A lean implementation of Workload control At Meijer sheet processing

(1)

A lean implementation of Workload control

At Meijer sheet processing

Author : Arjen Poppinga

Student number : 1480294

Study : Technology Management

Email : s148094@student.rug.nl

Faculty : Faculty of Economics and Business Supervisor : Prof. Dr. Ir. J. Slomp

Co-assessor : Dr. M.J. Land

Company : Meijer Plaatbewerking Supervisor : J. Oost

Co-assessor : H. Spoelstra

(2)

Preface

This thesis serves as my final research project for the Technology Management program at the University of Groningen. The project is best described by a quote of Mike Rother: “The research cycle never ends, of course, which means this text reflects a level of understanding at a point in time. There is more to learn and there are undoubtedly some mistakes here.” (2010).

In random order, I would like to thank the following people who have helped and guided me during the project.

Firstly I would like to thank all my colleagues at Meijer plaatbewerking. Special thanks to my supervisors John and Hans for their time, support and for making the project possible.

At the University, I would like to thank Jannes Slomp for his enthusiasm, inspiration and guidance throughout the project, Martin Land for his expert opinion and feedback and Nick Ziengs for his advice and help at the beginning of the project.

(3)

Management summary

This thesis is concerned with the implementation of a new production planning and control (PPC) system at Meijer Plaatbewerking; a supplier of semi-finished metal products. An increasing number of customer specific arrangements, high demand variety, small series production and growth of production capacity have resulted in a complex planning system at Meijer. The implementation of a new PPC system based on the Workload Control (WLC) concept is believed to provide the answer. By means of action research, small steps towards implementation of a new system are taken. Firstly a performance monitor is constructed, based on four performance objectives. The monitor enables performance measures, providing feedback on implementation efforts. After the performance measures are in place, three implementation steps are introduced. Firstly the nesting period is adjusted. Secondly a system is constructed to release jobs, based on available capacity at the shop floor. Thirdly the shop floor dispatching is improved, with the introduction of a new shop floor planning list.

A very important practical goal at the start of this research was to make Meijer less depended of its planner. Throughout this research a strategy surfaced to relief the planning process, consisting of a number of levels.

- Improving performance of the production system

A reliable production system with short throughput times needs less re-planning and adjustments.

- Decentralisation of planning responsibilities

A detailed capacity planning can be shifted to the shop floor. - Improving tools for planning

The goal is a decision support system which provides a foundation for adequate and timely decision making.

(4)

1. Introduction ... 5 1.1 Motivation ... 6 1.2 Research objective... 6 1.3 Methodology ... 8 1.4 Boundary conditions ... 9 1.5 Thesis outline ... 9 1.6 Résumé ... 10 2. Background ... 11

2.1 The production process ... 11

2.2 The production planning and control ... 14

2.3 Business performance ... 17

2.4 The workload control concept... 18

2.5 Résumé ... 19 3. Measuring performance ... 21 3.1 Lead times ... 21 3.2 Delivery reliability ... 22 3.3 Work in Process ... 23 3.4 Utilisation/throughput ... 24 3.5 In addition ... 25 3.6 Résumé ... 26

4. Changing the nesting period ... 27

4.1 Planning action ... 27

4.2 Developments... 28

4.3 Résumé ... 30

5. Introducing the release decision ... 31

5.1 The introduction ... 31

5.2 Release and ICT Support... 32

5.3 The release... 33

5.4 Developments after introducing the release decision... 34

5.5 Résumé ... 37

6. Shop floor control ... 38

6.1 Planning action ... 38

6.2 Evaluating Action: Shop floor control ... 39

6.3 Résumé ... 40

7. Conclusion and Recommendations ... 41

7.1 Conclusions ... 41

7.2 Recommendations for further implementation ... 43

8. Discussion and further research... 45

Appendix ... 48

Appendix 1 Initial design ... 48

Appendix 2 Organizational structure ... 55

Appendix 3 Products ... 56

Appendix 4 Shop floor layout ... 57

(5)

Introduction

This thesis is concerned with the implementation of a new production planning and control (PPC) system at a sheet metal production facility in the north of the Netherlands. The newly implemented system is based on the Workload control concept; A PPC system that has been thoroughly discussed in literature as robust and especially suitable in a job shop environment (see Land, 2004). Although literature on workload control dates back a number of decennia (see, for example, Bechte, 1988), the concept is not widely adopted in practice today. More recent studies have increasingly focussed on practical implementation of the concept (see, for example Stevenson and Silva, 2008). This thesis is in line with this trend, focusing on implementation of the concept at Meijer Plaatbewerking contributing to the still small empirical base of Workload Control.

Meijer was founded in 1921 as a small family company in Sint Jacobiparochie. In its first years it evolved into a small machine factory, mainly aimed at producing agricultural equipment. In later years the company performed a great variety of activities, gradually increasing in size but remaining a family business. In the nineties two activities remained, divided over two separate companies. The first company, “Meijer sheet processing”, grew out to be a supplier of sheet metal-, trunk-, tube- and bar-products. The second, “Meijer Special Equipment” (MSE forks), the adjacent sister company evolved into a producer of telescopic forklift attachments. The two companies are still a family business, have close links and collaborate to a great extent. This research will only be concerned with Meijer sheet processing.

Meijer occupies approximately 120 employees, all based at the facility in Sint Jacobiparochie. The company produces a wide variety of products out of sheet metal and tubes, serving a wide selection of customers. Products produced by Meijer are mainly half fabricates. Examples are frames for egg sorting machines, frames for milking robots, parts for bicycle carriers as well as for example a one-off memorial plaque. Pictures of more products are included in appendix 3. In 2008 this resulted in a realized turnover of nineteen million euro’s. An overview of the organisation structure is included in appendix 2.

(6)

1.1 Motivation

By answering to market requirements, in the last decade Meijer has experienced a shift from production in large series, to a make-to-order (MTO) environment. Globalization resulted in the emergence of competition from production facilities in low-wage countries. In cases of large production series where price is of great importance, it became impossible for Meijer to compete with these foreign production facilities. The tighter market demands resulted in a shift of competitive factors. A focus on small series with high delivery reliability and low delivery lead times became of greater importance. As a result, a larger variety of products in small series became the standard. Customers constantly renew the specifications of the parts they order and also order a great diversity of parts, resulting in a customer demand that is highly variable in both frequency and quantity of orders. Kingsman (2000) characterises such a company as a versatile manufacturing company; A company that has to supply a variety of products, usually in small quantities, varying from a range of standard products to all orders requiring a customized product.

