Achieving higher delivery reliability at Ramix/Romit B.V.

(1)

1

Achieving higher delivery reliability at

Ramix/Romit B.V.

"The mix for your success"

Applying the WLC concept

Master Thesis

Author: Ronald Vinke Student number: S1900188

E-mail: ronaldvinke@hotmail.com

Phone: 06 29 32 5993 1st Supervisor: Erik Soepenberg 2nd Supervisor: Jan Riezebos

(2)

2

Introduction

Company Background

Ramix / Romit B.V. (hereafter referred to as RaRo), is a small to medium sized production company with approximately 65 employees. The company is located in Dedemsvaart, and this research will be performed at the production facility in Dedemsvaart. Ramix provides over 50 years of pipe fitting experience with the newest moulding technologies, delivering their quality products all over the world. Romit industrial supply was founded in 1955, and is nowadays part of the Rollepaal Holding, their roots lay in the plastics and moulds industry. Romit’s product range differs from industrial supply like milling, lathe turning and metal machining/fabrication, to delivering complete supplies towards clients in energy technology, maritime technology and rubber/plastics technology. Since 1996 Ramix B.V. is part of Romit B.V. and continued as one company called Ramix / Romit B.V.

The company produces and engineers a wide range of moulds for the plastics industry and is a partner in machining projects. Because of the wide variety of products that are tailored to customer specification, almost every product is engineered to order (ETO) or made to order (MTO). The main products are built for the plastic industry (moulds), and there are several other large customers, like Siemens, who asks for turbine parts and subassemblies. Customers of RaRo are most of the time production companies as well, like Wavin and Rollepaal, companies that produce plastic pipes and fittings. Next to the fact that the products RaRo delivers need to be of a high quality, delivery reliability is one of the most important factors. Fast deliveries (rush orders) are often scheduled when a machine breakdown occurs at one of RaRo clients. The routing of the production process is often complicated and differs per product, which makes scheduling of the orders also complicated. The sales department of RaRo makes sure that RaRo receives orders. Orders are coming in for three different product ranges and these orders can differ in quantity, but even more in size and complexity. The three product ranges are:

Moulds:

As explained above, these orders include the injection moulds for the plastic industry. RaRo also assembles, tests, transports and maintains these moulds for their customers.

Rollepaal:

These orders come in from the holding to which RaRo belongs. The orders that are coming in from Rollepaal contain parts of extruders (equipment to produce plastic pipes, see figure 1 on this page). These parts are mainly barrels, screws, breaker plates and inlets.

Jobbing:

This contains every order that is not included in the above two categories. Parts like cranks, shafts and others belong to this category. These are mostly orders for the maritime and energy technology. This includes also the large orders of, for example, Siemens.

(5)

5

Reason for Doing Research

Due to unforeseen changes in production schedules, the management of RaRo gets the feeling that the production flow is rather uncontrolled, which gives pressure on delivery dates but also results in a lower utilization rate for the machines due to switching times and longer set-up times. At the moment RaRo’s director believes the delivery reliability of RaRo is around 70 percent (meaning that 70 percent of all orders is delivered on time to customers). The software package (Limis Planner) that RaRo uses for their planning gives some insight on the short term planning, however the system does not show directly the causes of unforeseen production changes. The system is more applicable for the medium term planning (6-8 weeks).

The workers on the shop floor can see/create a “to do” list from the planning system. On this list the projects for the coming week are scheduled. Behind every order on the “to do” list several information about the orders is given, for example: the delivery date, how many process hours there are scheduled for the specific machine, but also if the previous process is still in progress or not. A red sign means the previous process didn’t start, a yellow sign means that the previous process is in progress and a green sign shows that the product is available for processing the next operation. Because several disturbances can occur during or between the process steps, management wants a more detailed short time production schedule. In other words: a more detailed production schedule for the shop floor.

Because almost every order differs, the requested products have (most of the times) to be drawn from scratch. This means RaRo is a company that constructs products completely based on customer specifications, better known as Engineer to Order (ETO). This complexity results in products that differ in the amount of product steps; there are for example some products that have over 30 process steps that need to be taken. One can imagine that it is quite tough to make a correct pre calculation in these cases. Because RaRo already recognized that problems could arise somewhere in the planning phase, the production planner created a list of problems that are likely to occur, these are:

 Rush orders

 Repair assignments

 Outsourced parts not on time

 Extra work due to rejection

 Orders planned on wrong machine

 Machine breakdown

 Wrong pre-calculation

 Illness

 Tools not available

 Drawings not available

 Diverge from routing

 ERP System performance

(6)

6

jobs as well as input and output control (Hendry & Stevenson, 2005). Rush orders, repair assignments, outsourced parts not on time, order planned on wrong machine and diverge from routing are the most closely related to these PPC decisions and will therefore, if possible, included in this research as further explained in the limitations part (page 7) and conceptual model (page 10). The outcome of this research might be influenced by the other factors that not directly relate to planning, like a machine breakdown, illness or ERP system performance. Although not being measured or taken into consideration, this might still be important factors that might result in not meeting the required utilization rates and/or delivery dates.

Pre Research

Management of RaRo has the feeling that the order flow is rather uncontrolled because of unforeseen changes in production schedules. These changes are time consuming and lead to inefficiencies in production, which leads according to the management to a delivery reliability around 70 percent. If for example extra work due to rejection occurs at the beginning of an order, the other process steps will get delayed as well, which gives pressure on the delivery due dates. PPC systems are crucial for meeting high customer expectations in the competitive manufacturing market (Hendry & Stevenson, 2005). The management of RaRo mentioned that there is a lack of detailed planning for the shop floor to overcome this problem. By creating a more detailed planning system, the employees should have more insight in the consequences of changes in the production schedules. The management of RaRo believes that by giving the workforce more responsibility, they will communicate more, which will hopefully lead to early awareness and solving of disturbances. There is however not a clear picture how this can be achieved and moreover, there is not much information available over the exact problem. Therefore some pre research should be done to get a clear picture of the current problem with respect to delivery reliability. In other words, the current delivery reliability needs to be measured.

