Mitigating nesting performance implications when applying workload control in high variety MTO companies – An explorative case study

(1)

Mitigating nesting performance implications when applying

workload control in high variety MTO companies

– An explorative case study

MSc Technology & Operations Management

Kevin Schreur

Supervisor: Dr. M. J. Land Second assessor: Dr. J. A. C. Bokhorst

Company supervisor: Pim Jansen Abstract:

Workload control (WLC) is considered one of the most suitable production planning and control concepts for high variety make-to-order (MTO) companies. Through an order-pool that acts as buffer against external dynamics, high variety MTO companies are able to control work in progress, improve flow and decrease throughput time. Within the concept, a trade-off exists between balancing shop floor workload and processing the most urgent orders. In practice, several complexities exist e.g. sequence-dependent setup times, batching and nesting, that complicate this traditional trade-off. Additional considerations in processing the most urgent orders and balancing workload have to be made to handle these complexities. When not managed effectively, these complexities can result in performance implications as loss of capacity, increase in material waste, reduced delivery reliability or increased throughput time. In literature we find not much attention has been granted to nesting in relation to WLC theory, despite that this complexity can be found in many industries. This paper therefore aims to generate more knowledge about nesting and how performance implications that result from nesting when applying WLC can be mitigated. By means of an explorative case study at an advanced high variety MTO company in the Netherlands, we derived knowledge that supported us in proposing several production planning and control decisions (PPC) that could be applied to mitigate nesting implications when WLC principles are applied. The proposed PPC decisions in this study have been drawn per hierarchical control level. With this study we aim to contribute to the limited empirical literature available concerning complexities that complicate application of WLC principles.

Key words: Workload control; nesting; high variety make-to-order industry; case study; job shop June 2019

Faculty of Economics and Business Nettelbosje 2

9747 AE Groningen k.schreur.1@student.rug.nl

(2)

2

1. Introduction

High variety make-to-order (MTO) companies grow in importance due to increasing demand for more customised products (Stevenson, Hendry and Kingsman, 2005; Hendry et al., 2008). MTO is a widely applied strategy allowing companies to deliver a greater variety in offerings with higher degree of customisation. This complex environment requires a production planning and control (PPC) system that corresponds with the production system characteristics. For high variety MTO companies, workload control (WLC) is considered a very suitable PPC concept (Hendry, Kingsman and Cheung, 1998; Henrich, Land and Gaalman, 2004; Thürer, Stevenson, et al., 2012).

WLC allows companies to deal with a high degree of complexity by controlling queues on the shop floor via three hierarchical control levels that act as buffer against variability of incoming orders (Land and Gaalman, 1996; Hendry, Kingsman and Cheung, 1998; Kingsman and Hendry, 2002). In this way, workload and work in progress (WIP) can be controlled as well, which improves flow and decreases throughput time (Stevenson, Hendry and Kingsman, 2005; Land and Gaalman, 2009; Thürer, Silva and Stevenson, 2011; Thürer, Stevenson, et al., 2012). The three hierarchical input levels are, order entry, order release and priority dispatching (Tatsiopoulos and Kingsman, 1983; Land and Gaalman, 1996).

At the order entry level, decisions regarding accepting/rejecting and delivery date promising of orders are made based on e.g. material availability and available capacity. Accepted orders with assigned delivery dates first enter a pre-shop pool awaiting their release (Hendry, Huang and Stevenson, 2013). At the order release level, decisions are made concerning timely release of orders to the shop floor to meet promised delivery dates while keeping workload within acceptable boundaries to sustain output (Thürer et al., 2012; Fredendall, Ojha and Patterson, 2010). Often processing orders and balancing workload on the shop floor are in conflict, resulting in a trade-off. Urgent orders are typically considered first for release, however, when workload of an order does not fit within boundaries of all work centres, its release will be postponed. To balance workload on the shop floor load gaps are filled with less urgent orders, which avoids premature idleness and waiting time of orders (Land and Gaalman, 1998). At the priority dispatching level, workload is balanced and moved between workstations on a daily basis by dispatching rules that may change priorities over time (Kingsman, 2000; Land, 2004).

In practice, several complexities exist, e.g. nesting, sequence-dependent setups or assembly (Henrich et al., 2004), which hinder effective application of WLC and require additional attention to be managed effectively (Cransberg, Land, Hicks and Stevenson, 2016). Complexities have implications for WLC as they disrupt flow on the shop floor and complicate order release and priority dispatching decision. Cransberg et al. (2016) therefore proposed a framework to determine on which hierarchical level complexities are most suitable to be handled, however, it only provides limited insight in how complexities can be treated in practice. Despite several solution directions are proposed, these are rather limited. To fully benefit from advantages WLC has to offer, high variety MTO companies should be able to make the correct decisions on the appropriate hierarchical control level. Given that only a limited number of practical solutions for handling complexities are proposed in WLC literature, and implementations of solutions to mitigate implications are limited, a gap between theory and practice remains.

(4)

4 Nesting complicates the traditional trade-off in WLC between timing and load balancing, due to additional considerations that have to be made to prevent material waste (i.e. nesting efficiency). (Cransberg et al., 2016). For an efficient nesting orders that could be combined based on material type are often pulled forward, which conflicts with processing the most urgent orders. This can result in excessive release of less urgent orders to the shop floor and prevent more urgent orders from being released if workload norms are not to be exceeded. On the other hand, taking into account current workload levels and norms of the work centres in the nesting process can reduce the nesting efficiency due to restricted possibilities in combining orders.

The results of this study contribute to WLC literature by extending knowledge concerning mitigating performance implications for MTO companies that apply WLC principles. To fulfil the purpose of this paper, a case study has been executed at an MTO company in the metalworking industry that makes extensive use of nesting for its laser cutting operations. This company applies a wide variety of tools to control their operations and has useful data available in their Enterprise Resource Planning (ERP) system. The case study comprehends an in-depth analysis of the company’s current performance and planning structure. This allows us to expose the implications of nesting and learn how this complexity is currently being handled. A set of PPC decisions to mitigate nesting implications when applying WLC principles has been proposed.

(5)

5

2. Theoretical background

In this section, relevant WLC literature for this study has been reviewed. It starts with a brief explanation of the WLC concept followed by describing difficulties to improve performance in high variety MTO companies and why complexities are difficult to handle. Next, implications are described that result from complexities in real life high variety MTO companies that complicate application of WLC theory. Moreover, it describes how existing literature deals with complexities and the resulting implications. In the last section, emphasis has been put on challenges for handling the nesting complexity, which performance implications we can expect and where is room for further improvement of knowledge.