The vision of Meijer is that there will only be room in Western Europe for manufacturing companies who closely collaborate within chains; and in this way find the solutions to differentiate from competition worldwide. The mission of Meijer is, to be the first choice for the manufacturing industry with the best “total solution”, by focussing on reliability, quality, logistic service, lead-time reduction and chain integration. This strategy of close collaboration results in an increasing number of customer specific arrangements. In combination with the steady shift to small series, high variety and the growth of production capacity, a complex planning system arose. As a result, the efficacy of the planning has suffered. Currently only the planner is capable of performing the planning activity and absence of the planner leaves a risk to the company of not meeting its strategic goals. Focussing on on-time delivery and short lead-times, stresses the need for high performance of the (PPC) system even more.

This led the management of Meijer to investigate the current production planning system. In 2009 Remon Peters (technology Management student at the University of Groningen) conducted a master thesis research on the subject of the production planning at Meijer. The goal of the research was to find a suitable PPC system for Meijer. After an analysis of the production environment, the current planning system, and a thorough literature research, the implementation of a Workload Control PPC system was recommended. Almost two years have passed since the research, without further implementation of a workload control based PPC system. The management at Meijer is now interested in the actual design and implementation of a workload based production planning. The former research already provided the company with a diagnosis of the problem and a recommendation. Therefor the approach taken is this research is somewhat different, taking the former research as a starting point and focussing on the implementation of a new system.

1.2 Research objective

After the former research, management believes that workload control can help in achieving higher efficacy in the planning process and meeting their strategic goals. How a new WLC system should be implemented is still unclear. This leads to the goal of this research as stated below.

The goal of the research is to provide an implementation plan for a “workload control” system, in order to increase business performance

(7)

increasing exogenous variability as a result of the changing market environment. Examples are variable processing times, order quantities and demand variations. The challenge of the production planning and control system is to align this demand variability with the production resources in order to maximize the business performance. A detailed description of what is meant by business performance is also described in the next section. The implementation of workload control is expected to change the production planning and control system, resulting in a new system that also has the objective to align supply and demand in order to maximize performance. The exact design of the new system and all of its variables is this part of the research still unclear though. As shortly appointed before, WLC is not yet widely adopted in practice and there is still a lot to learn about implementation. The WLC system design will in this way emerge throughout the implementation process.

“There are perhaps only three things we can and need to know with certainty: where we are, where we want to be, and by what means we should manoeuvre the unclear territory between here and there. The way from where we are to where we want to be next is a gray zone full of unforeseeable obstacles, problems, and issues that we can only discover along the way.” (Rother, M, 2010)

Figure 1.2.1: Conceptual model

Production system

Dem and variability

Production planning

and control Business Perform ance

Im plem entation Process

Production system

Dem and variability

(8)

1.3 Methodology

A structured approach to realizing the research goal and “to manoeuvre the grey zone” is provided by Action Research (AR). AR is an approach to research that aims both at taking action and creating knowledge or theory about that action (Coughlan and Coghlan, 2002), also typified as research in action as opposed to research about action.

“Action Research is appropriate when the research question relates to describing an unfolding series of action over time in a given group, community or organisation; understanding as the member of a group how and why their action can change or improve the working of some aspects of the system; and understanding the process of change or improvement in order to learn from it” (Coghlan and Brannick, 2001).

During the research the aim is to explore the assumptions and understand the relationships, through action research, resulting in an increased ability to make choices concerning design and implementation of the new PPC system. The implementation process will eventually determine what the new WLC system will look like

AR aims at developing holistic understanding during a project and recognizes complexity. The action researcher is participant in the implementation process. Coughlan and Coghlan (2002) present an AR cycle for operationalization of this methodology. By following the cycle, the new PPC system is built while simultaneously building up a body of scientific knowledge. The cycle consists of three steps. 1. A pre-step – to understand context and purpose of the project. For gathering of information active involvement on a day-to-day basis in the organisational processes is combined with desk research. The ERP database and the research performed in 2009 provide a source of quantifiable data. Other examples of data gathering are interviews, discussions, informal conversation and observation. 2. Four basic steps

• Diagnosing, involves naming what the issues are. This is the basis on which action will be planned and taken. As the implementation process continues it is expected that issues will change.

• Planning action. Planning action follows from the two former steps. In this part, the first steps towards the implementation will be taken.

• Taking action. Implementing the planned action involves making the desired changes and following through on the plans in collaboration with relevant key members in the organisation.

• Evaluating action. Reflecting on the outcomes of the action taken in the former step. 3. A meta-step to monitor.

(9)

Through action learning the new PPC system is implemented, evaluated, revised and rolled out to a wider constituency to enact it. The approach facilitates learning through a collaborative process. When the four basic steps are taken, a new cycle can be started. The actions taken have led to changed situation accompanied by a changed set of issues. An iterative process leads to optimization of the process. Hence the opportunity for continuous improvement exists; the ability to move toward a new desired state through an unclear and unpredictable territory by being sensitive to and responding to actual conditions on the ground( Rother, M, 2010). This is in conformance with the lean philosophy, which has been a basis for improvement efforts at Meijer for a couple of years. Cycles recur as particular actions are planned and implemented, resulting in an on-going iterative process towards optimization.

1.4 Boundary conditions

Boundary conditions that the research and the new PPC system design have to conform to are constructed next. The conditions must ensure that the scope of the research is limited and clarify the direction of the research. Because the conditions are very diverse in character, they are structured by type.

Scope limitations

- The research is limited to a time span of five months. Keeping the research within this period requires a sharply set scope and boundary conditions.

- The research is limited to the planning of the production process. Other activities that increase lead times, like engineering and work preparation, are left outside the scope of this research.

- Because complete implementation is too extensive for the research period, the succession of the project is included in a subsequent study.

