A Bottleneck Analysis to Find a
Winning Throughput Design
A Bottleneck Analysis to Find a
Winning Throughput Design
Ruud Pepping
Universtity of Groningen
Faculty of Economics and Business
January 2011
Acknowledgement
I traveled a long way before this thesis obtained its current shape. I started my research in February 2010 in Canada at Hudson Boat Works where it took about three months to collect data. In June I returned to Europe to work through the summer for Hudson and in November, I started to write which resulted in this thesis it in January 2011.
First of all, I would like to thank Glen Burston who gave me the opportunity to write my thesis at Hudson Boat Works. I also would like to thank him and the other staff of Hudson for taking great care of me in the period I was in Canada. Also, I would like thank Nicky and Erik for all the time they spent on me.
Management Summary
Hudson is dealing with tight due dates and a growth in demand. In order to keep on handling the due dates and to be able to fulfill all orders in the future, it is important to know what the current capacity of the plant is and how the lead times can be kept short.
The production process is a tandem line process with a fixed product mix. These properties make it plausible that one bottleneck determines the plant capacity. Data analysis has shown that the paint preparation department is the bottleneck.
Paint preparation is overutilized. This is due to a few reasons. At the input of the department the product quality is not always high enough. Furthermore, the short term planning which is performed by the team leader of paint preparation causes a high time pressure on the workers. This in turn results in output quality issues and causes rework which increases the process times.
Index
Acknowledgement... 3
Management Summary ... 4
1. Introduction ... 6
2. Research design and methodology ... 7
2.1 Introduction to the current situation ... 7
2.2 Thesis structure ... 8 2.3 Conceptual model ... 9 2.4 Methodology ... 9 2.5 Overview... 10 3. Bottleneck identification ... 11 3.1 Plant analyses ... 11 3.2 Bottleneck analysis ... 12
3.3 Consequences for the bottleneck ... 14
4. Bottleneck exploitation ... 15
4.1 Process time reduction ... 15
4.2 Impact for the bottleneck... 15
5. Subordinate processes to the bottleneck ... 16
5.1 Increasing labor capacity ... 16
5.2 Process time reduction ... 16
5.3 Expected impact for the bottleneck ... 17
6. Conclusion ... 18
1. Introduction
About Hudson Boat Works
Hudson Boat Works is a global rowing boat manufacturer located in Canada. The company builds racing boats competing on an Olympic level. Hudson has thirty production employees producing about 440 boats per year. About twenty different types of rowing boats can be ordered in a particular weight class, shape, and type of carbon.
The motivation for this research
Recently, the demand at Hudson has been growing. Hudson expects that it cannot fulfill its due dates in the near future due to the shortage in production capacity. The due date is fixed and therefore late orders are unacceptable. To deal with the growing demand, the management would like to know the current capacity of the plant and would like to know what it has to do to extend this capacity. To give an answer to these questions, the company asked for external help with the objective to obtain more insights into the current capacities and the expansion possibilities of the resources.
Thesis overview
2. Research design and methodology
At this point in time, employees work overtime to reach due dates. Furthermore, because the market is growing over all the continents of the world, Hudson expects capacity problems in the near future. A growth in demand of 5% per year is expected. Since there is no insight in how many boats the plant can produce on the long term, insight is needed in the current capacity of the plant,its expansion possibilities and its lead times. Regarding to this problem description, the next research question is obtained:
What is the capacity of the plant and how can this capacity be expanded, without creating a growth
in lead times extending lead times, when demand is rising with 5% per year?
Three sub-questions are composed to answer this research question:
1. Where is the bottleneck located?
2. How can the bottleneck optimally be exploited without investment in resources? 3. How should all processes be subordinated to the bottleneck?
However, first the design of this research is explained to describe the nature of these sub-questions.
2.1 Introduction to the current situation
Rowing boats are high cost items. This is due to high costs in raw materials and a high labor intensity. Furthermore, due to the size of the product family and the customer preferences the production process is a make-to-order process. The total process of boatbuilding consists of five major construction stages. These stages are lay-up, construction, paint preparation, paint, and final assembly. The process is visualized in figure 1 as a basic form.
Raw materials
Lay-up Construction Paint Final assemy
CODP Welding Part assembly Raw materials Finished products Assemble d parts CODP (parts) Paint preparation Raw materials Hull cutting (CNC) Lay-up parts Oven Part cutting (CNC) Raw materials Oven Part cutting (CNC)
2.2 Thesis structure
The production process at Hudson is a tandem process which is, in fact, an individual line with a single routing. If the product mix is solid, the capacity for this type of line is identical to the rate of the bottleneck (Hopp & Spearman, 2008).