2.1 The WLC concept

Over the past decades WLC has been widely researched and has become a mature PPC concept. WLC is a concept based on input/output control (I/OC) that allows MTO job shops that manufacture high variety low volume products, to meet high customer demands and promised delivery dates (Land and Gaalman, 2009; Stevenson et al., 2005; Thürer et al., 2011, Thürer et al., 2012a). Due to the complexity of high variety MTO companies, it is not appropriate to implement other PPC concepts such as just in time (JIT) or theory of constraints (TOC), because MTO companies do not have the benefits of repetitive manufacturing (Henrich, Land and Gaalman, 2004). This does not allow MTO companies to configure their shop floor in a simplified layout (Hendry, Kingsman and Cheung, 1998; Kingsman and Hendry, 2002). Being able to control shop floor queues and, therefore control of lead time, is very important for high variety MTO companies as delivery dates often have to be quoted to customers in pre-production stage and can be crucial in winning orders (Hendry, Kingsman and Cheung, 1998; Stevenson, Hendry and Kingsman, 2005).

Controlling shop floor queues is achieved by integrating production and sales in a hierarchical system that acts as a buffer against external variability (Tatsiopoulos and Kingsman, 1983). The hierarchical system consists of three control levels, order entry (i.e. control of total workload), order release (i.e. control of planned workload) and priority dispatching (i.e. control of WIP) (Thürer et al., 2012a; Fernandes et al., 2016), which are depicted in Figure 1. The order entry level is where decisions are made whether to accept or reject customer orders and concerns their medium-term production planning, including due date allocation to orders. Once accepted, orders move to the release level awaiting release to the shop floor in a so-called

pre-shop order/job pool (Henrich et al., 2004). The order/job pool acts as buffer, protecting the shop floor from the variability in workload caused by incoming customer orders (Land and Gaalman, 1996). The order release level comprehends short-term production planning decisions and controls WIP by releasing orders to the shop floor, without compromising promised delivery dates and exceeding workload norms. The order release therefore decouples the order pool from the shop floor (Thürer, et al., 2012a). The release of urgent

(6)

6 floor are often in conflict, resulting in a trade-off between timing and balancing. The timing function relates to the sequence in which jobs are placed in the order/job pool in order to meet promised delivery dates, which indicates its relative urgency. The load balancing function entails the selection of jobs in the sequence based on their impact on workload. When the job exceeds the set workload limits, the job is not be released and its release will be re-evaluated at the next periodic release (Land, 2004). Order release methods should be applied to effectively deal with the trade-off between timing and balancing workload to create a smooth flow of jobs to the shop floor (Henrich, Land and Gaalman, 2004; Land, 2004). At the priority dispatching level, day-to-day shop floor workload is controlled by dispatching rules that may change priorities over time (Hendry, Kingsman and Cheung, 1998; Kingsman, 2000; Henrich, Land and Gaalman, 2004).

We now know the principles of WLC and how it enables control over shop floor queues to decrease throughput time. However, companies still struggle with implementing WLC and improving performance when complexities are present as these complicate the traditional timing/balancing trade-off.

2.2 Performance in high variety MTO companies

MTO companies have to deal with increasingly high customer demands. Consequently, it is important for MTO companies to remain competitive without compromising on performance. Key performance indicators for MTO companies are lead times and delivery reliability (Hendry, Kingsman and Cheung, 1998; Stevenson, Hendry and Kingsman, 2005). Short lead times are particularly important to MTO companies as they are unable to forecast demand or produce upfront as make-to-stock (MTS) companies are able to do (Stevenson, Hendry and Kingsman, 2005). Therefore, lead times are considered crucial for order winning. Delivery reliability refers to the punctuality of actual order delivery compared to promised delivery dates. Delivery reliability performance can be indicated by the average lateness of orders (Soepenberg, Land and Gaalman, 2012a). According Soepenberg et al. (2012c), delivery reliability can be improved by controlling both the average lateness and variance of lateness.

(7)

7

2.3 Handling complexities for WLC

Limited empirical knowledge is available concerning implementation of WLC and how to cope with complexities that exist in real life MTO companies that complicate application of the WLC concept. In WLC literature, many studies are based upon simulation wherein performance of order release methods is being assessed (Fredendall, Ojha and Wayne Patterson, 2010; Thürer, Stevenson and Silva, 2011). Many of these studies do not take into account the existing complexities of real-life high variety MTO companies, hence, the gap between theory and practice still remains. Over the years, however, there have been several studies that addressed complexities in MTO companies that hinder application of WLC. Henrich et al. (2004) identified in their study on WLC applicability that assembly and sequence dependent setup times may cause issues for WLC implementation. They argued that before WLC could be effectively implemented these obstacles should be removed, or at least its impact be minimised. For WLC to be most effective, setup times should be kept to a minimum and assembly should not have a dominant structure (Henrich, Land and Gaalman, 2004). The argumentation of Henrich et al. (2004), is from a different perspective as WLC becomes less appropriate when complexities are present, therefore the impact of the complexity itself should be minimised. However, this is contradictory with what we want to accomplish with this study, which is to improve applicability of WLC when complexities are present, and mitigate performance implications when applying WLC, i.e. find out how could WLC be made more applicable when complexities are present.

Only recent studies have investigated the issue of sequence dependent setup times in more detail and address that these can have significant impact on performance (Fernandes and Carmo-Silva, 2011). Fernandes et al. (2011) concluded that handling sequence dependent setups on a local dispatching level does not provide optimal results. The study of Thürer, Silva, Stevenson and Land (2012b), has indicated sequence dependent setups are best to be tackled on dispatching level by means of a setup oriented dispatching rule complemented by one that is not entirely focused on dispatching. On the order release level, they suggest controlled order release can be combined with dispatching, although, workload norms should allow for some lenience. In a later study, Thürer, Silva, Stevenson and Land (2014) conclude, that the LUMS COR order release method delivers the best performance within the WLC design concept, even with sequence dependent setups.