Design conditions

- Efficacy of the system is of great importance. The planning procedure has to become simpler to provide an easy insight in the current situation and also executable by other personnel. - The proposed system has to be in line with the competitive strategy of Meijer, enhancing

performance.

1.5 Thesis outline

(10)

1.6 Résumé

(11)

2. Background

This section provides the background to the conceptual model presented in section 1. In the AR Cycle of Coughlan and Coghlan (section 1.3) this relates to step one, by providing context to the research. A detailed assessment of the company and its resources, which provides an understanding of the research environment, is presented. Subsequently, the current production planning and control system is discussed. And additionally the concepts of workload control and business performance are further elaborated on.

2.1 The production process

This section gives a short overview of the production process at Meijer, to provide a better insight in the company’s operations. Visits to the shop floor, discussions, the company information system and a former research performed at Meijer in 2009 serve as input for this description.

The production facility of Meijer is arranged in a functional structure, machines performing similar operations are grouped together in departments. Four functional departments are identified at Meijer, with in addition a dispatching department. A shop floor map included in appendix 4provides an overview of the facility and the positions of the departments and machinery.

The routing of products through the factory is characterised by a lot of variation in length and processing times. Production is typically performed in large quantities of small series. This shop floor configuration is in literature often identified as a job shop. When looking closer at the production system, despite the variations, a general flow can be identified. The gross of the products undergo the operations in a certain sequence. This can be seen in a from-to-matrix and flow analysis included in appendix 5. In a general flow shop a movement between any combination of two stations may occur. The difference is that the flow has the same direction, as can be seen in Figure 2.1.1 (Oosterman et al., 2000).

Figure 2.1.1: Flows in (a) the pure job shop; (b) the general flow shop

(12)

The structure of the network of order-routings is an issue in the character of problems of logistical control on a factory level. Convergent flows originate in case of assembly (Bertrand et al, 1998), when multiple parts are required before an operation can be initiated. In case of the Meijer production process, two convergent flows are identified. In the welding and assembly department, orders can only be started when all needed parts have arrived. An additional convergent flow is identified at dispatching. Multiple parts belonging to one order come together for shipping to the customer.

Inventory is stored in racks on the factory floor. The racks have three different colours, blue is for raw material, green is for finished products and red is for Work in process (WIP). After an operation, semi-finished products are put in a rack for the next department in its routing. In the receiving department the processing on the part can begin when capacity is available.

The Cutting department is in a large amount of the cases the initial operation on the production routing. At this department, the required shapes are cut out of metal sheets or tubes. To perform this operation, the department houses a number of machines. First of all the department consists of four Bystronic laser-cutting machines, one of which is especially used for stainless steel. These machines account for the largest part of the sheet metal processing. Additional are two CNC punch- nibble machines that can also be used for sheet metal processing. Two CNC tube lasers are used for the cutting of the metal tubes, recently accompanied by a third laser.

For the lasers cutters to cut a shape out of the raw material, an order has to be programmed first. Programmers prepare orders that have not been produced before or make alterations to a re-order if necessary. When programmed, the laser cutters have the possibility to produce unattended. In this way batches can be produced after normal working hours or in the weekends.

A nesting procedure is used for the laser cutting of metal sheet; this is the combining of orders that can be cut out of a sheet of metal. At the tube lasers optimal material use is also pursued, but to a less extent. The main aim of the nesting procedure is to maximize material utilisation and minimize setup times. Clustering products with similar sheet thickness and material characteristics results in an output sequence of jobs based on material characteristics. The products that require subsequent processing steps will thus arrive at the next station on a moment in time depending on their material characteristics. Before the products can go on to a next operation, they have to be separated from the sheets manually. This can be done by low skilled workers or operators that are temporarily unoccupied.

The second functional department is the press brake department. It consists of a total of fifteen CNC controlled press brakes. Three of the press brakes are especially equipped for Non-ferrous metals and Stainless steel and are located at the non-ferrous section. There are also variations in power of the press brakes, making not every job executable on every press brake. Operators are mostly assigned to a particular press brake. The machines are aligned next to one another and are mainly supplied by the adjacent cutting department described before. When jobs are processed they are stored in racks, waiting to be picked up for a next processing step or to be sent of to the dispatching department. Dispatching is the last step before the product is shipped to the customer. Quality checks are made prior to internal transport to this department. Products that are part of the same order are picked together, for joint shipment. Customer orders can generally only be transported when the order is complete; this means that there is a possibility that orders have to wait until a batch is complete.

(13)

in several capacity groups, this will be elaborated later. Groups are formed based on welding skills because not all jobs can be done by every welder. The distinction provides a more accurate insight in available welding capacity.

For some welding operations, a fixture is used. These fixtures are made in a special workshop that is part of the work preparation department. In case of a new order these templates have to be made prior to the welding process.

Additionally to the functional groups are a number of other non-initial-non-assembly operations. The additional machines are drills, a grinding machine, a winding drum and a milling machine. These operations are mostly performed somewhere in the middle of the routing of a job.

(14)

2.2 The production planning and control

Where the former subsection described the production process; in this section the subject matter is narrowed to the production planning and control of Meijer.

In general the planning is characterised by complexity and uncertainty. The planning process embodies innumerable variables and irregularities that the planner has to take into account when performing the process. Giving an exhaustive list of all these complexities is perceived as an impossible task in the timeframe of this research. This section is for that reason aimed at giving a general idea of the current planning system. The action research methodology described in the former section prevents a generalisation of the research, by encountering the complexity in action and overcoming the issues one step at a time.

Figure 2.2.1 is a planning framework based on the planning framework of Bertrand, Wortman & Wijngaard (1998). It gives an overview of the different hierarchy levels of a general planning process. This short theoretical passage is included, to provide structure for the description of the production planning at Meijer, treated subsequently. The high level control in the framework is concerned with long term strategic planning, like for example capacity acquisition, maximum lead time and maximum WIP levels agreements. Material coordination is about accurately setting the delivery due dates and accurately timing the arrival of parallel flows of assembly parts. Throughput time norms can be useful in this case, but already accepted work also has to be taken into account. Possibly capacity adjustments have to be made or due dates have to be reset. Workload control has a strong link with material coordination. When the workload exceeds the capacity, it has to be reduced or the capacity has to be adjusted. This can be done by realizing extra productivity, by for example using extra temporary employees, or working overtime. Material- and work release refers to the release of orders to the shop floor in accordance to the material and workload coordination. Lastly, shop floor control, at the bottom of the figure, is concerned with the detailed planning and is the responsibility of the shop floor supervisors.