Dr. Eliyahu M. Goldratt introduced the Theory of Constraints (TOC) in his book “The Goal” (1984). The TOC is used as a guideline to answer the research question. More particularly, the first three steps of the TOC correspond to the three sub-questions of the research question.
1. Identify the constraint
2. Decide how to exploit the constraint 3. Subordinate all processes to the constraint
2.2.1 Identify the constraint
The first step in the TOC, “identify the constraint”, is the first sub-question “What is the bottleneck?” The next theory will be used to answer this question:
Hopp & Spearman (2008) show that at an individual line, the bottleneck is the department with the highest long term utilization (u). Therefore, to calculate the utilization of the departments, the next formula can be used:
u= ra/re
In this formula, (ra) is the arrival rate per workstation and (re)is the effective production rate.
At Hudson the maximum effective rate of a department is mainly determined by labor capacity and not by machine capacity. Since workers are interchangeable within their departments, the next formula can be used to define (labor) capacity at the departments (Hopp & Spearman, 2008):
re = (n/ T0) * h
In this formula, (n) is the number of workers, (T0) is raw process time, and (h) is labor hours
A second method to identify the constraint is to look at the station with the longest queue (Goldratt) This method can be used when process data is not available.
2.2.2 Exploit the constraint
The second step in the TOC, “Decide how to exploit the constraint”, is used as a guideline to answer the second sub-question. Since the bottleneck sets the pace of the plant it is important to exploit the constraint optimally. It has to be made sure that the constraint is always working and that it is at no time idle (Goldratt).
2.2.3 Subordinate all processes to the constraint
2.2.4 Total lead time
As explained before, the department with the longest queue is the bottleneck department. To shorten lead times, queues need to be as short as possible.
Queue time is caused by utilization and variability. A reduction of utilization can be done in two ways: One, increasing the bottleneck rate (adding effective process time) or two, reducing the arrival rate into the bottleneck. Reduction of variability will also reduce queue times. Process variability can be reduced by reducing repair times, improving quality or to reduce rework. Reducing arrival variability will also positively affect the queue length. To smooth the material flow, process variability has to be decreased at upstream stations, which can be done by improvements of scheduling. A possibility is to install a pull system. (Hopp & Spearman, 2008).
2.3 Conceptual model
The conceptual model shows that the three variables n, T0, and hdefine the effective process time for
a department (re). The bottleneck can be found by the utilization of departments. This can be
determined by using the arrival rate and the effective process times of departments. A second method to find the bottleneck is by measuring queues.
Lead times Queue length
Arrival rate Labor Hours
Workers Raw Proces time
+ Utilization + Variety + + + -Effective process time department +
Figure 2 Conceptual model
To understand which decisions have to be made to expand capacity in an economical way, the bottleneck needs to be analyzed. Therefore, the second and third sub-question are postulated:
“How can the bottleneck be exploited optimally without investment in new resources” and “How should non-bottleneck processes be subordinated to the bottleneck.”
2.4 Methodology
For finding the bottleneck, quantitative data like process times and the capacities per department is collected. Interviews are used to support quantitative data that needed to be clarified. Furthermore, to elevate the bottleneck throughput capacity, adjustments need to be made in one or both of the variables: labor capacity and the process time per unit. Qualitative research supports the analysis of the bottleneck.
2.4.1 Qualitative data
In order to justify any remarkable findings in the quantitative data, interviews with the team managers and the operations manager are used.
2.4.2 Quantitative data
To define raw process times, arrival rates, queues and labor hours, item tracking sheets are used. For a period of one month, tracking sheets are attached to the order sheets which travel with every boat through the process. Team leaders in each department are instructed to collect data such as start times, finish time, and the amount of hours that are spent on a product. This data is collected for every boat that was built between the 31st of March and the 30th of April, 2010, which amounts to about twenty rowing boats. To start off, the teams are given a clear instruction as to how they should fill out the data on the sheets. Next, the actual data per department is collected: first, the starting date and time of the production on the boat are recorded. Second, it is registered at what date and time production on the boat is finished, and third, it is calculated how many hours have been spent on the boat. This is done by adding up the hours every employee worked on a specific boat.
To identify the number of workers and the working hours of employees, employee punch cards and the order release list is used. To get an overview over this gap, extra data is collected. Employee punch cards and the order release list over the period from 1st of January until 1st of May are collected. The registered employees are divided over the departments and so are the effective process times and utilization in the departments clarified. Unfortunately, it was not possible to collect required data for all departments. Therefore, queues are analyzed to make sure the right bottleneck is found.
Finally, few departments work overtime. Since this is not desirable for the company, this can be called “overutilization”. Therefore, utilization sometimes reached over a hundred percent.