The objective of the pre research phase is to gain insight in the problem and to achieve a solid base for further research. Through analyzing the planning system, the delivery dates could be analyzed. All the orders since the

beginning of May 2011 have been analyzed (278 orders in total). There were in total 22 moulds related orders, 14 main extruder parts orders, 166 other extruder parts and 76 jobbing orders. The table below shows the percentages of orders that were delivered on time. Each order receives an order number; the first 3 numbers indicate the type of order, as shown in table 1. Most of the orders are 910 and 917 orders, and have a delivery reliability around 60 and 65 percent, see also Appendix B. The feeling the management has about a relatively poor delivery reliability of 70 percent is therefore

(7)

7

The reason that the 905 order (moulds) have such low delivery reliability is the amount of re-cut / fine tuning that needs to be done after delivery to the customer. After delivery to the customer the order will not be closed in the planning system, because it is due to financial and planning reasons easier to keep the order open in the system. The planning reason is that it is quite complex to get an order back in the system after it is closed, and financial reason is for assigning the hours to the correct order. The main extruder parts (909 orders) are mainly ovens and screws of Rollepaal, which are inventory based items with a due date in the far future. To be able to absorb fluctuations in demand, these parts are processed in times when there are not a lot of orders; this is the reason that this category has a delivery reliability of 100 percent. The main part of the total amount of orders contains of 910 and 917 orders, which are requests of customers (Jobbing) or Rollepaal (other extruder parts). In these last categories it is also more likely that rush orders come in, for example when a customer calls with the message that a machine/extruder has broken down and they need a sub- or spare part. The main part of the orders will be investigated in this research. In section 3 will be further elaborated on delivery reliability regarding the product categories.

1 Problem Statement

The incoming orders are accepted by the sales department and this department also promises the delivery dates to the customers, often based on a quoting process which is highly uncertain. It is determined if a quote/offer has a high or low probability to become an order. If the probability is high, in some cases (for example Siemens) the order will already be preliminary booked in the production planning. The sales department communicates often with the engineering and planning departments, and they have a team meeting every week to discuss progress of orders. Nevertheless, a lot of rush orders are accepted to keep customers satisfied. This gives pressure on the delivery reliability, which the delivery reliability rate of approximately 65 percent shows. Achieving high delivery reliability is a combination of controlling average throughput times and controlling the progress of individual orders as mentioned by Land (2004). The objective of this research will be focused on diagnosing the company and give advice for improvements in planning and production decisions to achieve higher delivery reliability. The research objective of this research could therefore be described as:

To advice RaRo in which way they can improve their delivery reliability, by giving insight in the problems that cause disturbances in the order flow.

The research question can be derived from the research objective and focuses on several elements that influence the order flow. This process starts already when an order is accepted by the sales department and ends when it will be transported to the customer. The research question belonging to this investigation will be:

How can RaRo improve their planning and control system to be able to meet due dates? (I.e. improving delivery reliability to at least 85 %)

Limiting Conditions

(8)

8

validation of these motivations, together with several factors that influence the performance of the delivery reliability, like for example the order flow and the planning and control decisions will be reviewed and thirdly, it will be examined how this is influencing the delivery performance of the company. The paper will therefore not focus on implementing these changes and only give guidance on how improvements can be achieved. Therefore this paper can be seen as the diagnostic phase of the issue regarding delivery reliability. Theory will be reviewed, problems will be investigated and last but not least the difference between theory and reality will be shown which will result in several recommendations for improvement.

2 Research Design

Theoretical Background

RaRo produces several different products, but mostly moulds for the plastics industry and industrial supply to customer request which is mostly milling and metal fabrication. A (injection) mould is complex equipment; they differ in size, specifications, connection and differ therefore in amount and kind of parts and subassemblies. Because these moulds differ so much, every time new detailed drawing needs to be made according to customer specification. This is also the case for milling and metal machining orders, most of the times these operations need to be performed at customer specification. Marucheck and McCLelland (1986) considered manufacturing extending from pure engineer to order (ETO) on the one hand to pure make to stock (MTS) on the other. The increasing need for companies to deliver customized products has increased the number of companies that operate in the make- or engineer to order (MTO/ETO) base (Hicks & Braiden, 2000). Because RaRo produces products according to customer whishes, they operate in an Engineer to Order environment. As mentioned by Gelders, F. (1981) “in an Engineer to Order environment, a company designs and produces products to customer order”. Typical examples of such kind of companies can be found in the mechanical engineering domain where an order is partly or completely tailor made. This is also the case for RaRo. ETO products may require a unique set of item numbers, bills of material, and routings. These products are typically complex with long lead times (Brian & Watson, 2010). Unfortunately, literature has mainly focused on make to stock environments as mentioned by (Hendry and Kingsman, 1989). There are however some products RaRo produces which can be categorized in the MTO environment. In an MTO environment, stocking happens at the level of raw materials, or direct purchasing of materials from suppliers after receipt of a customer order according to Schönsleben (2007). Orders that come in from Siemens, but also moulds can be categorized at MTO level. In this case the final product is produced using materials from the raw materials store or acquired through customer procurement orders. Problems in planning arise due to a demand that can highly fluctuate, with each unit of demand representing a large proportion of the design and manufacturing capacity. Besides that, there usually is also a service and spare parts business associated with the major products in MTO companies. (Hicks and Braiden, 2000).

(9)

9

ETO/MTO company to be able to identify the status of specific components and orders that are required by particular customers in order to meet delivery dates. Meeting delivery dates is extremely important nowadays for MTO/ETO companies to remain competitive. Next to quality, delivery reliability is one of the order winning factors. To achieve a high delivery reliability, the variance as well as the average lateness needs to be controlled. By relating these two aspects to production planning and control (PPC) decisions, the Work Load Concept (WLC) is a helpful diagnosing tool as mentioned by Soepenberg et al. (2007). According to Stevenson et al. (2005) the shop floor layout is a major factor in the applicability of production planning and control. Four configurations are considered, namely: the pure flow shop, general flow shop, general job shop and the pure job shop. In a pure flow shop, the materials visit work centers in a strict order. In a pure job shop the routing sequence is random. Jobs can start and finish at any work center allowing complete freedom for customization. However, researchers as Oosterman et al. (2000) doubt that a pure job shop can be found, because in reality a dominant flow can be found. Because RaRo has three different product categories and produces numerous different products, a common flow is hard to find and has therefore characteristics of a pure job shop.