High level control and production planning decisions are the responsibility of the management and left outside the scope of this description. An ERP software package currently assists the planner. The ERP system is a sophisticated tool, but is not perceived as sufficient for performing the production planning. For the level of workload control the planner constructed an additional Excel spread sheet, in order to create insight in the capacity available on the shop floor. The spread sheet includes eighteen different capacity groups. The four floor supervisors communicate the weekly specifics of

(15)

available capacity of the groups. If for example, employees have time off, the hours are adjusted in the available capacity planning. Based on this information, and the necessary capacity, weekly capacity adjustments are made. When capacity seems insufficient the planner attempts to increase or decrease capacity. A limited amount of workers can be shifted between departments. In addition temporary workers from an employment agency can, if necessary, complement the standard workforce. Feedback on the capacity adjustments is then briefed to the floor supervisors in advance of a calendar week.

The material coordination level is concerned with accurately setting the delivery due dates. When an order is placed, the customer communicates an expected delivery date. If an order cannot be filled from stock (for a small amount of products Meijer holds a minimal stock), the planned throughput time of the production process is calculated to give an indication of the delivery date. An order of an article that is new to the company is prepared for production by the engineering department. Subsequently, the order is passed on to the planner, who checks if the due date required by the customer can be met. The purchasing department gets the green light for acquisition and the order is checked for timely availability of necessary materials. For assembly orders, the routing of the assembly part with the longest throughput time is checked for setting the nearest delivery date. Throughput time of a job for a certain system is defined as the time between its arrival at the system and the time it leaves the system (Land, 2004).

A planned throughput time of three days per operation for internally processed parts and five days per operation for outsourced production steps is used for this calculation. The actual processing time has little influence on the planned throughput time, only assembly operations with very large processing times get additional days. For an order requiring six different operation steps, a total minimum of eighteen days will be reserved for shop floor processing. In reality the actual processing time is perhaps no longer than two hours. The remaining portion of the planned throughput time, the order will be somewhere in the production process, waiting on a shelf.

(16)

The shop floor control is, at Meijer, in the hands of the shop floor supervisors. Once the planning is released to the shop floor, the floor supervisors have autonomy to plan the initiation of processing an order. The shop floor supervisors decide on “who, where and when”, based on local dependencies. For example, at the laser cutting department the material characteristics are an important variable, resulting in nests. At the press brake department some operators largely perform the orders of specific customers; other orders can only be processed on a certain machine. A buffer of jobs on the planning list, enable the shop floor supervisors to create local optimums, based on their experience and insight. Figure 2.2.2 shows an example of a shop floor planning list.

(17)

2.3 Business performance

In the research goal and the conceptual model it is made clear that the new PPC system should help increase the business performance. Choices made in the design of the new PPC system are expected to have an effect on the performance. As a result the success of implementation efforts can be evaluated by monitoring business performance. Currently a number of performance objectives are already measured, but these measures are perceived as insufficient for accurate monitoring of the system. For this reason the concept of performance is extended in this research. The research subdivides the performance into a number of performance objective definitions, based on literature, in order to make the term more concrete. The four performance objectives for a job shop (Land, 2004) are used.

Short lead times

The lead time of a given routing or line is the time allotted for production of a part on that routing or line (Hopp and Spearman, 2000). It is in contrast to throughput time an appointed time period and it could be set differently if this is found necessary. From a competitive point of view, the date should be set as early as possible. As mentioned earlier this performance objective has become of increasing importance in the competitive environment. Despite this fact, there are currently no measures of lead time performance at Meijer present.

High delivery reliability

Delivery reliability is of major importance in an MTO environment. In contrast to ‘Make-To-Stock’, orders cannot be delivered out of a finished goods inventory. Accurate setting of delivery due dates in combination with accurate setting of lead time, is of major importance for delivery reliability. The On Time In Full (OTIF) measures at the different departments, currently give an indication of the delivery reliability to internal/external customers. Although these measures are a step in the right direction, they only give information about the fraction of orders delivered on time. To what extent an order is delivered early or late is not measured.

Low inventory level

Inventory comes in different forms, examples are raw materials-, WIP- and finished goods inventory. Most important in the job shop environment is the WIP inventory (Land, 2004). A lower average WIP level translates into shorter manufacturing cycle times (Hopp and Spearman, 2004). Regulating and minimizing WIP creates a better flow of orders through the factory by minimizing waiting time. Similar to lead time, Meijer does currently not keep track of WIP levels.

Utilisation/ throughput

(18)

2.4 The workload control concept

Lastly, this section explains what is meant with the Workload control concept. The discussion is in this part of the research limited to the theoretical concept of WLC, because there are currently no aspects of workload control present at Meijer. A more detailed assessment and operationalization will be given later on.

Workload control is a relatively simple concept based on the relationship between workloads and throughput times. By controlling the workloads of a certain capacity group or workstation, the throughput times of that workstation can be reduced.

A frequently used metaphor to explain the WLC concept is that of a bathtub. The bathtub represents a workstation and the water it contains represents the workload of the station. When water flows out via the drain, the water-level in the bathtub drops. By opening the tap, new water is released to re-establish the former water-level. Synchronising the inflow of water with the outflow will result in a constant water-level. A constant water level results in a stable and predictable time a certain amount of water stays in the tub before it flows out via the drain. When the water level can be held stable, the next step is gradually lower the water level, without the tub running dry.

Land (2004) describes the WLC concept as follows:

“WLC conceptualises the shop floor as a queuing system. In front of each work station, an arriving job finds a queue of jobs waiting to be processed. The principle of WLC concepts is to control the length of these queues.”