Cransberg et al. (2016) have designed a framework to determine at which hierarchical control level complexities should be handled. In their research, several complexities have been identified from literature that require additional attention to be controlled effectively. Their framework is depicted in Figure 2. The Y-axis refers to the position of the complexity in the routing, whereas the X-axis refers to the criticality of the complexity. The position of the complexity in routing determines on what level the complexity should be addressed. Supported by Soepenberg et al. (2012) and Henrich et al. (2004), the more downstream the complexity is positioned in the routing, the less impact a decision will have when executed at the order release level, therefore implying the complexity should be managed by priority dispatching decisions. The criticality dimension is separated into two sub-dimensions (i) criticality of the bottleneck and (ii) criticality of release sequence. It refers to the impact of a complexity on output of an operation. A complexity at the bottleneck machine is considered more Figure 2: The conceptual decision framework to determine

(8)

8 critical compared to non-bottleneck machines. The criticality of the sequence relates to the impact the release sequence of jobs/orders would have on the performance of an operation. Criticality of sequence is considered high if the impact of the release sequence has a high impact on the output of the operation. This is the case for e.g. sequence dependent setup times or batching, as their output can reduce and cause long waiting times if release sequence of orders is not taken into account properly. If release sequence of orders only has minor impact on performance and the machine is not a bottleneck station, the sequence criticality is considered low (Cransberg et al., 2016). For nesting, the criticality of sequence refers to the impact of release sequence of orders on material waste. Planning advantages made by release sequence should not be at the expense of unlimited material losses, which would increase cost significantly. The criticality of sequence should be considered high if it significantly impacts nesting efficiency and low otherwise.

By determining both routing position and criticality, complexities can be appointed to one of the four quadrants which provide direction for handling complexities at either the job release level or priority dispatching level. The framework proposed by Cransberg et al. (2016), already narrows the gap between theory and practice, but only provides limited answers on how to mitigate implications resulting from complexities. Neither does the proposed framework include the order acceptance level as option for handling complexities, which could be a possibility to handle complexities. The authors have provided a good starting position to determine at which hierarchical level complexities should be handled and provide several solutions to mitigate implications. However, they do not further elaborate on them. Therefore, it still remains unclear how to mitigate implications that result from complexities even though it may be known at which level the complexities should be handled.

2.4 The nesting complexity and implications

Nesting is a process that aims to reduce waste of material and loss of machine capacity by maximising the number of parts, or finding an optimal combination of parts, that can be cut from or printed on a (irregular shaped) sheet or raw material (Heistermann and Lengauer, 1995). The nesting is encountered in several manufacturing industries, such as textile, metal, glass, leather and paper industry, and is known as a NP-hard problem (Chryssolouris, Papakostas and Mourtzis, 2000).

The nesting process itself is performed by bundling orders that require parts made from the same material often referred to as a job. The orders that form the job are typically selected based on their combination of parts that form the largest possible surface that can be cut from or printed on a sheet of material. However, by only looking at forming jobs based on optimal combination of parts in the orders to prevent waste, is a somewhat simplistic approach. In many industrial settings both the nesting problem and the scheduling problem have to be taken into account (Chryssolouris, et al., 2000). Nesting can be considered a complexity as the objective of nesting and scheduling can be contradictive. Whereas nesting focusses on reducing waste, scheduling aims to optimize sequences to improve efficiency and agility (Sakaguchi, Matsumoto and Uchiyama, 2018). In an ideal situation, the effect on overall production objectives, e.g. lateness, tardiness, production cost and capacity utilisation, should be taken into account (Chryssolouris, et al., 2000).

(9)

9 as set. When spread in urgency between orders that are nested becomes greater, flow on the shop floor is likely to be disrupted and delivery reliability performance is likely to decrease.

Whereas performing nesting before order release allows orders in the order pool to be nested, performing nesting after release only considers orders on the shop floor. Nesting before release is therefore expected to be more efficient in terms of material waste but causes more difficulties in the timing/balancing trade-off. Performing nesting locally at the operation, is likely to affect the nesting efficiency as chances of combining orders that can be cut from the same material is lower. This is especially the case for materials that are uncommon. Consequently, even less orders will be available for nesting. It is likely this will result in an increase of material losses and therefore costs. The effect of nesting on delivery reliability and throughput time, when it is performed locally, is likely to be lower. Especially when the traditional trade-off did not include nesting considerations. Therefore, is expected spread between urgency of orders becomes smaller and disruptions in flow decrease, but material loss (i.e. cost) will increase. Nesting locally is anticipated to be less efficient in terms of material loss, but it does not interfere with the trade-off between timing and balancing.

The release decisions when nesting locally can have a significant impact. If a high nesting efficiency is desired, the waiting time for orders to be nested will likely increase. If waiting times must be short, it is on the other hand unlikely a high nesting efficiency can be realised. A high nesting efficiency is more likely to be realised when more orders can be weighed against each other. On a local level this can be realised by postponing the nesting process and wait for released orders to arrive for creating an efficient nesting. However, this will increases waiting time. The release sequence of orders can therefore play an important role, because waiting times can be reduced, and nesting efficiency can be increased if orders arrive in sequence based on material type. However, the release sequence could also have impact on output of other operations. Depending on whether the operation affected most by nesting is a bottleneck or not, one may decide whether to take nesting considerations into account in the release sequence of orders. This is much in line with the developed framework by Cransberg et al. (2016).

(10)

10

3. Methodology

Literature wherein complexities and resulting implications have been investigated in practice is particularly scarce. One of the reasons is that the majority of WLC literature has been based on simulation studies (Thürer, Stevenson and Silva, 2011), and only recently more attention has been spend on identification and handling complexities (see e.g. Alblas, 2014; Cransberg et al., 2016). Only a limited number of practical solutions for handling complexities are proposed in WLC literature, and implementations of solutions to mitigate implications are very limited, a gap between theory and practice remains.

The purpose of this paper is to generate knowledge concerning mitigating performance implications that result from nesting when applying WLC principles, thereby narrowing the gap between theory and practice. The methodology described in this section aims at providing answers to the research questions below:

1) What performance implications for implementing WLC principles result from nesting? 2) Why do these performance implications occur?

3) How can these performance implications be mitigated?

When aiming for solutions, field research is considered an appropriate method (Dehoratius and Rabinovich, 2011).Consequently, for this paper is chosen to carry out a case study in a high variety MTO company.

3.1 Case selection criteria

For a sound investigation of implications that result from nesting, an appropriate case company has to be selected. Therefore, the following criteria have to apply for the company in order to fit the purpose of this study. The case company should manufacture according the make-to-order strategy. Eventually, WLC is also considered as most suitable PPC concept for MTO companies (e.g. Henrich et al., 2004; Stevenson et al., 2005). Furthermore, the case company should make extensive use of nesting, and nesting should have a considerable amount of impact on shop floor performance in terms of delivery reliability and lead time. In addition, a certain degree of complexity that is usually seen in job shops should be present in order to reveal relevant implications.