Measures of Performance

Numerous studies focused more on planning decisions in relation with MTS or ATO manufacturing companies. Little attention was given to pure job shop environments and MTO companies. However over the last decades customers request more and more for tailored and customer specific products, which lead to an increase of MTO companies. These companies are most of the times small to medium sized enterprises, which have limited financial resources (Hendry and Kingsman, 1989).

Delivery reliability is one of the order winning performance criteria in job shop environments, according to Land (2004). However, planning of those kinds of products is characterized by uncertain operation durations, finite capacity resources and multilevel product structures (Earl et al., 2003). Acquiring high delivery reliability can be achieved by controlling the average lateness and variance of lateness of orders as mentioned by Soepenberg et al. (2007). These two aspects are influenced by production planning and control decisions (PPC). According to Kingsman & Hendry (2002) as well as Soepenberg et al. (2007) PPC decisions relates to input control (job entry, job release and priority dispatching) or output control (adjusting capacities). Because this study focuses on the diagnostic phase of PPC at RaRo, the before mentioned factors will be used to investigate the current situation with respect to input and output decisions.

(10)

10 Conceptual Model Explanation of Factors

The definitions of the conceptual model (figure 2) are explained beneath, to make them measurable. All the factors can positively or negatively influence delivery reliability. However, by adjusting / improving each of the factors, the delivery reliability should be positively influenced. Therefore, the assumption is made that all the factors that influence the order flow have a positive influence on the delivery reliability.

Delivery reliability

Delivery reliability is a common performance goal (Stoop & Wiers, 1996) and achieving high delivery reliability gains a competitive advantage, as assigning exact due dates and delivering the goods on time to the customers will enhance customer satisfaction (Hendry, 2010). According to Soepenberg et al. (2007) this factor is especially important for MTO companies, because it is one of the order winning performance criteria. Performance measures of delivery due dates can be split into earliness and tardiness. Earliness means that an order is finished to early and will result in a higher level of finished goods inventory. Tardiness of an order results in an order being delivered to late. Economic implications of tardiness are often related to penalties for not delivering on time and losing customer goodwill (Alpay, 2009).

Job entry

Creating an appropriate way to select orders for release is the key to the successful use of an order pool. However, without job entry control, power will have been lost before an order reaches the job release or priority dispatching stages, which results in orders staying too long in the pre-shop pool (Tatsiopoulos, 1997). The customer enquiry stage is very important to an MTO company, although when customer enquiries cannot be controlled; by stabilizing lead times and relating workload to capacities, in using effective PPC, the company can make delivery date quotations to customers with a greater reliability (Stevenson & Hendry, 2005).

Job release

Because capacity is often finite, it is important to select orders for release that provide capacity groups in the shop with a balanced load. This will support the control of the average lateness (Land, 2004). Load balancing results in smooth flow on the shop floor and avoids

Job entry Job release Capacity Delivery reliability

+

Priority dispatching

(11)

11

long queues in front of certain capacity groups. Kingsman and Hendry (2002) state that controlled job release means maintaining a pool of unreleased jobs at the planning office. Jobs will only be released if it will not cause the planned queues. This will result in a work in progress (WIP) reduction. The queue of jobs in the pool also need to be controlled, otherwise orders might stay in the pool too long and miss the delivery date.

Priority dispatching

The day-to-day shop floor control level is called the priority dispatching level. Of the three input factors, priority dispatching received most attention, but it is a relative weak level if used alone (Kingsman and Hendry, 2002). Because the workforce has to fulfill the day-to-day operations and because they also have a certain impact on the day-to-day planning of orders, priority dispatching is of high importance in this research. Delivery reliability can be improved by using dispatching rules; nevertheless, this will have a limited influence on the average and variance lateness (Soepenberg et al., 2007).

Capacity

Output control decisions can adjust capacity to those machine groups where orders are queuing. Capacity changes are generally triggered by large sets of orders that have a high possibility to be delivered late. Output control focuses therefore on the average lateness of orders (Soepenberg et al., 2007). Resource utilization is an important performance measure that is of particular managerial interest to ETO/MTO manufacturers (Weng, 1998). Capacity adjustments can be made according to medium, short or daily capacity adjustments. For job shops these capacity adjustments are in terms of outsourcing, allocating overtime, assigning operators to machines.

Besides capacity there are other factors that influence the performance of delivery dates. Several problems have been discovered by the planning department of RaRo, and of those eleven problems, factors like tooling or outsourced parts not available do not relate to capacity and will therefore not be taken into consideration.

The above mentioned input and output control decisions can be visualized in the Work Load Control (WLC) framework that has been introduced by Land & Gaalman (1996), see figure 3 below:

(12)

12

Sub Questions

The sub questions can be determined from factors in the conceptual model. The sub questions should give an answer on the subparts that ultimately will answer the research question how the delivery reliability can be improved. The following sub questions have been defined:

 How are orders accepted?

 How are orders released?

 How does priority dispatching influence delivery reliability?

 How are capacity adjustments made?

The sub questions relate directly to the factors used in the conceptual model and are raised to gain insight in the way how certain processes function at RaRo, and how this mentality of working affects the delivery reliability. These are common factors that affect delivery reliability in job shop environments and are stated here because of their importance in meeting delivery dates. Note here that meeting the delivery date means on time delivery. An order delivered to early has the same poor delivery reliability as an order delivered late. Because most of the products RaRo produces are long lasting and according to customer specification, changes to the product, delays in outsourced parts or purchased materials as well as several other unforeseen changes can occur. It can in these cases happen that the delivery date will be rescheduled. Always the agreed delivery date with the customer will be taken as the delivery date.