2.5 Overview
To come to the answer of the research question: What is the capacity of the plant and how can this
capacity be expanded, without creating a growth in lead times extending lead times, when demand is rising with 5% per year?
Next, three sub-questions are composed. These questions are answered in respectively section three, bottleneck identification; section four, bottleneck exploitation and section five, subordination of processes to the bottleneck .
1. Where is the bottleneck located? (Section three)
2. How can the bottleneck optimally be exploited without investment in resources? (Section four)
3. Bottleneck identification
3.1 Plant analyses
The needed data for identifying the bottleneck is shown in table 1. The table contains information about process time, queue times, lead times and work in process per department.
The average labor hours per week and the number of workers per department are collected from employee punch cards and together it shows the available labor capacity. The process time per boat is collected by means of tracking sheets. Due to different process times per boat type, the process times are averaged over the product mix. The theoretical capacity of boats per week is calculated out of this data. From the order release list are the processed boats per week collected which is the demonstrated capacity. With these data, the utilization per department is calculated.
Due to a high variation in tasks and performing activities on the different products at the same time it was too complicated to measure process times for the CNC department, paint department and final assemblies. Furthermore, the table shows a higher number of processed boats per week at paint preparation than theoretically is possible. Here, compromises in quality must be made since the demonstrated capacity is higher than the theoretical capacity.
Table 1 Analysis of departments
Average labor Hours/ week Nr. Of workers Process time per boat in hours Theoretical capacity Boats/ week Demonsrate d capacity boats/ week Utilization Lay-up 41 6 32 7,69 6,76 88% Construction 38 4 11 13,82 6,76 49% CNC 44 1 X X 6,76 X Paint preparation 43 6 42 6,14 6,76 110% Paint 40 4 X X 6,76 X Final assemblies 32 2 X X 6,76 X Total
Operations downstream of paint preparation have almost no queues and all departments have process times that are no longer than one day so these departments have no difficulty in handling the throughput. For CNC and final assemblies it was not possible to measure throughput due to a lot of sub activities which blurred the data. Nevertheless, working in the production line has first priority for the workers at CNC and final assemblies and since they perform a lot of sub-activities, there is a lot overcapacity in the line in these departments.
Table 2 Queues, lead times and WIP
Processed boats per hour Average queue in boats Service time in hours WIP in boats Lay-up 0,06 26,4 1,49 Construction 0,06 0,1 74,4 4,20 CNC 0,06 0,1 2,4 0,14 Paint preparation 0,06 5,0 45,6 2,85 Paint 0,06 0,1 14,4 0,82 Final assemblies 0,06 0,4 9,6 0,56 Total 5,8 172,8
The question “Where is the bottleneck located” is clearly answered and the bottleneck is identified. Paint preparation has the highest utilization and the longest queues’. Therefore, this department is the bottleneck.
3.2 Bottleneck analysis
By the bottleneck identification is showed that paint preparation has the highest utilization of all departments. Furthermore, the work in process caused by paint preparation is the highest of all departments. Since the research design shows that the throughput in the bottleneck identifies the throughput capacity of the facility, paint preparation has to be analyzed in depth. To learn how the bottleneck is functioning, and to understand the effect of changes in the production process, this section discusses in more detail on the properties of the paint preparation department.
Table 3 Variations in order throughput CV in process times CV till shipping date Lay-up 0,25 X Construction 0,23 X Paint preparation 0,35 0,34 Paint 0,23 X Final assemblies X X
3.2.1 Variation in input quality slows down process times
Processing carbon is a complex process which requires craftsmanship on a high level. In fact, some technical activities need to be automated but as no resources are available, these are performed by hand. Therefore, some activities upstream of paint preparation have high error sensitivity which makes it impossible to be fully consistent in quality output. Furthermore, due to the complexity of the production process, variety in craftsmanship irrevocably results in variety in quality. The paint preparation department is the final department where inconsistencies can be repaired in an efficient way. Therefore, inconsistencies in the lay-up, construction, and the CNC department all have to be repaired in the paint preparation department which results in a high variation in process times. Moreover, the research and development department is continuously working on improvements in the materials. Experiments with new materials cause high variation in the process times. Nevertheless, these activities are inevitable as the product needs to be innovative to be competitive. As the R&D activities are not always planned within the production schedule, this leads to high variations in labor hours per item.