The shorter a job has to wait in queue, the shorter the throughput time of a job. Controlling the length of the queues is mainly done by controlling the release of jobs to the shop floor. The release of new jobs is triggered, based on the length of a queue in front of a work centre. When a queue length exceeds the workload norm, no new jobs that contain processing at that centre are released. Only when jobs on the shop floor are completed and new capacity becomes available, new jobs are released and join the queue. The shop floor is in this way subjected to a stable and predictable flow of workload. The main tool for establishing this stable load is the job pool; a buffer of orders that is ready for processing, but not yet released to the shop floor. In this way, the job pool intercepts the external dynamics off incoming orders, in order to prevent the shop floor from exposure to this variability.

(19)

2.5 Résumé

(20)

Implementation

The next section describes the actions that are planned and taken towards the implementation of a WLC production planning system. It includes a series of small steps in a pilot environment with the goal of learning and carefully monitoring the effects of the actions taken. The section starts off with the construction of performance measures to enable monitoring. After that, small steps provide the opportunity to tackle the complexities of the system one at a time.

Before WLC can be introduced as a suitable concept for a particular company, it is important to know its strengths and weaknesses with respect to the specific job shop characteristics. (Henrich, 2005) Appendix 1 provides for this reason an initial design of a WLC system that is the result of a literature study. In the design, terminology, background and basic issues emerge, giving more insight into the variables of importance when establishing a WLC system. This basic theory is used throughout the implementation process to motivate action steps and present key concepts to members within the organisation.

Because the researcher acknowledges that there is still a great gap between theory and practice, the design is only used to give direction and prioritization to the action research steps. Initial actions are planned based on the discrepancies between the desired future state and the current situation at Meijer. Next, the action research cycles, containing small steps, will bring Meijer closer towards adoption of a suitable design. Tailoring the initial concepts with every action step and assimilating on basis of action learning.

(21)

Throughput times Short routing 0% 5% 10% 15% 20% 25% 30% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Working days % o rd e rs 1 step 2 steps 3 steps

3. Measuring performance

Section 2.3 already mentioned four performance objectives of a job shop. In this section performance indicators are derived from these performance objectives. The indicators, combined form a monitor to analyse the effects of the actions taken. To establish a baseline measure, the system is already monitored for a period in advance of implementation. The required data is extracted from the ERP database tables. In addition to these quantitative measures, the shop floor is consulted for qualitative data. This data is important to put the quantitative data into perspective and check the reliability of the data.

It should be noted that the measurements are taken in practice. The system is for this reason influenced by numerous variables that are not under the control of the researcher. Consequently, the performance measures are also influenced by variables not concerning WLC.

3.1 Lead times

The delivery lead time is an important performance objective for Meijer based on its competitive implications. The importance of improvement efforts on this subject are stressed by the fact that in the period from August 2010 until January 2011, approximately 30% of delivery due dates were set later than desired by the customer. Apart from high delivery reliability, the desire to shorten lead-time is in this way communicated by the customer. Due to the narrow lead-timeframe, the scope of this study is limited to the part of the lead time resulting from time allotted for the production and planning process. Contributions to the lead time resulting from processes like drawing and engineering are for this reason not examined in detail, but recommended for further research. The actual time a job spends on the shop floor is called the Shop Floor Throughput Time (SFTT). The SFTT is measured as the time between the reported production start of an order and the time the order arrives at the dispatching department. In the current planning system orders are immediately released and a release moment cannot accurately be measured. For this reason, the initialization of production of an order is used as a start date for measuring the SFTT for the baseline measurements. Figure 3.1.1 shows the SFTT for the orders with one, two and three production steps in October. Because all orders have different routing lengths, the orders are grouped by the amount of processing steps they undergo in production. The rest of the descriptive statistics on SFTT, for orders containing more steps, are not included in this report but are part of the performance measures.

The SFTT can in turn be subdivided in several station throughput times. The throughput time of a 1 step Mean: 3.9 days Stdev: 3.1 days CV :0.80 2 steps Mean: 7.4 days StDev: 4.6 days CV : 0.63 3 steps Mean: 8.6 days StDev : 3.6 days CV : 0.42

(22)

Deburring 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Throughput time (working days)

% o rd e rs Drilling 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

% o f o rd e rs Press brake 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

%

o

rd

e

rs

station. Figure 3.1.2 shows distributions throughput times for three different departments in October. In the first department, deburring, 80% of the orders are processed within 3 days. The coefficient of variance of 0.65 indicates a relatively small variance of throughput time at this department. The press brake department shows a mean of 4 days with a standard deviation of 3 days and a slightly higher coefficient of variance of 0,70. 48% of the orders in this department have a throughput time longer than 3 days. Obviously this can cause problems when 3 days of throughput time are allotted in the planning system. At the drilling department the standard deviation is 2.3 days with a mean of approximately 4 days. The lead times that are perceived as too long by customers are in this situation not even long enough to ensure on time delivery. The next subsection elaborates further on this subject.

3.2 Delivery reliability

For measuring the delivery reliability, the indicator often used in practice is the percentage of jobs delivered late. Lateness is defined as the difference between the promised delivery date and the realised throughput time of an order” (Soepenberg, 2010). Notice that lateness can be positive (indicating a late job) and negative (indicating an early job) (Hopp and Spearman, 2000). By mapping the lateness distribution, two approaches for the reduction of lateness become apparent. The first is to reduce the average lateness; the second is to reduce the variance of lateness. The lateness distributions in figure 3.2.1 (Soepenberg et al., 2008) visualise these approaches. The vertical line marks the on-time delivery, with on the left side the negative lateness and on the right the positive lateness. The shaded area represents the tardiness. The tardiness is defined by Hopp and Spearman (2000) as the lateness of a job if it is late and zero otherwise. Tardiness is useful, because average tardiness can be used as a measure for delivery performance. With average lateness, for every job delivered late, a job delivered early can compensate. Because of this, low average lateness does not necessarily mean high delivery performance.