Data Gathering Methods

The aim of this research is to explain and understand the why and how of the decision making process in order to meet delivery dates, therefore a diagnostic research should be performed. To find out how processes work and why certain phenomena happen, a qualitative research is appropriate (Thomas, 2004). According to Blumberg et al. (2005) quantitative studies rely on quantitative information like for example numbers and figures, whereas qualitative studies base their assumptions on qualitative information, such as words and sentences. Besides quantitative research, qualitative research needs to be done. Figures about delivery reliability, waiting time and throughput times for example are necessary measurements to supplement the qualitative data. A combination of qualitative and quantitative research is preferable for several reasons: first of all, it allows gaining practical knowledge in addition to theoretical knowledge. Secondly, it provides in-depth information about specific characteristics on planning methods, which can be approved by figures. And thirdly, it provides a first insight for empirical research concerning the improvement plan that might arise from this research. The following information gathering methods will be used:

Desk Research

(13)

13

Field Research

The field research will be performed at the company RaRo. The delivery reliability problem will be analyzed from a PPC point of view at this company. The WLC concept is a helpful tool in analyzing the delivery performance of RaRo, because the WLC concept aims at diagnosing and improving delivery reliability in job shop environments. Most of the MTO/ETO factories are small to medium sized enterprises and RaRo is also part of this category. The WLC concept is especially suitable these kind of companies. The strength of this research is the practical use of this theoretical model. When executing the field research at RaRo, collecting new information through interviews and observations takes place. The employees of RaRo need to be interviewed. It is definitely necessary to interview the people who operate on the work floor, to gain insight in their needs. After all, the workforce needs to have on time and correct information about the planning of orders, but they also need to use this information. By interviewing the workforce, shortcomings in the current information needs might be discovered. Moreover, by interviewing the workforce the current method of working can be analyzed. Other departments like planning, work preparation and sales need to be interviewed as well, because they have a direct influence on several aspects of delivery reliability. Blumberg (2005) mentions: using personal interviews within a qualitative research is a good way to derive examples and background information about managerial interpretations on decisions leading to success or failure. The importance of discussions with key informants can nevertheless, give rise to the threat of the researcher becoming too dependent on them (Blumberg, 2005). Relying too much on just a few key informants can jeopardize the validity of a study if the informants present a biased picture of the case issue. Observations need to be done to measure and control the order flow (the quantitative part). The expert insights and theoretical findings can be compared with the observations to find shortcomings in meeting delivery dates. In other words, the qualitative part should give insight in where problems arise and the quantitative part shows why those problems occur. Thus, clearance on the problems that arise in delivery reliability can be shown by facts from analyzed data and the interviews. A limitation in the quantitative data part is the time period that will be taken for analyzing (the progress of) certain orders. The researcher will take relevant data around the period that the research includes, namely the orders of May till September 2011. The vacation in between this period might bias the progress and delivery date performance of a few orders. A longer analyzing period can create a more complete/correct view as well.

(14)

14

machine to point out. Most of the times this was the sawing machine, but this was not always the case. Besides that, the sawing machine doesn’t use the scanning tool that Limis uses. To figure out the release date for all the orders would therefore be almost impossible. The qualitative data (interviews) will however support the quantitative data in terms of opinions, experiences and findings from the employees of RaRo.

Data Analysis

Analyzing the sub questions will be done by using qualitative data in the first place. Like interviews with the sales managers, planning department and workforce, but also by analyzing the procedures with respect to the job entry, job release and priority dispatching phases to gain insight in the current situation of how procedures work and capacity adjustments are made. Furthermore, the way the workforce itself thinks delivery reliability can be improved will hopefully be the result of the interviews.

Next to collect this qualitative data, the quantitative data showed already that the delivery performance is around 65 – 70 percent, and resulted in a confirmation of the feeling that management had. Other quantitative data, like for instance a deeper analysis on delivery performance figures and throughput time diagrams give more insight in the way decisions and/or machines had on the delivery performance.

(15)

15

3 Delivery Reliability and the WLC Concept

After redefining the ultimate goal for this thesis from detailed planning for the shop floor to improving delivery reliability, the first analysis about the company characteristics already had been made. RaRo is a company that, as said before, produces to customer specification. Almost every order differs and some products contain a lot of process steps and different routings, where others only have a few. This means that RaRo can be categorized as a complex job shop, a company where a large number of different products are produced according to customer specification. Another characteristic of complex job shops are highly variable routings and processing times (Land, 2004).

Research on job shop control has been performed from different perspectives and delivery reliability is one of these topics. Static scheduling perspectives result in a certain set of jobs to be available; however the uncertainty regarding the routings and the processing times of job shop products, scheduling exact start and stop times for operations of an order becomes unrealistic. Such a problem is called a NP-hard problem, requiring long computing times for only a reasonable problem size. New orders together with the uncertainty and disturbances that can occur on the shop floor create instability in schedules and require a lot of feedback from the shop floor.

In most job shops, queues arise before each workstation to act as a buffer in order to keep machines utilized. Control of these queues is one of the main principles of the WLC theory (Land, 2004). According to Soepenberg et al. (2007) competitiveness of job shop environments imposes needs upon delivery reliability. Both the average lateness and the variance of lateness need to be controlled in order to achieve high delivery reliability. In the pre-face of this research, it turned out that the current delivery reliability of RaRo is between 65 and 70 percent. To increase the delivery reliability, a deeper analysis should be made and should be linked to Production Planning and Control (PPC) decisions. Soepenberg et al. (2007) mentioned that PPC decisions can relate to input control (order acceptance, order release and priority dispatching) and output control (adjusting capacities).

Input and output decisions can both influence average lateness and the variance of lateness. Lateness can be divided in to subclasses: positive (orders delivered late) and negative lateness (orders delivered early). Many people have in mind that an order delivered to early is not a problem. Nevertheless, an order delivered early can’t compensate for an order delivered late, which means that an order delivered 10 days early can’t compensate for an order delivered 10 days late.

(16)

16

To show the effects of controlling average lateness and the variance of lateness, I will refer to Soepenberg et al. (2007), see figure 4 above.

After making a histogram of the orders RaRo processed and delivered in the period between May and September 2011 (see Appendix B for the data set), it can indeed be said that there is a certain combination of average and variance of lateness, as shown below in figure 5. There is a certain amount of orders that is between 1 day and a week to late, and if throughput would have speeded up, these orders would have been on time. There is also a certain amount of orders that doesn’t even come close to the pre-defined delivery date.

