a single-metric MTO/MTS suitability score, the products are ordered from low to high MTS suitability based on their score calculated by that metric. Having established this single dimension MTO/MTS suitability metric, we seek to explore the system’s performance under different levels of MTO and MTS production. Hence, in this series of experiments, we will study the system’s performance under different cut-off values as described in subsection 3.2. The design of experiments is given below in Table 5. Although these experiments resemble the experiments described above, they are described separately here, as they serve a different purpose and are therefore presented in a different subsection of the results.
Table 5: Design of Experiments: Optimal MTO/MTS cut-off
Suitability formula Levels of MTO production employed
(%) Output measures 𝑆𝑐𝑜𝑟𝑒 = 𝑆ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒 7, 15, 20, 25, 30, 35, 40, 50, 60, 70, 80 𝑆𝑐𝑜𝑟𝑒 = 𝑆ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒 𝑀𝐴𝑃𝐸 𝑆𝑐𝑜𝑟𝑒 =𝑆ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒 ∗ 𝑉𝑜𝑙𝑢𝑚𝑒 𝑀𝐴𝑃𝐸 𝑆𝑐𝑜𝑟𝑒 = 𝑆ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒 ∗ 𝑉𝑜𝑙𝑢𝑚𝑒 𝑀𝐴𝑃𝐸 ∗ 𝐶𝑜𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 7, 15, 20, 25, 30, 35, 40, 50, 60, 70, 80 7, 15, 20, 25, 30, 35, 40, 50, 60, 70, 80 7, 15, 20, 25, 30, 35, 40, 50, 60, 70, 80
Fill rate, deteriorated stock, average stock
Packaging stage’s % MTO production (independent variable)
Output measures (dependent variable)
18
4.3.3. Changing levels of MTO/MTS production to cope with a capacity change
After having conducted experiments to demonstrate the workings of the proposed MTO/MTS classification approach for single stage production, we will continue demonstrate how the MTO/MTS capacity load can be altered to cope with a decrease in available production capacity. Following that line of research, we will study the effects of a change in the assembly stage’s MTO/MTS capacity load on the capacity requirements it imposes on the packaging stage. Then, we demonstrate how the packaging stage’s MTO/MTS capacity load should shift as a result of the changing assembly stage’s MTO/MTS capacity load. To this end, we employ a full factorial design, where the MTO/MTS capacity loads of both stages are altered independently of the other.
Hence, we will first study the effect of changing the MTO/MTS capacity load in the assembly stage on the capacity it requires from the packaging stage. The design of these experiments is provided in Table 6 below.
Table 6: Design of Experiments: Effects assembly stage's MTO/MTS capacity load on capacity required from packaging
After having established the effects of the assembly stage’s MTO/MTS capacity load on the capacity requirements it imposes on the packaging stage, we will study how the packaging stage should shift its MTO/MTS capacity load in order to cope with the changing capacity devoted to the assembly stage. In order to do this, we will employ a full factorial experimental design, where the two independent variables are the MTO/MTS capacity load in the packaging and assembly stages. The design of these experiments is shown in Table 7.
Assembly stage’s % MTO production (Independent variable)
Output measures (dependent variable)
19 Table 7: Design of Experiments: Effects of assembly's MTO/MTS capacity load on optimal packaging’s MTO/MTS capacity load
4.3.4. Operationalization of output measures
In table 8, the KPI’s of interest to this research are briefly enumerated. Furthermore, reasoning for selecting these is provided. The stock and deterioration metrics will be expressed as percentages, compared against the all-MTS scenario which is subsequently expressed as 100%.
Table 8: KPI's of interest in this research
Packaging stage’s % MTO production Assembly stage’s % MTO production Output measures
0 7 15 20 25 30 35 41 50 60 70 80 100 0 7 15 22 28 36 40 44 48 59 73 87 100
Fill rate of packaging stage
Output measures Reasoning for selection
Fill rate Most important KPI to the case company
Stock levels
Product deterioration
Percentage of packaging stage’s capacity required for producing components for assembly
Indication of holding cost of goods
Significant expense and increases capacity requirements due to rework – expected to fall with more MTO production
20
5. Results
In this section, the results of this work are provided. The results are presented following the sequence with which the experiments are defined in subsection 4.3. After the results of each experiment are provided, a short interpretation of the key observation from the results are provided in the text. A more in-depth discussion of the findings will be the subject of Section 6.
5.1.