In order to learn from the company’s PPC decisions to mitigate nesting implications, not every high variety MTO company with nesting processes can be selected. To be of use, the company should be exemplary for its industry as it is more advanced and mature compared to the typical high variety MTO companies. In other words, the case company should apply advanced methods and tools for controlling their operations and nesting process. Additionally, the company should be able to provide useful data for the analysis. Section 3.2 below provides a brief description of the company that has been selected for the case study.

3.2 Company description

Company characteristics

(11)

11 assembling. In addition, quality inspections and logistical processes are performed to assure high quality and timely delivery of orders. Their product routings show typical job shop characteristics as shown by the from to matrix in Appendix I. The shop floor layout is characterised as a process or functional layout.

This company is more advanced than other high variety MTO companies, as it is working in a very structured way by controlling its production with smart planning tools which are integrated in their ERP system. Their own IT department allows them to improve applications continuously to support workers in making better PPC decisions. For continuous improvement the company collects a wide variety of data, which is beneficial for this case study.

3.3 Data collection

In this study data will be collected through the following data sources, interviews with domain experts within the case company, direct observations and quantitative data exports from the company’s ERP system. The latter acts as main data source for the analysis, whereas the two former sources are used for clarification, validation and inspiration. Combining multiple methods whereby both qualitative and quantitative data is collected will strengthen the construct validity of the study (Voss, Tsikriktsis and Frohlich, 2002).

The collected data are stored in a database readily accessible by the researcher. The quantitative data contains information concerning individual orders. This includes, but is not limited to, measurements of order entry, order release to the shop floor, planned production times, actual production times, promised delivery dates and actual delivery dates. Actual time measurements are based upon the company’s scanning system. Scanning orders at the moment of arrival and dispatch before and after a workstation is integral to working procedures. The scanning system is integrated with the company’s ERP system to enhance traceability.

3.4 Data analysis

By collecting data from multiple sources more reliable data could be obtained which enhances internal validity and tend to lead to more reliable results (Voss, Tsikriktsis and Frohlich, 2002). Together with domain experts at the case company, data will be carefully reviewed and verified. Extensive data sets have been checked on outliers, which are removed when encountered by only taking into account data that falls within the 5% to 95% boundary.

To gain insight in performance implications caused by nesting when applying WLC theory, we want to focus on operations that are closely related to nesting. Operations are therefore analysed on the frequency they occur and how workload is distributed among them. Moreover, production orders are analysed on their routing to find which one is dominant. Combined with workload distribution and frequency that operations occur a focus area is determined.

(12)

12 of which at least one activity has been performed in 2018, have been used. In order to relate findings to nesting, multiple input and output curves have to be drawn - related to different moments in the progress of orders - and the laser cutting process will be analysed in great detail as it is directly related to nesting. This step gives insight in which performance implications result from nesting.

The performance analysis is followed by analysis of the planning structure of the case company graphically in a business process model. A detailed description of the planning process is made to provide insight in PPC decisions on all hierarchical control levels. Moreover, the decisions made on each hierarchical control level during the planning process are related to how the nesting complexity is currently being handled by the case company. Decisions related to handling the nesting complexity and their impact are backed up by former findings in the data analysis. Additionally, domain experts have been involved during this step to strengthen the validity of findings. In this step PPC decisions made by the case company provide answers to why certain performance implications that result from nesting occur. Moreover, it forms the foundation for PPC decisions to mitigate nesting performance implications.

3.5 Developing PPC decisions

(13)

13

4. Results

In this section results of the empirical research are discussed. It starts with determining in which area of the production process nesting has the greatest impact. This part is followed by the analysis of the planning structure wherein PPC decisions to handle nesting are linked to performance implications. By linking PPC decisions with performance we gain insight in which performance implications result from nesting, why these occur and what impact they have. Moreover, we evaluate PPC decisions made by the company to mitigate performance implications.

4.1 Operations affected by nesting

For the performance analysis, we want to focus on the operations that are affected the most by nesting. Therefore, the current production process has been analysed on frequency the operation occurs and distribution of workloads among operations. Figures 3a and 3b, show the main operations performed by the company in 2018 by means of a pareto diagram. Figure 3a depicts the frequency of operations that have been performed. Figure 3b illustrates the workload per operation expressed in percentage of total workload. Shown in Figure 3a, the most frequently performed operations are quality control, production planning, laser cutting, packaging, folding, grinding, machining and welding. In Figure 3b we see that operations with highest workload are welding, laser cutting, machining, folding and montage.

The above figures provide already some insight on which operations can be focussed as these are most frequently visited by production orders and process the highest workloads. The analysis of followed routings by production orders in 2018 has resulted in Figure 4, wherein the dominant routing has been depicted. The percentages on the arcs relate to the quantity of production orders that moved from one operation to the other and are based on the total number of production orders in 2018. Percentages in the operation boxes account for the number of production orders of 2018 that have

Figure 3b: Pareto analysis operations (workload). Figure 3a: Pareto analysis operations (frequency).

Figure 4: Overview most dominant routings of production orders in 2018 and focus area. Closely related

(14)

14 visited that operation. For example, in 2018, 89,7% of production orders visited the laser cutting operation, of which 80,37% of all production orders moved from production planning to laser cutting. Taking into account the above, the focus for performance analysis will be on laser cutting, grinding / deburring, machining and folding operations (see Figure 4). These form the focus areas because these are most closely related to the nesting complexity as these follows directly from laser cutting and are part of the most frequently visited work centres. Additionally, data of these operations are the most accurate compared to that of other operations in the dominant routing. For more detailed information concerning routing data, we refer to Appendix I, which contains a from to matrix of all operations performed by the case company.

4.2 Performance implications

For analysing the delivery reliability, lateness has been analysed of operations within the focus area. Additionally, lateness and throughput time are analysed with the throughput diagram, which also illustrates the link between them. The link with nesting and performance implications is discussed and supported with tables and figures.

4.2.1 Delivery reliability

For the lateness analysis, a dataset containing production orders that have been accepted and completed in 2018 has been used. Figure 5a illustrates the average lateness distribution of production orders started in 2018, based on actual order completion and planned order completion. In Figure 5b, a lateness distribution is depicted, based upon actual order completion and promised order completion.