(17)

17

To show variances between product groups, histograms have been made of the 3 different product groups, see appendix C. The moulds category (905 orders) shows a relative high percentage of orders being delivered late, the reason that moulds are often delivered late, is due to the fact that they need to be installed at the customer and that fine-tuning and testing needs to be done. This means that although the order is delivered, the order will not be closed in the system yet, and therefore the system “thinks” the delivery due date is crossed. The Rollepaal inventory orders (909 orders) show the opposite figure of the moulds category; these orders have delivery dates in the far future. When there are for example not a lot of orders in a certain time, these orders are being processed. Both categories (905 and 909 orders) contain only a few orders, respectively 22 and 14 orders, which is only a small part of the total amount of 278 orders. When determining the delivery reliability it turned out that the 905 orders had a delivery percentage of 18.2 percent and 909 orders 100 percent. Although the amount of orders is small, these orders might bias the main parts of orders that have a delivery reliability of approximately 60 percent (see also table 1, page 6). To get a clear picture of the current delivery reliability, the 905 and 909 orders are left out, resulting in the histogram as shown below in figure 6. Note that in this histogram the

Rollepaal inventory orders (909 orders) are left out, but the other Rollepaal extruder parts

(910 orders) are included. This figure shows that if the orders that are delivered between 0 and 15 days (=approximately 2 weeks) too late would have been delivered on time, the delivery reliability would already improve to a percentage around 90 to 95 percent.

Now the delivery reliability problem has been shown in more detail, it can be seen that the moulds (905) and Rollepaal inventory (909) categories had a significant influence on the variance of lateness. From this point, the focus will be again on the input and output decisions that influence the delivery reliability. As mentioned before, the WLC framework as first introduced by Land & Gaalman (1996) gives a good overview of the relevant input and output decisions, see also figure 3, page 11.

(18)

18

The remaining parts of this thesis will focus on the 3 input control decisions, respectively: the job entry stage, the job release stage and the priority dispatching stage together with the output decisions in terms of controlling capacity that belong to these stages. Thus, each decision moment gives the possibility to influence the input to the load of a specific subsystem of the job shop. An input decision might happen together with a decision to adjust capacity, which affects the output of the job shop (Land, 2004). The current situation of RaRo will be analyzed and the theoretical findings in contrast to the WLC with respect to the processes of RaRo will be shown. Besides that, personal interviews given by the researcher are done with the workforce to gain insight in their feeling and use about the planning (system) and what kind of priority dispatching function they think is best suited to RaRo. The purpose of these interviews is creating knowledge about the exact problem. For example: do they think the information they have is correct and on time, or is there something missing? Why is it missing? And how should delivery reliability be improved according to the workforce? This qualitative data can complement to the quantitative data in order to show shortcomings in delivery reliability tables or throughput diagrams.

(19)

19

4 Job Entry Stage

The logical first stage of the WLC concept is job entry, after all: this is the first stage in where orders enter the company (see figure 7). The input decisions regarding job entry are order acceptance and delivery date promising. The entry level guards against excessive fluctuations of the total accepted amount of work (Land, 2004). Job entry is the moment customers have the first contact with the company, for example with the sales or planning department to request for a quote or ask if it is possible to produce an order in a certain period. This chapter will focus on the job entry stage of RaRo and moreover, the job entry decisions in contrast with delivery reliability.

Customer Enquiry at RaRo

There are three main product categories available at RaRo, namely: Jobbing, Rollepaal and Moulds. Each of these categories has their own sales manager and work preparation employee. Because RaRo belongs to the Rollepaal holding, there is no sales procedure available for Rollepaal products. The other categories “Jobbing” and “Moulds” have their own procedure that differs on some aspects. For a complete description of the procedures is referred to appendix D. The main differences that show up between the procedures are that the procedure of moulds has a “pre-planning phase”, where the procedure of jobbing has not (see figures 8 & 9 on the next page). The reason that the moulds procedure has a pre planning phase is because of the size and processing time of large moulds orders that require a lot of capacity.

(20)

20

Figure 8: part of the jobbing sales procedure

Figure 9: part of the moulds sales procedure

“Inplannen calculatie” means checking if there is enough capacity available for the requested order, to make sure RaRo is able to meet the delivery date. When it is highly certain that an order comes in, it will already be pre-planned (called budget hours). After a discussion with “Jobbing” sales manager Leo Kuier (see also appendix E), to gain more insight in the Jobbing procedure, it turned out that in some cases a capacity check is done for jobbing enquiry. Only if an order request with a delivery date is sent to RaRo, the sales manager contacts the planning department to check if it is possible to deliver the order within the pre-set delivery date that the customer wishes. If there is only a request for a product, the work preparation department of RaRo will define the process steps and (standard) time necessary to produce the order. In the quote to the customer a throughput time will be given. For example: “after

acceptance of the quote, it will take 5 weeks to deliver the product to your company”. When

this is done, there is no “available” capacity check stage included, but the sales managers are quite aware if there are a lot of jobs on hand or not. If this is the case, the profit margin on the estimate might be set higher, to make sure it is still valuable to outsource orders to partners in order to meet delivery dates. Although, outsourcing rarely happens in this stage and products will most of the times be outsourced on short notice (when drawings and cost calculations are already made and when raw materials already have been bought).

(21)

21

can rise to more than a week in certain periods. Besides the waiting time, the actual estimated processing time also needs to be calculated. Most of the time this is done by using historical data from jobs that are quite similar to new jobs. If an entirely new project needs to be pre calculated, this will be done together with engineers and if necessary an employee from the shop floor. To clarify the above text with an example: if a customer quotes for a product and it turns out that this product needs at least 5 process steps (e.g. sawing, milling, metal machining, scrubbing and a quality check) the standard waiting times for these operations will be added, which could be 8 hours, 16 hours, 24 hours, 16 hours and another 16 hours. This means that in total the product has to wait for 80 hours in front of all the machines where it needs to be processed. This is 5 working days because there are 2 shifts. If for example the total processing time of the order is estimated on 32 hours (for example 2 hours sawing, 5 hours milling, 17 hours of metal machining, 7 hours of scrubbing and 1 hour for packaging), there will at least be 6 days added to the estimated waiting time. Not 5 days, while the 17 hours of metal machining can’t be processed in one day. This means that in the quote to the customer at least 5 days (estimated waiting time) + 6 days (estimated processing time) = 11 days as the throughput time to the customer is mentioned. Next to the calculated shop floor time, an estimated time for work preparation and transportation is added to this quote.