Single-metric MTO/MTS suitability
Here, the formulation of a single-metric MTO/MTS suitability assessment is developed, and its conceptual approach is validated. By incrementally adding new characteristics to the formula, we confirm both the added value of incorporating each characteristic in the formula, as well as the conceptual approach of using a multiplication formula to combine multiple characteristics into a single metric. The latter will be determined by observing if the multiplication formula yields a better performing MTO/MTS classification then a single-characteristic assessment method, which was used in earlier academic work.
The results of this procedure are displayed below in Figure 4. The x-axis shows the cut-off values employed, while the y-axis displays the system’s results in terms of fill rate. The four plotted lines represent the four different classification formulas used, as described in subsection 4.3.1.
Figure 4: Performance of MTO/MTS suitability formulas in terms of fill rate
We observe from Figure 4 that the suitability formula yields the best fill rate when shelf life, forecast accuracy and demand volume are incorporated. We observe that demand variance should be excluded. Before drawing conclusions on the best performing suitability metric, we first present the performance of the suitability formulas in terms of resulting stock levels and the level of deteriorated stocks. These results are provided in Figure 5. Here, the x-axis denotes changing levels of MTO production, while the y-axis in the two graphs show the stock levels and levels of deterioration respectively.
97,20% 97,40% 97,60% 97,80% 98,00% 98,20% 98,40% 0 10 20 30 40 50 60 70 80 Fi ll rate p ac kagi n g
% MTO production in packaging stage
Shelf life Shelf life + MAPE Shelf life + MAPE + Volume
21 Figure 5: performance of MTO/MTS suitability formulas in terms of stock levels and deteriorated goods
From figure 5 it can be concluded that making a product classification based on shelf life only, yields both higher stock levels and higher levels of product deterioration. The other suitability formulas yield similar results in terms of stock levels and product deterioration.
As Figure 4 shows, the best fill rate is achieved with a combined classification metric of shelf life, forecast accuracy and demand volume. Figure 5 indicates that, when inspecting the results in terms of stock levels and product deterioration, this formulation of the classification metric performs equally well as the other formulations of the metric. Both figures indicate that the combined metric of shelf life, forecast accuracy and demand volume performs better than the single-characteristic metric of shelf life. Hence, we have developed a multiple-characteristic MTO/MTS suitability metric and compared its performance against previously used single-characteristic metrics. As we observe the multiple-characteristic metric yields a better performance than the single-characteristic metrics, we will continue our experiments using that metric. The finding that this multiple-characteristic formula yields better results when demand variance is excluded will be discussed in the next section. The best performing MTO/MTS suitability formula yielded is displayed in Formula 4.
𝑀𝑇𝑂/𝑀𝑇𝑆 𝑠𝑢𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑚𝑒𝑡𝑟𝑖𝑐 =[𝑉𝑜𝑙𝑢𝑚𝑒] ∗ [𝑆ℎ𝑒𝑙𝑓 𝑙𝑖𝑓𝑒]
[𝑀𝐴𝑃𝐸] (4)
The multiplication formula above produces MTO/MTS suitability scores per product. Now, the products can be ordered from low to high MTS suitability. Having done this, the subsequent allocation of products
0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% 0 20 40 60 80 100 Sto ck le ve ls
% MTO production in packaging stage
Shelf life Shelf life + MAPE
Shelf life + MAPE + Volume Shelf life + MAPE + Volume + Variance
0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% 0 20 40 60 80 100 D e te ri o rate d g o o d s
22 to MTO or MTS production will be based on the capacity loads the products constitute. In Figure 6, the products are ordered from low to high MTS suitability (x-axis), while the y-axis represents their individual capacity load. Recall from Section 2 that this capacity load is determined by dividing a product’s total demand by the product’s processing speed.
Figure 6: Resulting MTO/MTS suitability and capacity loads
The products classification above will be used as follows. If we choose to devote 30% of capacity to MTO production, we will assign products to MTO production from left to right in Figure 6, until the capacity loads of products assigned to MTO production constitute 30% of the total product portfolio’s capacity load. Following this approach, we will now continue by studying the system’s performance under multiple MTO/MTS capacity loads.
5.2.
System performance under different levels of MTO/MTS production
In this subsection, the packaging stage’s performance under different levels of MTO/MTS production will be explored. In the previous subsection, the products were ordered from low to high MTS suitability. In this subsection, we will present the results of simulating the packaging stage under multiple levels of MTO production, in order to explore the system performance under different levels of MTO/MTS production, and to determine the optimal cut-off value between these.