The average lateness in Figure 5a is -1,15 days (var. 13,8). Figure 5a shows that quite a large percentage of orders are tardy (28,9%). The left side of the curve shows a pattern that corresponds with the nest horizon (i.e. the time between earliest and latest starting date of orders in one nesting) of production orders as shown by Figure 6. In this figure we can see that for a large part of orders the nest horizon is between the one and seven days, of which the effect can be seen at the left side of the curve in Figure 5b. The applied FCFS dispatching policy enables this pattern as this policy does not interrupt with order release decisions.

In Figure 5b the average lateness is -3,74 days (var. 34,34). This indicates that between planned order completion and promised order completion an average slack is exists of approximately 2,6 days. Although the realised average slack between planned and promised order completion is 2,6 days, the curve does not have a similar shape. Despite reducing tardy orders to 21,9%, variance has increased from 13,8 to 34,34. This is revealed by the increased spread in negative lateness. The assigned slack to production orders, i.e. the planned time buffer between planned order completion and actual order

Figure 5a: Average lateness of production orders 2018 based on planned delivery dates.

(15)

15 completion, is 3,82 days (var. 15,12).

This is much in line with the realised average lateness, however we see in reality variance has increased. This could be caused by nesting as orders are pulled forward and processed further according a FCFS. However, when relating this back to Figure 6, we see that on average orders are pulled forward 2,09 days (var. 9,33) and only small percentage of jobs has a greater nest horizon than 8 days. It is more

likely that variance increased due additional slack that is assigned when production orders if routing includes operations as montage, welding or an outsourced operation. Additionally, Figure 5b illustrates that decrease of tardy orders is only marginal, indicating that efforts to deliver orders on time has been spent to the wrong orders.

To understand how nesting affects lateness performance, similar distributions of lateness have been constructed for operations in the focus area in Figure 7. The lateness of these operations is based on actual operation completion and planned operation completion. Table 1 specifies the lateness and variance per operation. In the figure can been seen that lateness of the operations in the focus area is not in line with the average lateness.

This indicates that on average more downstream operations in the routing are speeded up (i.e. completed sooner than planned) or the planning included sufficient slack to correct lateness. Moreover, the laser cutting process seems to cause lateness of its successive operations as their lateness increases. This also indicates the slack between these operations is very narrow and therefore only tiny corrections could be made in terms of lateness. Additionally, the applied dispatching rule FCFS, does not allow for changes in sequence after release which makes correcting the lateness more difficult. Lateness performance could also be related to orders that start too late (46,2%), because more time was required for pre-production preparations or material was not on stock at the planned starting date. This can also apply for orders tardy and average tardiness. When the production process already starts late, it obviously becomes harder to deliver orders on time. When we relate the lateness performance of laser cutting to nesting, we would expect lateness to be negative, as orders are pulled forward therefore starting sooner than planned and also be completed sooner than planned. However, this effect does not seem to apply. This can be explained by orders that start too late or by high direct load at the laser cutting machine, due pulling orders – and therefore workload – forward. This would increase queueing time for jobs increasing chances to be completed too late at the laser cutting operation. Nevertheless, more evidence is required to relate the lateness performance to the nesting process.

Figure 7: Average lateness of production orders 2018 per operation.

51,9% 7,3% 8,5% 7,6% 7,5% 5,0% 2,7% 3,1% 1,2% 1,2% 0,9% 0,8% 0,7% 0,4% 0,7% 0,4% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0% 10% 20% 30% 40% 50% 60% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Pe rce ntag e o f p ro du cti on o rd er s Nest horizon

Nest horizon production jobs 2018

(16)

16 Operation Avg. Lateness (days) Var. Lateness (days) Tardy orders Avg. Tardiness (days)

Laser cutting 0,66 16,3 53,4 % 3,49

Folding 0,84 16,0 48,0 % 4,04

Machining 0,73 21,0 46,1 % 4,22

Deburring 1,17 13,9 52,2 % 3,90

Complete order -1,15 13,8 28,9% 3,16

Table 1: Overview of lateness and tardiness in 2018.

The nest horizon is an important decision made during the nesting process, as this determines the spread in urgency of orders released to the shop floor. When plotting lateness against the nest horizon, a relation between them can be identified. If the nest horizon increases, it correlates with a decreased average lateness. Figure 8 illustrates this trend. The negative relation between nest horizon and average lateness based on order completion and planned completion (-0,198), and between nest horizon and order completion and promised delivery (-0,190), are not perfect, but are significant (p < 0,01). The relation between nest horizon and lateness seems somewhat obvious as for nesting efficiency reasons, orders are pulled forward (i.e. orders start sooner than planned) and therefore are more likely to have a lower lateness when the nest horizon increases. Especially when applying a FCFS dispatching rule. The effect however begins to level around a nest horizon of 16 days. When a nest horizon becomes greater than 16 days, we see average lateness to increase again.

Plotting nest horizon against average throughput time, a relation can be identified between them. Figure 9 shows a positive relation between nest horizon and average order throughput time (0,256), which is significant (p < 0,01). It seems to keep throughput time as low as possible, the nest horizon should be kept is small as possible. To see how many orders are pulled forward, we refer back to Figure 6. This figure shows that the number of orders pulled forward is actually relatively low. More than 90% of the production orders fall within a nest horizon of 6 days and almost 52% of production orders that are nested have the same starting date. An explanation can be that by pulling production orders forward for increasing nesting efficiency WIP levels increase and therefore throughput time becomes longer. The next section addresses performance in terms of WIP and throughput time in relation with nesting in more detail.

4.2.2 Throughput time

In the throughput diagram cumulative input and output are presented on different moments in time. The horizontal axis of the diagram represents cumulative time in days, whereas the vertical axis indicates the workload measured in hours of an operation. The throughput diagram is applied to gain insight in PPC decisions that caused deviations from the ideal situation. Deviations can be derived from

Figure 8: Nest horizon vs average lateness.

Figure 9: Nest horizon vs average throughput time.

-8 -7 -6 -5 -4 -3 -2 -1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Av er ag e l ate ne ss (d asy )

Nest horizon (days) Nest horizon vs average lateness

Actual vs planned Actual vs promised

0 2 4 6 8 10 12 14 16 18 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Av er ag e th ro ug hp ut ti me (d ay s)

Nest horizon (days) Nest horizon vs average throughput time

(17)

17 sudden peaks in the cumulative input and output curves. Ideally, when applying WLC principles, average throughput times and small constant buffers or queues on the shop floor are desirable. While successfully maintained, a limited stable WIP level is present on the shop floor, indicating a stable output (Soepenberg, Land and Gaalman, 2012c). This will show a relative steady line in the throughput diagram.