Figure 3.2.2 shows a selection of the lateness distributions of October for every department individually. The distribution on the left is that of the laser cutting department, an initial operation only supplied with direct workload. The average tardiness of this department is 0.5 days and the standard deviation of lateness is approximately 2.5 days. When moving on to the press brake department, that is usually later on in the production routing, the distribution is much more spread. In this department the average tardiness is approximately 1 day and the standard deviation of lateness is approximately 3.3 days. For the departments usually even further on in the routing, the

Figure 3.1.2: Throughput time distribution per department

Figur 3.2.1: Lateness distributions (Soepenberg et al, 2008)

(23)

Laser Sheets 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 Lateness distribution (working days)

% o f o rd e rs Press brake 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 Lateness distribution (working days)

% o f o rd e rs Customer 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14

lateness distribution (working days)

%

o

rd

e

rs

tardiness becomes even greater. This gives an indication that the delivery reliability becomes more difficult to control with increasing routing lengths. Eventually, when orders are sent to the customer, at the dispatching department, variety is much smaller. A standard deviation of 2.4 days and average tardiness of 0.7 days are measured. This could indicate a time buffer at dispatching, to ensure a timely delivery.

The distribution of lateness of the press brake department was investigated in more detail by plotting individual orders in an order progress diagram (Figure 3.2.3). The lines run from the lateness of the starting date to the lateness of finishing date of an order at the press brake department. The figure shows a number of orders that arrived ahead of planning and where delivered even further ahead of planning, while other orders arrived late and were delivered with an even greater delay.

A number of such orders may indicate poor priority dispatching (Soepenberg, 2010). The orders in graph 3.2.3 are deliberately chosen because they indicate poor description and do not prove that there is a structural problem. A further qualitative investigation, by interviews, showed that dispatching at the press brake department is largely based on whatever lies at reach.

3.3 Work in Process

A minimum amount of WIP on the shop floor will result in smaller queues in front of workstations. When an order’s queuing time is shortened by limiting the amount of WIP, the overall throughput time of an order will also decrease. The WLC concept uses this fact when releasing orders to the factory floor. The goal of the WLC experiments is to firstly keep WIP constant and then gradually bring WIP levels down to an optimum. WIP levels are within WLC the controlled variable, based on which new workload is released.

Figure 3.2.2 Lateness distributions

Order Throughput Press brake

-8 -6 -4 -2 0 2 4 6 8 29-09-10 4-10-10 9-10-10 14-10-10 19-10-10 24-10-10 29-10-10 date la te n e s s ( d a y s ) 10435268 10436496 10435215 10436245 10436176 10435588 10434103 10434095 10435651 10436183

(24)

Deburring 0 5 10 15 20 25 30 35

13-nov 18-nov 23-nov 28-nov 3-dec 8-dec 13-dec

Date D ir e c t W o rk lo a d ( h o u rs p ro c e s s in g t im e ) Press brake 0 20 40 60 80 100 120 140 160 180 200

13-nov 18-nov 23-nov 28-nov 3-dec 8-dec 13-dec

Date D ir e c t W o rk lo a d (h o u rs p ro c e s s in g t im e )

workloads for the press brake department and deburring, measured in processing hours required on the workstation. The measurements presented show that WIP levels in the current situation fluctuate heavily. The throughput time of an order at a station varies as a consequence of these fluctuations. In a situation with low WIP levels orders will spend less time waiting in queue before processing, then with high levels of WIP. An accurate prediction of the throughput time, for the setting of a promised delivery date, becomes impossible. In order to maintain delivery reliability, a large time buffer is necessary in setting the delivery date, in this way increasing lead time. A maximum direct workload at the press brake department of 176 processing hours was measured. With the capacity at the time being approximately 260 processing hours a week, this stands for a workload of more than three days. With this workload level the maximum throughput time of three days currently used, is hard to achieve.

At the laser cutting department, an initial operation, the direct load was not measured. Because jobs are immediately released to the shop floor, the queue of jobs in front of this station can not be defined accurately. At the welding department another difficulty occurs. A welding order mostly consists of the assembly of a number of different orders. The orders that are completely available for assembly can be expressed in amount of processing time required at the station. For the orders not yet arrived in fully this is problematic. In the baseline measure only the direct load that was complete for processing was measured.

3.4 Utilisation/throughput

Utilisation is defined as the probability that a station is busy. Maximum utilisation is, as stated before, only a performance objective when the workstation under consideration is a bottleneck station. All extra throughput realized at a bottleneck station contributes to the overall throughput. The high variability of orders and processing times at Meijer, results in moving bottlenecks depending on the mix of workload situated on the shop floor.

If a station’s utilisation is increased without making any other changes, average WIP and throughput time will increase in a highly nonlinear fashion (Hopp and Spearman, 2000). The probability that an order has to wait for completion of preceding orders increases. Shorter queues in front of workstations result in shorter SFTT, but setting the queues too short will result in a decrease in utilisation caused by starving of machines. Starving relates to the absence of input for production, with the station standing idle as a result. The variability in processing times and routing of jobs can result in a low utilisation of certain workstations when workloads are not adequately balanced. The workloads at Meijer show great variations over time, the capacity is adjusted accordingly and is for that reason also highly variable. Table 3.4.1 shows the throughput levels, recorded by time spend on jobs as registered by workers. The throughput levels are compared to the planned capacity of five different capacity groups, for a period of four weeks. The capacity adjustments are made by shifting workers between stations and by hiring temporary workers. The actual processed times in table 3.4.1 sometimes exceed planned time. Possible explanations for the rates above 100% are overtime, extra

(25)

non planned help from other departments or differences in calculated machine time and actual time. The amount of throughput a workstation can realise in a release period is of great importance when setting the height of the workload norms.

week 46 week 47 week 48 week 49

Cutting 82% 106% 78% 54%

Processed time on jobs (hours) 146,82 148,54 108,89 67,08

planned capacity 180 140 140 125

Press brake 102% 123% 68% 83%

Drilling 87% 86% 33% 162%

Deburring 113% 89% 41% 54%

Welding 102% 88% 70% 88%

Table 3.4.1 Utilisation/ throughput

3.5 In addition

In addition to all the quantitative measures represented before, the shop floor is also monitored for observations and frustrations. The shop floor supervisors and operators are interviewed about experiences in practice. This section gives a number of these observations from before the start of implementation.

“The large amount of rush orders have a negative effect on the performance of normal order production”. Additionally, the amount of rush orders increased to a level, that almost every order was a rush order and it lost its credibility.

Last minute additions to the planning cause a stressful situation for the nesting programmers. “Last minute additions to the planning result in the loss of a clear overview”. The programmers experience this as disruptive. “This results in extra searching for orders and occasionally loosing job cards”.