The first results that are presented, show the effects of different levels of MTO/MTS production on a production system’s fill rate. In Figure 7, the x-axis represents the level of MTO production in the case company’s packaging stage, while the y-axis represents the resulting fill rate.
1 2 4 8 16 32 64 128 Cap ac ity lo ad (Lo gar ith m ic scale)
23 Figure 7: Effect of level of MTO/MTS production on fill rate
As can be concluded from Figure 6 above, the fill rate is to a large extend impacted by the level of MTO/MTS production in a system. Continuing to study the effects of the levels of MTO/ MTS production, we present its effects on the stock levels carried by the production system, and the level of deterioration. Figure 8 below shows the effects of changing levels of MTO production on the stock carried by the production system and the deterioration levels. The x-axis represents the change in the percentage of MTO production, while the y-axes show the levels of stock carried by the system and product deterioration respectively
Figure 8: Effect of level of MTO production on the system’s stock levels and deterioration levels
Figure 8 shows that the level of stock carried by the production system decreases linearly with the percentage of MTO production, while the level of product deterioration decreases in a convex manner. Hence, a higher level of MTO production yields lower stock levels and less deterioration.
84,00% 86,00% 88,00% 90,00% 92,00% 94,00% 96,00% 98,00% 100,00% 0 10 20 30 40 50 60 70 80 90 100 Fi ll r ate p ac kagi n g
% MTO production in packaging stage
0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% 0 20 40 60 80 100 Sto ck le ve ls
% MTO production in packaging stage
0,00% 10,00% 20,00% 30,00% 40,00% 50,00% 60,00% 70,00% 80,00% 90,00% 100,00% 0 20 40 60 80 100 D e te ri o rate d g o o d s
24 In this series of experiments, the system performance as a result of different levels of levels of MTO/MTS production was studied. It was found that a system’s fill rate, stock levels and product deterioration levels are very dependent upon how the cut-off between MTO and MTS products is made. The performance in terms of fill rate is optimal between 20% and 50% of MTO production, after which it starts to decrease in a convex manner. Figure 7 shows that, in order to minimize stocks, a high level of MTO production is preferred, and that the decrease in the levels of stock deterioration is subject to the law of diminishing returns. Taking these three output metrics into account, we find that the optimal level of MTO production for this production system is around 50%, as at that point nearly the highest fill rate is achieved, while minimizing stock and deterioration levels.
5.3.
Changing levels of MTO/MTS production to cope with a capacity change
After having demonstrated a manufacturing system’s sensitivity to different levels of MTO production in a hybrid system, we now show how changing the level of MTO production can be wielded to cope with a decrease in available capacity. As described in subsection 5.3, we will first present how a change in the level of MTO production in the case company’s assembly stage interferes with the capacity the packaging stage has left to produce its own end-products. Then, we will present how the packaging stage’s percentage of MTO production should change in order to maintain an optimal system performance. Hence, the first results we present show how the packaging stage’s capacity required to produce components for the assembly stage changes as a result of changing levels of MTO production in the assembly stage. To that end, the x-axis in Figure 9 denotes the percentage of MTO production in the assembly stage, while the y-axis represents the portion of the packaging stage’s capacity that is used to produce components for the assembly stage, denoted as ‘% pack capacity for assembly’.
Figure 9: Effect of level of MTO production in assembly stage on required capacity from packaging
31,00% 31,50% 32,00% 32,50% 33,00% 33,50% 34,00% 34,50% 35,00% 35,50% 36,00% 0 10 20 30 40 50 60 70 80 90 100 % p ac k cap ac ity for asse m b ly
25 From Figure 9 we observe that, as expected, the capacity the packaging stage requires to produce components for the assembly stage changes as a result of changing levels of MTO production in the assembly stage. The observation that the capacity requirements initially drop will be discussed in the next section. Nonetheless, we observe the capacity the packaging stage has to produce end-products of its own is a function of the assembly stage’s MTO/MTS distribution.
Now, we will demonstrate how the packaging stage’s level of MTO production can be altered to cope with a change in available production capacity. To this end we employ a full factorial experimental design, as described in subsection 4.3.3. Recall from that section that the two independent variables are the percentages of MTO production in both stages, and that the dependent variable will be the fill rate of the packaging stage. The results of the full factorial experiment are displayed in Figure 10. Here, the rows represent the percentages of MTO production in the packaging stage, while the columns show the levels of MTO production in the assembly stage. The values in the cells represent the fill rate of the packaging stage, expressed as percentages. The colors of the cells range from green to red, to indicate high and low values respectively.