In Figure 10, the order pool can be derived from the vertical distance between the acceptance curve (1) and order release curve (2). WIP can be derived from the vertical distance between the order release curve (2) and operation completion curve (3). The horizontal distance between these curves give an indication of the average operation’s throughput time (Soepenberg, Land and Gaalman, 2008). Comparing the operation completion curve (3) and planned operation completion curve (4), it shows how many hours have been completed on time (as planned) and how many are too late, indicating whether the operation’s planned due date has been set with too much lenience or too tight (Soepenberg, Land and Gaalman, 2008).

As the laser cutting operations is directly influenced by the nesting process, we start by analysing the throughput diagram in laser cutting hours. In Figure 10, four curves represent, (1) order acceptance, (2) order release, (3) order completion and (4) planned order completion.

The figure below shows that the acceptance curve (1) is characterised by some large fluctuations, although, most fluctuations are absorbed by the order pool. While observing the order release curve (2) we see this curve is much smoother compared to the order acceptance curve (1). Interestingly, the operation completion curve (3) is fluctuating as well. This indicates the output of workload coming from the laser cutting operation is not stable and causes an unstable WIP pattern, since order release is not always adjusted accordingly. Order release (2) and operation completion (3) do not seem to

1 2 3 4

(18)

18 complement each other. The order completion does not seem to keep up with the order release, which is shown by the fluctuating difference in vertical distance between the two curves. The horizontal difference between these curves is increasing at certain moments in time as well, which increases the operation’s average throughput time.

The variations in WIP in the order pool and WIP on the shop floor of the laser cutting operation can be seen in more detail in Figure 12. The average WIP in laser cutting hours in the order pool is 352,16 hours and WIP at the shop floor at the laser cutting operation 176,96 hours. As indicated by the graph, the WIP levels are not stable and show some major fluctuations on a daily level. This can be seen in the standard deviation of the WIP in the order pool and WIP at the laser cutting operation. These are 98,29 hours and 69,58 hours, respectively.

To see what influences the laser cutting operation has on successive operations, similar overviews of WIP and buffer levels have been plotted for folding, machining and deburring in Figures 13, 14 and 15 respectively. These operations are greatly impacted by choices made concerning the laser cutting operations, because more than 86,6% of production orders at the laser cutting operation proceed to either folding, machining or deburring. The WIP in order pool and WIP on the shop floor are measured in hours of the operation. WIP at the shop floor includes direct load and indirect load of the operation, specified as aggregate load. Workload is considered direct load when the previous operation is finished up to completion of the operation itself. The aggregate load at the laser cutting operation is equal to its direct load as this is the first operation after order release. In Figures 13, 14 and 15, we see that WIP levels at the operations downstream of laser cutting show large deviations in aggregate and direct load levels. We also see that fluctuations in selected load that has been released are visible in the direct load of the downstream operations. Release decisions made for the laser cutting machine therefore have a high impact on the direct load of its successive operations. Table 3 summarises the measures of Figures 12, 13, 14 and 15.

Figure 13: WIP levels and direct load in folding hours in 2018. Figure 12: Order pool WIP level and direct load in laser cutting hours in 2018.

(19)

19 Operation Measure WIP in order pool (hrs) WIP at shop floor (hrs) Direct load

(hrs)

Laser cutting Average 352,16 176,96* 176,96

Standard deviation 98,29 69,58* 69,58 Folding Average 424,79 291,89 148,18 Standard deviation 120,22 76,71 53,85 Machining Average 460,59 279,32 171,43 Standard deviation 107,77 67,47 57,81 Deburring Average 61,34 25,32 14,91 Standard deviation 18,87 14,28 10,05

Table 2: Average, St.Dev. of WIP in order pool, WIP at shop floor and buffer per operation. *equal to direct load because released workload is considered directly available for the laser cutting operation after release.

4.3 The planning structure

This section comprehends the analysis of the current planning structure of the case company related to nesting. In this analysis we evaluate how the trade-off between timing and balancing is formed at the case company and how nesting complicates the trade-off decisions. We evaluate PPC decisions made by the company to mitigate performance implications that result from nesting.

Figure 16 illustrates a simplification of the current situation the case company has to control by making PPC decisions. Each point represents a moment where PPC decisions are made in order to maintain control over production processes. Points 1, 2 and 3 represent PPC decisions moments related to nesting, as the nesting process is performed before order release. Points 4, 5, 6 and 7 concern PPC decisions not directly related to nesting, but to handle nesting consequences. Below a brief description is given which addresses the most important PPC decisions made.

At point 1, decisions that concern assigning delivery dates and production starting dates to accepted customer orders are made. A delivery date is based on available capacity per day, order workload, material availability and requested delivery date by the customer. By taking into account a time buffer of 3 days between planned order completion and promised delivery date, a starting date can be determined as well. The time buffer increases the chances of timely order completion. The assigned production starting date, and/or assigned urgency (typically on request of the sales department or customer itself for speedy delivery), determine priority. This priority determines the sequence in which orders are considered for the nesting process executed at point 2.

By taking into account the available capacity and order workload during assigning production starting dates and delivery dates, load balancing at a general level is applied. The company plans only with 80% of their capacity, so orders that are pulled forward due nesting can be processed without delay. If capacity of an operation’s work centre in the order’s routing is insufficient, its planned starting

(20)

20 date for that operation will be adjusted to the first day on which capacity is available again. When the production planner is satisfied with the planning, the order becomes available in the order pool and capacity planning will be updated.

At point 2, the nesting process is executed. At this point there is decided which orders will be combined in a nesting to form a job, based on priority determined at point 1. The standard procedure is to upload the complete order pool into specialised nesting software which allows orders to be combined to form a nesting. This software also generates the cutting programs that will be loaded onto the laser cutting machines. Usually, orders with a starting date up to three days ahead of the current date and can be cut from the same material are combined in a nesting. The nester evaluates the nesting on waste percentage. If not efficient enough, the nester has some lenience meaning orders can be pulled forward up to seven days ahead in time to include them in the nesting. If the nesting is still not efficient enough, orders will be nested in such a way that a residual piece can be created. Residual pieces are usually created for expensive materials as waste of these materials is much more costly. Residual pieces are not desired, as relative setup time increases when a residual piece has to be cut, and therefore machine efficiency is lost.