WIP levels before the press brake department occasionally rise very high. As a result the storage racks flood and orders are stored elsewhere in the factory, wherever room is available.

At the assembly department, the supervisor appoints that it is very difficult to meet their delivery reliability objectives, mainly because the assembly parts are delivered late.

(26)

3.6 Résumé

(27)

4. Changing the nesting period

4.1 Planning action

The first step that is noticeable on the factory floor is concerned with initiation of production, in particular the nesting procedure that precedes the laser cutting of sheets. For the greatest part of the production mix this is the initial step of production. The main reason why the nesting period is of interest to the implementation of WLC is because it concerns initiation of production. A number of reasons, for changing of the nesting procedure arise. The dynamics of practice are the first reason. The system under consideration is dynamic and not every adjustment is initiated by the researcher. The desire to change the nesting period, already present within the company, originated from a frustration of frequent last minute adjustments to the planning. Although not planned as such, the researcher grasps the opportunity to align the actions with the workload control plans.

Changing the procedure facilitates order release, a fundamental part of WLC as mentioned in the initial design of appendix 1. In the initial design, it becomes apparent that a release period is ideally set as short as possible. By changing the nesting period, the release of workload can be synchronised with the nesting period. In this way, every two days, jobs can be released shortly before every new nesting period.

Thirdly there is the desired production sequence. In the desired future state, a released order relies on a steady factory flow. For realizing all intended delivery due dates in this system, orders have to be produced in the right sequence. The shop floor production sequence is controlled by the shop floor dispatching rules. Implicitly, a number of dispatching rules are already in use on the factory floor. Most of these implicit rules are based on gut feeling and experience of the shop floor supervisors, like for example limiting the amount of set ups. The dispatching rule currently used for the sheet cutting is of special interest with respect to the implementation of WLC. As stated before, generally the cutting is the initial operation. The priority dispatching at this capacity group and thus the output sequence is a result of a nesting procedure. In section 1 the nesting procedure was already shortly discussed. Currently, a weekly planning is released for the forming of nests, consisting of two blocks of clustered jobs. The first block of jobs has to be ready at the cutting department on Wednesday and the second block on Friday. The sheets that are processed contain numerous different orders with different due dates. A nest can contain from two up to forty orders. When the orders are finished cutting they are stored in racks, waiting to be manually separated from the sheets. At separation, the different orders are collected from the sheets and provided with a card containing processing information, before they are transported to the next operation.

Jobs with scheduled start on Friday are currently possibly already processed on Wednesday and vice versa, because they are part of a nesting. This means that with the nesting, the production sequence of three days’ worth of workload is re-planned. The initially planned processing sequence based on the due date of a job, is altered to a processing sequence based on material efficiency and setup reduction.

(28)

A new period of two days for nesting, is expected to provide the possibility for job release every two days. Additionally the shop floor frustration mentioned in section 3.5, is expected to be eliminated

“Last minute additions to the planning result in the loss of a clear overview”. “This results in extra searching for orders and occasionally loosing job cards”

Finally a shorter nesting horizon is expected to increase delivery reliability, resulting in a decrease of lateness.

4.2 Developments

Although the main goal of this implementation step was to align the nesting period with a possible future release period, in the period after implementation effects became visible on the shop floor. After a month of production with the new nesting procedure reactions were positive on the factory floor.

The supervisors of the departments supplied by the laser cutting department noticed more timely deliveries of supplies. Because more jobs arrived on time it became easier for these departments to deliver on time. In the period before implementation, delivery performance of the departments was frequently undermined by a lack of timely supply. As a result of this, motivation to deliver on time also decreased. Because of the newly established timely supplies, supervisors regained control over the delivery performance of their department. This increased motivation. The qualitative information was enriched by measurements from the performance monitor. The rest of this section describes the measurements that are of special interest to the evaluation of this implementation step.

Lateness

The lateness measure gives an overview of the discrepancy between when an order was planned to be ready and when it actually was ready. A decrease of the variance of lateness would indicate that products are processed closer to their planned production time.

When the lateness distributions of January were compared to those of the baseline measures, it showed a sharp decrease in the deviation of lateness and tardiness for all stations. This would indicate that the implementation step satisfied its purpose. Further enquiry revealed that the decline was already initiated before the start of implementation. Table 4.2.1 shows the standard deviation of lateness and the average tardiness for three workstations. A drop of average tardiness over a four month period becomes visible, that is most likely a result of a decreasing order intake. Nevertheless, after the implementation start in January, the reliability performance continued to increase.

The decrease in total shop load, initiated in November, resulted in a smaller workload on the shop floor and a better ability to keep up with the supply of workload. Figure 4.2.1 shows the direct workload at the press brake department. The descending line is illustrative for the decreasing shop loads. The delivery performance increase, as a result of the decreasing input, indicates that input control provides delivery performance gains.

Table 4.2.1 Standard deviation of lateness & Average Tardiness

Laser cutting (sheets) Press brakes Welding

St dev of Lateness Tardiness St dev of Lateness Tardiness St dev of Lateness Tardiness

October 1.59 0.50 3.31 0.98 3.85 1.62

November 1.42 0.33 3.03 0.75 2.61 1.41

December 1.34 0.30 2.28 0.32 2.11 1.10

(29)

Figure 4.2.2 shows the development of lateness distribution for the initial and the assembly operation over a period of three months. It shows that the effects are most prominent at the welding department. Welding, often the last station in the routing and a convergent line, is most dependent of timely supplies for its performance. Keeping the two approaches for reducing lateness (figure 3.2.1) in mind, it becomes clear that gains were mainly a result of average lateness reduction instead of variance reduction.

Throughput time

Figure 4.2.3 shows the distributions of the throughput times at three departments for January. The throughput time performance did not reflect the positive note of the lateness performance. At the drilling and deburring workstations average throughput time increased from respectively 3.9 to 4.7 days and from 2.3 to 2.9 days.

Remarkably these increases in throughput time did not deteriorate the delivery reliability of the stations. This is possibly explained by the supply of orders to these stations. Order supply was largely ahead of planned start date. When studying the lateness distribution at the laser sheet cutting department it shows that 71% of the orders were finished early. This enables the jobs to have larger throughput times at other stations, while still meeting the intended due date.