Next to the processing times and the waiting times, costs of materials need to be defined for which the suppliers are contacted. This is all done by the work preparation department and once an order is accepted, work preparation also needs to buy these raw materials, define the exact routing and print the routing cards.

The work preparation department, containing of (only) 3 employees for the different product groups, have a lot of tasks to perform, especially when there are a lot of requests for all three product groups. Since each category has its own employee, one can imagine what happens if that employee will become ill. Hans Bakker, preparation employee at RaRo, mentioned that it is possible to do tasks of the other 2 employees, however it is time consuming and only possible if they have not so many requests to handle themselves. Furthermore Hans mentioned that although Leo said that they want to reply to customers within 3 – 5 days, this differs a lot per product. The relative standard / small products can be quoted in 1 day, but complex products like for example a Siemens wind-turbine including the subparts can take more than a week. Also for repair work it is hard to define which steps need to be taken and how many hours of processing need to be done to repair products. The last three months, a lot of moulds orders have been accepted, as well as work for Siemens, which resulted in a highly utilized work preparation department, and most of the times the 3-5 days quoting period was hardly achieved according to Hans Bakker.

The customer enquiry, including the date of enquiry is stored by the sales manager in an enquiry file. Most of the times the sales manager also assigns a date of reply to this file, but this enquiry file is stored manually. Therefore it is quite hard to track quoting days and if the pre-set enquiry time work preparation has for products has not be exceeded. Furthermore it can be said, that hardly no capacity adjustments are made in this stage. A quote in one of the conversations with a sales manager confirmed this: “we never say no to

certain customers”. The studies of Land (2004) & Bergamaschi (1997) confirm that almost

(22)

22

job shops in practice generally say yes to each job” and “the behaviour of most manufacturers is that all orders received by the shop in the order entry phase will be accepted, regardless of shop conditions”.

Customer Enquiry in WLC Theory

Customer enquiry can arise in several ways, sometimes the customer determines the delivery due date, but in other cases the company can determine the due date, or have at least a strong influence on determining the due date. Dealing with customer enquiry is a complex process (Kingsman et al. 1996). Eventually the enquiry process will result in a number of confirmed orders that needs to be produced in a certain period, with a promised delivery date for each order, see figure 10.

Figure 10: process steps related to the job entry stage.

Soepenberg et al. (2007) mention that the average lateness will increase when a larger number of orders have to be produced in a certain period, considering the fact that capacity and promised delivery time will be the same. Increased queuing and waiting times will result in a longer average realized throughput time. Furthermore, average lateness will also increase if tighter delivery dates are promised with the same set of orders. Also an increase in the NOP (number of operations) will generally require larger throughput times. If the processing times of individual large orders are insufficiently taken into account, the variance is also likely to increase. Complex job shops manufacturing industries compete for orders on the basis of price, technical expertise, delivery time and reliability in meeting due dates (Kingsman et al. 1996). A lack of coordination between sales and production at the customer enquiry often leads to accepted orders that will eventually cross the promised delivery time and / or being produced at a loss.

Customer Enquiry and Delivery Reliability

When preparing the cost estimates and setting the delivery time to the customer, it is assumed that somehow the delivery date will be met or excuses will be found (Kingsman et al. (1996). In practice it turns out that this is not the case and that many orders are delivered later than promised or important customers receive high priority at the expense of other jobs. This also results in assigning overtime and subcontracting on short notice. Ultimately this will result in a poor delivery performance or much higher production costs than estimated, which could have been foreseen by proper planning. It is therefore essential to check the current workload when making an estimate and evaluate if extra costs above the standard need to be made for producing the product the estimate is made for.

(23)

23

the floor in the pre-pool phase or by releasing more work for a workstation that runs out of work, e.g. it provides load balancing. The quantity of accepted work generally results from an enquiry process. This can be influenced by the policy of pricing and delivery date promising to customers (Land, 2004).

How Does Customer Enquiry Affect Delivery Reliability?

When taking a look at customer enquiry of RaRo, the first thing that need to be said is that decisions that are made at RaRo are not per definition “wrong” decisions; consequently, it gives only insight in the processes at RaRo and how these decisions influence delivery reliability. The most noticeable reasons that affect delivery reliability with respect to the job entry stage of RaRo are:

- Delivery date promising is often too short;

Both sales, planning and work preparation does not know the exact state of the shop floor. If they monitor the shop floor by a daily check through observing they can see if they are fully occupied or not, however they do not have figures about the future state or exact waiting times. In the order release and priority dispatching chapters (chapter 5 & 6) the throughput diagrams will be discussed. Here will be shown that the standard waiting times that are taken into account will sometimes be exceeded. This results in a wrong calculation of waiting times, while the predefined waiting time is shorter in some periods than the actual waiting time. The result is a promised delivery date to the customer that will be crossed at normal processing.

The main reasons why the delivery date promising is too short are discussed below:

- Capacity control is only done if there is an estimate for a mould or an informal

capacity check is done for Jobbing when a pre-set delivery date is given;

Certain orders are accepted by the sales department without being checked if it can be produced within the delivery date that will be promised to the customer. The work preparation department calculates an estimated throughput time using historical data and taking into account a standard waiting time for workstations. However, as mentioned by Soepenberg et al. (2007) the average lateness will increase when a larger number of orders have to be produced in a certain period with the same defined delivery dates and capacity restrictions. After all, if the capacity per week is 40 hours, and 60 hours of work need to be processed there are three main possibilities: (1) the delivery date will not be met, (2) extra work needs to be done and/or (3) outsourcing needs to be done. Furthermore, there are a lot of orders at RaRo which have a lot of process steps (number of operations), which will generally require larger throughput times. It needs to be mentioned that RaRo is a difficult company because of the three product groups in this case. Because there is no sales procedure of Rollepaal orders, the sales department might not be aware of the amount of orders that come in from this product group. Correct data is therefore difficult to find and missing for the sales department.

- Work preparation has too much tasks;

(24)

24

Besides the task of making quotes in a very detailed manner, they have numerous other tasks like buying raw materials for the quotes that became an order. Also parts that have to be outsourced or subcontracted are mainly done by these employees and preparing the routing and discussing the drawings with the engineering department is their job as well. Because they prepare quotes most of the time as detailed as possible, which takes a lot of time in case of larger projects, this goes at the expense of other tasks. They pay for example less attention to the routing for the shop floor.