When it becomes busier, the nester narrows the nest horizon to one or two days and reduces lenience in nest horizon to four days. The nesting horizon becomes narrower in busy times, because the same amount of orders is available to combine in a nesting despite the narrower nest horizon. Moreover, it prevents laser cutting capacity is allocated to less urgent orders. An exception to the typically applied nest horizons is made when materials have to be cut from expensive or non-stock materials. In this case delivery time of the sheet material has to be included in the nest horizon, because material will be purchased after the nesting has been made to purchase the exact number of sheets required of that particular material. The nesting horizon can therefore increase significantly.

During the nesting process, the nester takes into account the workload for laser cutting hours that have been nested. The aim is to keep the workload at the laser cutting machine at a stable level by adjusting the nested workload for laser cutting. Workload nested for other operations and their current workload is neglected.

At point 3, decisions concerning release of jobs are made. Usually, all jobs are released to the shop floor on the same day. In the previous step, the nester already took laser cutting workload and its workload norm into account and therefore all orders could be released without exceeding the laser cutting workload norm. Workloads of successive operations are not considered at the release decision.

At point 4 jobs are assigned to laser cutting machines. Jobs are assigned to a specific laser cutting machine depending on material characteristics jobs and cutting time. Many jobs are able to be assigned to several laser cutting machines, which allows load balancing between laser cutting machines. At point 5 and 6 decisions are made concerning which jobs are combined in a handling unit and in which sequence these will be processed. The sequence in which jobs are processed is based on the planned production starting date (i.e. the earliest starting date of orders in the nesting). Decisions made at this level have minor impact as workload present at the shop floor cannot be affected at these points. Only marginal fluctuations in workload for downstream operations throughout the day can be realised by sequencing decisions.

(21)

21

4.4 Interpretation of results

Within the case company, we see that nesting is handled at the order release level, just before order release itself. As discussed in section 2.4, performing nesting at the order release can conflict with the traditional timing/balancing trade-off. This would result in an increase of less-urgent orders present at the shop floor and disruptions in flow as the nested set of orders cannot be split to balance load at the shop floor. Consequently, a balance has to be found between spread in urgency between orders in a nesting and nest efficiency.

In the results we have seen the company is dealing with large amount of tardy orders, especially for operations closely related to nesting. It is tricky to directly relate this to the nesting complexity as other pre-production preparations have caused orders to start late and thereby increasing chances of tardy orders. Moreover, the workload at the laser cutting machine is quite fluctuating, indicating order release is not adjusted accordingly or there is bad output control for the laser cutting process. It seems a combination is the case as fluctuations in order completion (see Figure 10) are caused by not having material on stock at planned starting dates. The release procedure therefore is not always taking material availability into account.

In general, we see that at the case company that a wide nest horizon correlates with a low average lateness. Despite that, the company aims to keep nest horizons low. When plotting nest horizon against throughput time, a low nesting horizon correlates with low average throughput times. Other reasons are that resource capacity is allocated to the most urgent orders and variance in lateness can be controlled. A higher spread in starting dates between orders in a nesting when applying FCFS dispatching policy should lead to increased variance in lateness as FCFS does not interferes with release decisions. When nesting would not be the case, sequencing of orders for release would not be disrupted and a FCFS policy would be more applicable.

The fluctuations in selected load that has been released are visible in the direct load of the downstream operations. Release decisions, which are primarily based on laser cutting loads, have a high impact on successive operations. This indicates only a marginal buffer exists between laser cutting and its successive operations. This should be taken extra carefully into account in the nesting process as performed at the case company. Currently, only laser cutting workload and laser cutting WIP at the shop floor is taken into account, whereas other operations are neglected. To control WIP at the shop floor of all operations these should ideally be included in the nesting process as well or considered at the order release.

At the order entry level delivery dates are assigned to orders based on a capacity planning. Within this planning the company plans with 80% of its capacity. This allows order to be pulled forward by absorbing the additional required capacity to finish orders in time with the remaining 20%. Other considerations related to nesting are not taken into account, as this is rather complex. A nesting is dependent on many factors which perhaps are even impossible to take into account in assigning delivery dates. One can argue if 20% of capacity is sufficient to cope with the additional that will be pulled forward by nesting. If the total daily capacity would have been allocated to orders, only orders that have the same starting date can be nested. Pulling orders forward for nesting efficiency purpose would mean more capacity is required than available and orders will start too late.

(22)

22 efficiency, spread in urgency between orders in a nesting and material cost. By allowing nesting for residual pieces, nest horizon can be kept narrow and material waste can be reduced.

(23)

23

5. Mitigating nesting performance implications

Based on literature and analysis of the case company’s performance and planning structure, the following PPC decisions have been proposed to mitigate nesting performance implications when applying WLC principles.

The framework of Cransberg et al. (2016) can be applied to decide at which hierarchical level it would be most appropriate to handle the nesting complexity. Although a remark should be made considering determining the criticality. In the current framework, criticality is divided into impact of release sequencing on output and whether the complexity affects the bottleneck. When the nesting complexity is considered, it is more appropriate to consider the impact of release sequence on cost of material waste. If release sequence impacts the nesting efficiency significantly and the material costs are high, the criticality can be considered high. If the impact of release sequence only slightly affects nesting efficiency impact of release sequence can be considered low.

When nesting is handled at the order release level, it is most likely orders will be pulled forward for nesting efficiency purposes. For assigning accurate delivery dates, taking into account a capacity buffer when planning an order can prevent tardiness. Additional capacity is required to process jobs when it contains orders that are pulled forward. For assigning an accurate starting date, it is important to include time required for pre-production operations and material availability to prevent orders starting too late and therefore become tardy.

At the order release level, the nesting performance implications could be mitigated in the following ways. By taking into account workload of orders, workload norms of operations and current WIP levels during the nesting process, balancing can be performed almost simultaneously with nesting. This requires however to have insight in workloads and WIP levels beforehand. Moreover, a trade-off between nesting efficiency, timing and balancing remains. When the sequence in which orders have to be considered is accurately determined on starting date, we propose two perspectives on handling the trade-off between nesting efficiency and balancing.

1. Focus on nesting efficiency

(24)

24

2. Focus on workload balance

When focus is on workload balance, orders are considered for nesting based on sequence of urgency. When workload levels have been reached, all selected orders will be nested. Some lenience in workload norms should be allowed, to improve nesting efficiency. However, workload balance should not become too variable. After jobs have been made, they can be released to the shop floor.

Depending on costs of raw material, variety of material and demand of material, a decision can be made between focus on nesting efficiency or on workload balance. When a company is processing expensive materials, in high varieties and low demand per variety, focussing on nesting efficiency would be more appropriate. When material cost is low, variety is low, and demand per variety is high, focus on workload would be more suitable. In general, there is expected that focussing on nesting efficiency will result in lower cost due material waste, but higher fluctuations in WIP levels compared to focus on work balance. For work balance we expect that nesting efficiencies are lower, but WIP levels are more stable due better balance of workload on the shop floor.