November Laser Sheets

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 Lateness distribution (working days)

% o f o rd e rs

December Laser Sheets

% o f o rd e rs

January Laser Sheets

% o f o rd e rs November Welding 0% 5% 10% 15% 20% 25% 30% 35% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 Lateness distribution (working days)

% o f o rd e rs December Welding 0% 5% 10% 15% 20% 25% 30% 35% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 Lateness distibution (working days)

% o f o rd e rs January Welding 0% 5% 10% 15% 20% 25% 30% 35% -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 Lateness distribution (working hours)

% o f o rd e rs

Figure 4.2.2 Lateness Laser Sheet Cutting and Welding

(30)

Throughput times Short routing 0% 5% 10% 15% 20% 25% 30% 35% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Working days % o rd e rs 1 step 2 steps 3 steps

The SFTT gives an overview of the overall factory throughput time performance of jobs, not restricted to a single workstation. Figure 4.2.4 also shows that for throughput time no improvements are evident. Throughput time gains at one station are apparently levelled out at subsequent stations. A production sequence based on the “planned operation due date” at the shop floor could be responsible for this. Based on their planned due date, orders are processed and short term capacity adjustments are made. When orders arrive at a station ahead of schedule, they are left waiting until they approach their intended due date.

4.3 Résumé

This section describes a change of the nesting period that precedes the laser cutting operation. The main reason for this implementation step was that it provides the possibility of releasing jobs every two days. Secondly it was expected to improve the production sequence, which should result in a lower standard deviation of lateness and an increase in delivery reliability of the laser cutting operation. Although lateness was reduced, this was likely the result of a decreasing order input. The shop floor frustration of last minute changes to the nesting was eliminated by making clear agreements about deadlines for order planning. Persistence of large station throughput times motivates additional implementation efforts described next.

1 step Mean: 4.48 days Stdev: 4.08 days CV :0.91 2 steps Mean: 7.23 days StDev: 4.6 days CV : 0.64 3 steps Mean: 9.34 days StDev : 4.7 days CV : 0.51

Figure 4.2.4 Shop Floor Throughput Time Figure 4.2.3Throughput times January

Deburring 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

% o rd e rs Press brake 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

% o rd e rs Drilling 0% 10% 20% 30% 40% 50% 60% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Troughput time (working days)

%

o

rd

e

(31)

5. Introducing the release decision

Now that the initial production step is altered, implementation of a fundamental WLC concept can be initiated. The performance measures after step 1 show that shop floor throughput times are longer than anticipated in the planning. There is a strong relationship between WIP levels and lead times and this relationship plays an important role within workload control concepts. (Land, 2004) In theory, lowering the level of WIP should increase flow. Shorter queues in front of workstations should reduce the waiting-time component of total throughput times. Based on this, the goal of the next step is composed. Figure 5.1 visualises this goal. In section 3.3 the levels of direct workload for workstations where presented in a graph. The left graph in fig 5.1 shows one of those graphs, with a fluctuating direct workload level over a period of 20 days. The goal of the next implementation step is to control these levels. The middle graph shows a hypothetical road to controlling the length of the queue, firstly taking of the sharp edges. The graph on the right represents the desired state, with constant and low workload levels.

In order to reach this goal, a WLC concept at the job release level is established. The introduction of an order pool and a release decision should create the possibility to control the input of workload to the shop floor. More details and theoretical background are provided in appendix1. By controlling the input it is expected that large peaks of direct workload as visualised in the left graph can be prevented. And supply of workload can be smoothened. For testing this in practice, substantial steps towards WLC implementation will be taken and support from the organization is required. For this reason a more extensive action planning is necessary. A basic release system is constructed and described next.

5.1 The introduction

Before the start of the test, the plans are communicated to the employees concerned in order to gain support and reinforcement for the project.

Firstly the management team is informed with the pilot plans. After their approval the plans can be communicated to the rest of the organisation. All the employees concerned with planning function, are introduced to the plan and asked for support. These employees occur at several positions in the organisation. Besides the basic production planning at a central level, the shop floor supervisors are accountable for the planning and control at shop floor level. Their support is necessary for successful implementation. The supervisors are expected to give feedback and evaluate during and after the tests, this qualitative information is used to support the quantitative information gained from the performance measures. In a session with the planner, the shop floor supervisors and the production leader, basic theory, motivation and the pilot plan are presented.

Variables Current situation 0 20 40 60 80 100 120 140 160 180 200 0 5 10 15 20 Date D ir e c t W o rk lo a d ( p ro c e s s in g h o u rs ) Experiment 0 20 40 60 80 100 120 140 160 180 200 0 5 10 15 20 Date D ir e c t W o rk lo a d ( p ro c e s s in g h o u rs ) Desired state 0 20 40 60 80 100 120 140 160 180 200 0 5 10 15 20 Date D ir e c t W o rk lo a d ( p ro c e s s in g h o u rs )

(32)

5.2 Release and ICT Support

To obtain control over job release, a release list is necessary. This section presents the design of this list. Currently planning is done in an ERP system called ISAH. For the implementation, some alterations are necessary. The choice of ICT support is done with the following in mind:

“Software developers have concentrated on advanced planning and scheduling methods, while Muda and Hendry [2002] conclude after a number of case studies that small make-to-order companies do prefer simple solutions, not requiring more than spreadsheet support. In this respect the WLC concept still fulfils the needs.” (Land, 2004)

Fig 5.2.1 shows the release list created in the ERP system. Jobs are no longer immediately released to the shop floor. To make a distinction between released and unreleased orders, an order status is assigned. New orders automatically get a status 10 and appear on the release list, but not yet on the shop floor planning.

The release list is arranged by placing the orders with highest priority for release on top of the list. The first priority sequence is that of earliest planned start date of the initial operation, depicted in the far left column. Composition orders, orders that are an assembly of a collection of other orders, are an exception to this rule. Figure 5.2.2 represents an example of the production progress of such a composition order. For the start date of these composition orders, the start of the assembly part with the latest starting date is taken. This ensures that assembly orders appear on the shop floor planning in time.