- Almost every order will be accepted, including a lot of rush orders and repair work; RaRo is a company that accepts almost every order that comes in. There are some cases when bids are refused, this occurs when RaRo simply doesn’t have the skills or machines to produce a certain type of product. Furthermore orders are sometimes refused if a potential customer already placed several enquiries but never became an order for RaRo. The result of this policy is that a lot of orders and enquiries are still accepted. Especially valuable customers which have repair work or which make an enquiry are accepted: “because it needs to be done”. The advantage in this case is that RaRo is a customer oriented company, but the disadvantage is that accepting a lot of rush orders and a lot of normal enquiries leads to the probability of exceeding the maximum available capacity that gives pressure on the delivery reliability.

To gain a schematic overview, the above factors that influence the delivery reliability at RaRo at job entry stage have been summarized in table 2 beneath:

Table 2: delivery reliability influencing factors related to job entry.

(1) Factor(s) that influence delivery reliability at order acceptance

(2) How/why does it influence the delivery reliability related factor?

(3) Why is that happening? (4) Why?

Delivery date promising is often too short

Throughput time estimation often

not correct Not every process

step is taken into account Not always an available capacity check Relationship with the customer

Routings are being copied from previous

orders Waiting time estimation is often

not correct Takes too much time/exact info is

missing

Work preparation has too much tasks

It needs to be done, too much work accepted in certain

periods

(25)

25

5 Job Release

The job release stage is the second phase of the WLC concept. The release level controls the amount of work on the shop floor (Land, 2004). It must keep the queues of jobs on the shop floor steady and small. The quantity of work on the shop floor is controlled at release level. To figure out how RaRo deals with order release, the process of planning and release of orders has to be investigated. This chapter will show how job release works at RaRo and how it affects delivery reliability.

Job Release at RaRo

After an order is accepted by the sales department, work preparation will determine the routing of an order. Raw materials need to be bought; machines that are best capable for making the product have to be defined etcetera. After this is done, the final routing with pre-calculation of the estimated time the shop floor will spent on the different subsets / process steps of an order is determined. The next step is “booking” the routing into the planning system, which is done by the planning department. The production planner synchronizes the program (Glovia) where the routing is determined with the planning program (Limis), which automatically schedules orders. Backward infinite scheduling is used here, the delivery due date is taken and the routing steps together with a pre-set waiting time for the machines (included in Limis planner) are calculated to the latest starting date of an order. At this time an order should be released to the shop floor. Still, if the orders are synchronized, the order routing is printed and when the drawings are available, the order is already available for processing. This indicates awareness of controlled release in terms of backward scheduling; however the release is uncontrolled, because orders can already be processed due to direct release. The production manager of RaRo, Michiel Timmerman, keeps track of the process of (rush) orders that need to be processed. Nevertheless he mentioned that he does not have a clear view about the amount and type of orders that are coming in the next week or upcoming weeks. Once an order appears in the system it is visible for the shop floor, with a workload for the next few days with a maximum of 2 weeks. Theoretical Findings on Job Release

The WLC concept is a production planning and control method designed to control queues in MTO Job Shop manufacturing environments. At job release, decisions are made regarding which jobs should be released to the shop floor so that processing can continue (Hendry et al. 1997). The centralized decision to release orders from the pool to the shop floor aims at load balancing, and besides this it considers the urgency of jobs. Therefore, order release requires a view of the current shop floor situation (Land, 2004).

(26)

26

Besides a reduction of work-in-progress (WIP) costs, keeping jobs in the pre pool has numerous additional advantages. It creates a transparent shop floor situation which allows faster feedback opportunities, that is of high importance in the turbulent job shop situation. Just in Time (JIT) literature extensively presents the benefits of a leaner shop floor. In addition to these benefits, the time jobs spend in the order pool enables the delay of decisions. The time between customer confirmation of a job (job entry) and the release of the orders to the shop floor allows for preparation activities with the due date in mind, such as the planning of activities (routings) and machines (CNC programs), checking and ordering of the available material (tooling and raw materials). In the broad sense: it deals with production order preparation and insertion to the pre-pool (Bergamaschi, 1997). Until release, the job only exists on paper. The advantage of this method is that changes to the product specification or drawings can be made without disturbing the shop floor and wasting material. Once the release decision has been made, a job should stay on the shop floor till all activities have been completed. Therefore, job release is the last moment to create balanced capacity requirements on the shop floor. By postponing the start of the first operation, excessive work in progress can be avoided, and facilitates the decision to select the right job for processing in order to deliver on time (Land, 2004).

The release decision within the WLC concept is made periodically, like for example weekly or daily. The procedure results in decisions which jobs should be released from the pool on the shop floor (e.g. opening the tap of the bathtub). This procedure considers the current workload situation on the shop floor in combination with the workload contribution of the jobs, together with the relative urgency of jobs. This can be seen as respectively the load balancing and the timing function. The release procedure has thus 2 stages: sequencing and selecting of orders that are to be processed.

Land (2004) distinguishes between 2 main types of order release, namely infinite and finite loading methods. Several subtypes of infinite and finite loading methods are shown in table 3 on the next page. Finite loading methods use a restriction on the amount of work that can be released, where infinite methods do not and therefore assume infinite capacity.

(27)

27

Table 3: release methods categorized

How Does Job Release Affect Reliability?

This paragraph shows theoretical insights compared with order release at RaRo to answer the question how order release influences the delivery reliability. It will give insight in the way delivery reliability is affected at RaRo.

- The loading method used at RaRo focuses more on lateness variance than on

improving throughput (average lateness);

The release method used at RaRo was not easy to define. According to table 3 and the planning system, RaRo uses an infinite loading method. Backward infinite loading is used because the orders are planned backward from due date to earliest start date. RaRo uses infinite loading, because capacity is not restricted. All the machines linked to the planning program do however contain a maximum capacity rate. The system shows if the planner books more orders than the machine can handle, but the planner is hereafter free to balance these operations on other machines and is allowed to schedule different delivery dates. However, scheduling a different delivery date is trying to be avoided. If scheduled capacity for a certain machine goes above the maximum, the first action is to balance the load to another machine that is also capable for the operation (as shown in the figure on the next page of the AF9 machine group. If for example an operation is planned of machine AF97, but if AF96 still has capacity, the first action is to balance the load to other machines in the machine group). The second action is trying to outsource (part) of the order or assign overwork. If these actions still result in too many hours for machines, the customer will be called to ask if it is possible to schedule a different delivery date / or informed that the order will be delivered later than promised. The delivery date might hereafter be modified in the planning system for planning reasons. Modified infinite loading is not used, but the planning system takes into account a certain slack time for machines. However, this slack time is not based on the current shop floor situation.