(25)

25

6. Discussion and conclusion

6.2 Discussion

This paper aimed at generating knowledge about the nesting complexity and proposes how performance implications that result from nesting when applying WLC principles can be mitigated. In this study we have indicated which implications can result from the nesting complexity when applying WLC principles, why these occur and how these can be mitigated.

In this study we found that in practice several complexities (e.g. nesting, batching, assembly) exist that interfere with the traditional trade-off between timing/balancing (Cransberg et al., 2016). Only limited research has been done on how to handle complexities when WLC principles are applied and proposed solutions are limited. Cransberg et al. (2016) have proposed a framework to determine at which hierarchical control level complexities should be handled. Based on position in routing and criticality, a hierarchical control level is proposed. To determine the criticality of a complexity, first to be determined is if it affects a bottleneck and second what the impact of release sequence is on the impact on output of the affected operation. Instead of relating impact of release sequence to output of the operation, it should be related to the impact on cost of material loss, when nesting is considered. This is the objective of nesting (Heistermann and Lengauer, 1995). Despite this remark, the framework remains applicable for addressing the nesting complexity.

The performance implications of nesting when applying WLC have been partially derived from literature as well as from results of the case study. Moreover, the case study gave insight why these implications occur and formed the foundation for the proposed PPC decisions to mitigate nesting implications. The performance implications encountered by the case company could not always be directly related to nesting. Discussions with domain experts and by ruling out other possibilities, we could infer that certain performance implications could be related to nesting, although not always directly visible in the data. To improve validity of the findings, a simulation study or multiple case study could be performed.

The proposed PPC decisions to mitigate performance implications that result from nesting are primarily based on findings in literature and performance analysis. However, these are mainly focussed on addressing the nesting complexity when the affected process is located upstream, and nesting is performed at the order release level. The applicability of the PPC decisions may therefore not be applicable when nesting affect an operation that is located downstream and is addressed at dispatching level. This provides opportunities for further research but limit the results of this study.

6.2 Conclusion

(26)

26 implications resulting implications when applying WLC principles. The following research questions have been formed.

1) What performance implications for implementing WLC principles result from nesting? 2) Why do these performance implications occur?

3) How can these performance implications be mitigated?

This explorative case study at an advanced high variety MTO company, has allowed us to gain insight in performance implications that result from nesting, why these occur and how these can be mitigated. The nesting complexity when not addressed carefully can result in increased fluctuations in WIP levels, variance in lateness and/or increased material costs. These relate to the trade-off formed between nesting efficiency, timing and balancing. All three need to be considered to mitigate the implications. On the order entry level where delivery dates are assigned to orders, additional capacity can be buffered to cope with the additional capacity required when orders are pulled forward. On the order release level, two approaches can be followed depending on product and demand characteristics. One approach focusses on reducing material costs, whereas the other focusses on load balancing. At the priority dispatching level, depending on spread in urgency of orders nested, different dispatching policies can be applied. When spread is small, e.g. First Come First Serve policy can be applied, whereas if spread is large dispatching rules that reduce variance e.g. Earliest Due Date first should be applied to reduce variance in lateness.

(27)

27

References

Alblas, J. (2014) ‘Exploring arising complexities when implementing workload control in a real life Make-to-Order company’, Master Thesis, (June), p. 44.

Chryssolouris, G., Papakostas, N. and Mourtzis, D. (2000) ‘A decision-making approach for nesting scheduling: A textile case’, International Journal of Production Research, 38(17), pp. 4555–4564. doi: 10.1080/00207540050205299.

Cransberg, V. et al. (2016) ‘Handling the complexities of real-life job shops when implementing workload control: A decision framework and case study’, International Journal of Production Research, 54(4), pp. 1094–1109. doi: 10.1080/00207543.2015.1047974.

Dehoratius, N. and Rabinovich, E. (2011) ‘Field research in operations and supply chain management’, Journal of Operations Management, 29(5), pp. 371–375. doi: 10.1016/j.jom.2010.12.007.

Fernandes, N. O. and Carmo-Silva, S. (2011) ‘Order release in a workload controlled flow-shop with sequence-dependent set-up times’, International Journal of Production Research, 49(8), pp. 2443– 2454. doi: 10.1080/00207541003720376.

Fredendall, L. D., Ojha, D. and Wayne Patterson, J. (2010) ‘Concerning the theory of workload control’, European Journal of Operational Research. Elsevier B.V., 201(1), pp. 99–111. doi: 10.1016/j.ejor.2009.02.003.

Heistermann, J. and Lengauer, T. (1995) ‘The nesting problem in the leather manufacturing industry’, Annals of Operations Research, 57(1), pp. 147–173. doi: 10.1007/BF02099695.

Hendry, L. C. et al. (2008) ‘Investigating implementation issues for workload control (WLC): A comparative case study analysis’, International Journal of Production Economics, 112(1), pp. 452–469. doi: 10.1016/j.ijpe.2007.05.012.

Hendry, L. C., Kingsman, B. G. and Cheung, P. (1998) ‘The effect of workload control (WLC) on performance in make-to-order companies’, Journal of Operations Management, 16(1), pp. 63–75. doi: 10.1016/S0272-6963(97)00011-9.

Hendry, L., Huang, Y. and Stevenson, M. (2013) ‘Workload control: Successful implementation taking a contingency-based view of production planning and control’, International Journal of Operations and Production Management, 33(1), pp. 69–103. doi: 10.1108/01443571311288057.

Henrich, P., Land, M. J. and Gaalman, G. J. C. (2004) ‘Exploring applicability of the workload control concept’, International Journal of Production Economics, 90(2), pp. 187–198. doi: 10.1016/S0925-5273(03)00126-9.

Kingsman, B. G. (2000) ‘Modelling input-output workload control for dynamic capacity planning in production planning systems’, International Journal of Production Economics, 68(1), pp. 73–93. doi: 10.1016/S0925-5273(00)00037-2.

Kingsman, B. and Hendry, L. (2002) ‘The relative contributions of input and output controls on the performance of a workload control system in make-to-order companies’, Production Planning and Control, 13(7), pp. 579–590. doi: 10.1080/0953728021000026285.

Mitigating nesting performance implications when applying workload control in high variety MTO companies – An explorative case study