Figure 10: Joint effects of packaging and assembly stage’s levels of MTO production on the packaging stage’s fill rate
26 Having provided the results of the experiments conducted, these will be discussed in the next section of this work.
6. Discussion
In this section, the results provided in the previous section are interpreted and discussed. First, the results of the development of the MTO/MTS suitability metric are discussed, after which the findings on a system’s performance under different levels of MTO/MTS production are discussed. Finally, the manner in which the capacity load between MTO and MTS products was used to respond to a change in available capacity is discussed.
6.1.
Single-metric MTO/MTS suitability assessment
In this work, a novel, simple multiplication formula was used to assess the products’ MTO/MTS suitability. Figure 4 shows that the difference between the single-characteristic technique and the multiplication formula developed in this work constitutes a decrease in missed orders of around 15%. This significantly better system performance is particularly interesting in the light of the research by Van Kampen, Akkerman and Van Donk (2012). Here, the authors find that classification techniques increase in complexity as more quantitative characteristics are incorporated in the MTO/MTS decision. However, the results of this work show that a fairly easy assessment formula might perform well in assessing the MTO/MTS suitability of products by combining multiple qualitative classification characteristics into a single metric.
Although the determination of which characteristics should and should not be used in assessing a product’s MTO/MTS suitability is not the main aim of the present paper, an interesting observation was made. It was found that the best classification was yielded when demand variance was excluded from the MTO/MTS suitability formula. Demand variance is a commonly used MTO/MTS assessment criterion (Syntetos, Boylan & Croston, 2005; Bacchetti, Plebani & Saccani, 2013), but our results are in accordance with the findings of Van Kampen and Van Donk (2014), who found that demand variance should not be taken into account, as long as it can be accurately forecasted.
6.2.
System performance under different levels of MTO/MTS production
27 deterioration, as the managerial choice can be made to accept a minor decrease in fill rate in order to limit the stock levels and production deterioration. Although these findings follow from theoretical understanding of hybrid systems (see, e.g., Soman, Van Donk & Gaalman, 2004), to the author’s knowledge, this paper is the first that explores the effects of different levels of MTO and MTS production by means of simulating a real-life setting. By demonstrating their sensitivity to the level of MTO production employed, we stress the importance of taking the resulting level of MTO and MTS production into account when making an MTO/MTS product classification, as earlier identified by Soman, Van Donk & Gaalman (2004).
Finally, the observation that deterioration decreases in a convex manner indicates that the biggest deterioration gains are made when a small portion of the products is assigned to MTO production. As more products are produced with an MTO strategy, the gains in terms of lower deterioration decrease. This could be explained by the manner in which the products are ordered and assigned to MTO production, as the products that are the first to be selected for MTO production, have the worst forecast accuracy and shortest shelf life. Hence, assigning products from MTS to MTO seems to be subject to the law of diminishing returns, where the biggest gains in terms of product deterioration are made in assigning the first products to MTO.
6.3.
Changing levels of MTO/MTS to cope with a capacity change
In determining how the assembly stage’s MTO/MTS distribution affects the capacity it requires from the packaging stage, we observe an initial drop in required capacity. This observation is was not expected, as it is not in line with the scholarly consensus that more MTO production requires a higher capacity (Rajagopalan, 2002). This initial drop may be a result of the simplifications made in the simulation model, specific parameter settings or may be the product of the company’s specific two-stage production system. In any case, this observation should be met with suspicion. However, as the assembly stage’s MTO/MTS distribution was found to have an effect on the capacity available to the packaging stage, we continued to demonstrate how the packaging stage’s MTO/MTS distribution should change as a result.
28 2002). Finally, this finding implies that that a manufacturing company could dynamically alter the MTO/MTS classification of its product portfolio in response to changing conditions, such as increasing or decreasing market demand or supply chain disruptions.
6.4.
The application of the proposed MTO/MTS classification approach
The MTO/MTS classification approach employed proves to have several advantages. Although making an MTO/MTS product classification requires effort, it should be revised over time (Van Kampen & Van Donk, 2014). Using the the proposed MTO/MTS classification approach, we submit that it is fairly easy to revise the product classification, once the MTO/MTS suitability metric is defined. Hence, after carefully determining the characteristics relevant to the MTO/MTS classification in a specific context, the process of assigning products to either MTO or MTS production was found to be fairly easy and straightforward when applying the proposed approach. This is especially so when comparing this approach to other commonly employed approaches, where cut-off values have to be determined for every characteristic individually. Furthermore, the exploration of the system under multiple levels of MTO/MTS production shows that explicitly considering the amount of MTO or MTS production yielded by a product classification has a large effect on the system’s performance.