(28)

28

Figure 12: planned hours for orders at RaRo

improved to 85-90 percent. In that case and due to a relative low variance of lateness, RaRo performs quite well. The focus of this research is therefore more on controlling average lateness, but still keeping control of the variance of lateness.

Figure 12 and appendix G shows the capacity of the AF9 machine group, which has 200 hours of available capacity each week. The light blue part is already scheduled and ready for processing. The darker blue part contains of orders that are in the system yet but where the drawings or routings still have not been printed, or where (raw) materials are not yet available. After acceptance, work preparation can define the routing by assigning a machine or machine group, but when the exact routing needs to be defined the right (or best suited) machine needs to be picked. In other words, these orders are expected to be soon ready for production release or insertion in the order pool.

- There is no use of controlled release;

(29)

29

- Slack time not according to real time shop floor

When defining the loading method it turned out that RaRo does not take into account the real time shop floor situation with regard to slack time. Limis uses a standard value for waiting times. In the created throughput diagrams it can be seen that these slack times are sometimes much higher than predefined, see appendix F and figure 13 (next page). As mentioned earlier by researchers and as turned out in practice: it is hard to define the future state of an order, especially when the number of operations is increasing. This can result in a lower or higher utilization rate than expected for downstream machines on a certain time. This causes also variety in waiting lines in front of machines, and thus slack time. A load balancing release method can improve the throughput and machine utilization as further explained in the next chapter. According to Schönsleben (2007), theoretical capacity can also be adjusted in responding to foreseen changes such as scheduling downtime (due to vacation or preventive maintenance) or scheduling overtime (additional shift) for example. Figure 13 shows part of the throughput diagram of AD 83. Here it can be seen that sometimes a lot of hours are accepted by the machine (added to the queue), while in other periods almost no orders are added to the queue. Of course it is not possible that around the 21st of May the output rate is about 100 hours on 1 day. This was probably a group of (large) orders that have been signed ready at this point in time, resulting in an output rate of almost 100 hours in 1 day.

These orders resulted however in a throughput time for machine AD 83 of 1 week instead of the pre calculated day that the system uses. Accepting a lot of hours a week before the 21st resulted in orders queuing and competing for capacity, while around the 18th of June the arrival is steadier and the throughput rate is again around 2 days.

- Work preparation copies a lot of routings from previous orders.

Most of the times it is not easy to define which steps an order need to make or which machine suits best to process the product. Work preparation uses therefore in most cases the routing of a product that is most similar to the new one. This might lead to wrong estimates, especially for repair work of for example extruder screws. This is the main cause that certain process steps will not be signed ready in the system, because some steps simply don’t have to be done and will therefore not be signed ready. This leads not only to confusion on the shop floor, but also might result that orders need to take different or more process steps than was defined, which gives pressure on the delivery dates.

(30)

30

This chapter, like the previous chapter, also ends with a schematic view of factors that influences RaRo’s delivery reliability which is shown in table 4 below. The table shows the reason why the throughput time at the release stage is too long and gives an overview of relations between the factors (e.g. that the release phase affects the shop floor time). (1) Factor(s) that

influence delivery reliability at order release

(2) Why/how does it influence the delivery reliability related factor?

(3) What happens? / why is that

happening?

(4) Why?

Table 4: delivery reliability influencing factors related to job release.

Throughput time too long

Pre-phase takes too much time (time between acceptance and order pool)

Shop floor time too long / is disturbed

Work preparation has too much tasks

No load balancing at Release

No controlled release

No release method that focuses on load

(31)

31

6 Priority Dispatching

The decision to select the next job for processing is made locally and on time at each machine in (most) job shops. This decision is normally based on some kind of priority rule. Therefore it is called priority dispatching and besides this, it is also considered as the third stage of WLC. This stage differs from release, which schedules the operations for all work centers centrally and statically. The dispatching stage fulfills the capacity for a certain order at a certain workstation. This means, however, that selecting one job means that the processing of other jobs on that machine has to be postponed (e.g. orders compete for capacity). Considering the fact that each job has its own routing the dispatching decision also influences the availability of orders on downstream workstations. This section will review the dispatching decision at RaRo in theory and practice. As this stage has strong associations with the pre-defined subject of this thesis (detailed planning for the shop floor), also the workforce has been interviewed to gain insight in their view about the current situation. Priority Dispatching at RaRo

After the orders have been released to the shop floor they appear in the planning system that the shop floor also uses. This system is called Limis Planner and uses a bar code scanning tool to gain more insight in the progress of orders, but also for assigning the costs to orders. If an order has for example 5 steps, the first step (which might be sawing) with the processing date and estimated time will become visual in the system. At the computer of the sawing machine the planning system will now show a green sign, which means that the order is ready for processing. The next process steps (the downstream stations) can see the order as well, including the estimated process date and time. However, the sign on the computer at these stations will be red, which means that the order is released, but not available yet for that particular machine. The advantage of this method is that it gains insight in the upcoming (planned) work for workstations. In theory, by using this method, employees can exactly see which orders need to be processed in time. The system shows the planned orders for the coming days with a maximum of 10 working days (= 2 weeks) ahead. If a machine starts working on a certain order, the sign at the next workstation turns yellow. A yellow sign means thus that the order is currently being processed at the previous workstation, providing the information that soon that order will arrive at the workstation with the yellow sign. The order that is currently being processed at a workstation show a whole yellow row. Figure 14 on the next page is a print screen from the planning system, which visualizes the above information. At the workstations on the shop floor this kind of screens can be generated. At the moment of the print screen order 9172414 / 8-12 was being processed (yellow row), 4 orders were queuing (green) and 2 orders were scheduled but not ready for processing yet (red).

Achieving higher delivery reliability at Ramix/Romit B.V.