7. Conclusion
The aim of this work was to develop a novel MTO/MTS product classification approach that uses the preferred levels of MTO and MTS production as a cutoff value. Furthermore, the effects of different capacity loads between MTO and MTS production on a system’s performance were explored. Finally, a multiplication formula to combine multiple characteristics into a single-metric MTO/MTS suitability score was developed and tested against previously used single-characteristic metrics.
29
7.1.
Implications and further research
The findings have several managerial and theoretical implications. Firstly, we submit that more empirical work is needed to establish the performance of the multiplication formula as a classification technique. To that end, an interesting direction for further work would be the comparison of multiple product classification techniques. Hence, how would a simple technique, as the multiplication formula, perform against other, more complex classification techniques? We would expect a trade-off between complexity and resulting performance, and it would be interesting to plot these two against one-another. In doing so, substantial managerial insights could be developed, offering practitioners guidelines in choosing a classification technique. With our finding that demand variance should not be used in determining a product’s MTO/MTS suitability, we agree with Van Kampen and Van Donk (2014) but disagree with Olhager (2003) and Wanke and Zinn (2004). This implicates that more empirical work could be done to either in- or exclude demand variance from the classification characteristics, or to identify contextual circumstances which dictate the use of certain characteristics over others. To this end, we expect simulation studies or analytical approaches to be an effective tool in determining when certain characteristics should be in- or excluded from a product classification procedure. Finally, although we demonstrated the system’s performance under multiple cut-off values, no guidelines were developed on how to analytically determine this value. Therefore, it would be interesting to see if analytical approaches can be developed to determine a system’s optimal levels of MTO and MTS production.
7.2.
Limitations
30
References
Baccheti, A., Plebani, F., Saccani N., & Syntetos A. A. (2013) “Empirically-driven hierarchical classification of stock keeping units” International Journal of Production Economics, Vol. 143, pp. 263-274. Beemsterboer, B., Land M., & Teunter R. (2017) “Flexible lot sizing in hybrid make-to-order/make-to-stock production planning” European Journal of Operational Reseach, Vol. 260, pp. 1014-1023.
Cavalieri, S., Garetti, M., Macchi, M., & Pinto R. (2008) “A decision-making framework for managing maintenance spare parts” Production Planning & Control, Vol. 19(4), pp. 379-396.
Claassen, G. D. H., Gerdessen, J. C., Hendrix E. M., & Van der Vorst, J. G. (2016) “On production planning and scheduling in food processing industry: Modelling non-triangular setups and product decay” Computers & Operations Research, Vol. 76, pp. 147-154.
D'Alessandro, A. J., & Baveja, A. (2000) “Divide and Conquer: Rohm and Haas' Response to a Changing Specialty Chemicals Market” Interfaces, Vol. 30(6), pp. 1-16.
Van Donk, D. P. (2001) “Make to stock or make to order: The decoupling point in the food processing industries” International Journal of Production Economics, Vol. 69, pp. 297-306.
Huiskonen, J., Niemi, P., & Pirttilä, T. (2003) “An approach to link customer characteristics to inventory decision making” International Journal of Production Economics, Vol. 81-82, pp. 255-264
Van Kampen, T. J., Akkerman, R., & Van Donk D. P. (2012) "SKU classification: a literature review and conceptual framework" International Journal of Operations & Production Management, Vol. 32(7), pp. 850-876
Van Kampen, T. J., & Van Donk, D. P. (2014) “When is it time to revise your SKU classification: setting and resetting the decoupling point in a dairy company” Production Planning & Control, Vol. 25(16), pp. 1338-1350.
31 Olhager, J. (2003) “Strategic positioning of the order penetration point” International Journal of Production Economics, Vol. 85, pp. 319-329.
Soman, C. A., Van Donk, D. P., & Gaalman, G. J. C. (2004) “Combined to-order and make-to-stock in a food production system” International Journal of Production Economics, Vol. 90, pp. 223-235.
Syntetos, A. A., Boylan J. E., & Croston, J. D. (2005) “On the Categorization of Demand Patterns” The Journal of the Operational Research Society, Vol. 56(5), pp. 495-503.
Rajagopalan, S. (2002) “Make To Order or Make To Stock: Model and Application” Management Science, Vol. 48(2), pp. 241-256.