3.2.2 Variation in available time to finish a boat slows down process times
The customer delivery schedule is an informal guideline for the due dates of the paint preparation department. This delivery schedule is based on container deliveries or deliveries on regattas which usually consists of a bigger quantity of orders. Based on this delivery schedule, the team manager of paint preparation calculates the order due dates for his department. He calculates his due dates in a way there is just enough time for the departments downstream to finish the orders on time. Usually, this is two to four days before orders are shipped. Resulting from this way of planning is a high variation in workload, dependent on the size and moment of the shipping batch. This leads to varying process times, labor hours and throughput flow. Due to the high workload, errors are made which have to be repaired. This in turn, leads to longer process times.
3.2.3 Labor capacity
3.3 Consequences for the bottleneck
3.3.1 High utilization
When the delivery date of a transport approaches, the paint preparation team is taken by surprise with regards to the amount of work that needs to be done. Overtime and weekend work is usual at these periods. Next to these negative consequences for the workers in this department, this lack of planning also results in a bull-whip effect which works through in the departments downstream. As downstream of paint preparation no slack time is available, these departments have to rush and work overtime to finish the orders on time. This is not only an indication that the utilization at paint preparation is too high but is also is a basis for an increase in utilization in downstream departments.
3.3.2 Variation in output quality
Rushing at paint preparation is also the basis for inaccuracy in the output quality since the activities performed at paint preparation are very subtle. This creates a margin for interpreting the level of the output quality at this department. It is attractive to be less accurate when running out of time. Since the time available is subordinated to the delivery schedule, it results continuously varying quality. An acceptable quality for the product is not defined. Since the flexibility in process time, respectively flexibility in quality, is used to reach due dates it is clear that the planning is not solid enough to reach due dates with boats with a consistent quality. These problems in planning and quality needs to be solved before the throughput can grow in a controlled way.
4. Bottleneck exploitation
To exploit the bottleneck optimally, process times need to be reduced without investment in resources. To do this, the first step is to reduce variety in quality and throughput. To gain variety reduction, suggestions in the field of quality control and order throughput control are given.
4.1 Process time reduction
4.1.1 Quality control
Since the output quality differs due to different levels of craftsmanship and time that is available, it is interesting to create a new quality system. When output quality is defined and quality checks at the output of this department are executed, a more consistent quality in the hulls might be reached. Furthermore, this will help in stabilizing the process times of paint preparation. The stabilization of output quality can only be performed if time pressure is reduced and as this is related to the planning system, a quality system only works when a proper planning system will be adapted too. More consistent quality will reduce the amount of rework in repairing quality issues so the average process time will be reduced.
4.1.2 Change in order release system
At this moment the orders are released at the lay-up department. Since the paint preparation has less capacity than lay-up, queues appear upstream paint preparation. When changing the order release system to a capacity based order release at paint preparation, the system will be able to manage its own queues. This release system is a system where a limit of WIP is established and no release is allowed into the line whenever the WIP is at or above the limit. This results in a more constant WIP level but more important, this release system supports the idea of cross-trained workers: As upstream departments will be out of work when queues rise at paint preparation they can help this department until the queues are on an acceptable level.
Since late orders are not allowed. The medium term planning should have a due date for paint preparation which is derived from process times related to hulls with a consistent quality. Some slack in this due date might be handy since variation will never be reduced to zero.
4.2 Impact for the bottleneck
Exploiting the bottleneck this way will reduce process times and creates a possibility for cross-training workers. Furthermore, due to a more consistent throughput flow, errors will be reduced which will shorten the process times and gives more control in the production. It gives room for adjustments in the quality variety as the bottleneck sets the pace. This stability is a fundamental basis for cross training workers.
5. Subordinate processes to the bottleneck
As mentioned in the bottleneck analysis, there are few external factors that have influence on process times of paint preparation. First, the variation in input quality and the second is the R&D activities which put pressure on the process times of the bottleneck. Furthermore, idle workers at non bottleneck departments should focus on supporting paint preparation, for example, by helping in that department.
5.1 Increasing labor capacity
5.1.1 Add labor capacity: Cross train workers
Idle workers at non-bottleneck departments should work at the bottleneck. Since 90% of the workers in the plant have the skills in to work in this department, not much investment in training is needed. Furthermore, this way will workers upstream of paint preparation experience on first-hand the negative consequences of delivering bad quality; this could work as a self learning cycle.
5.2 Process time reduction
5.2.1 Reduce process times: Improve output quality at departments upstream of the bottleneck
Reallocating workers and maintaining a proper quality system upstream of paint preparation, helps in reduction the input quality for paint preparation. This will shorter the process times, since paint preparation doesn’t have to repair errors made by the upstream departments.
5.2.2 Reduce process times: R&D in low season periods
As R&D activities are performed to reduce the complexity of the process it is important to keep executing innovation experiments. Nevertheless, it would help when R&D activities are planned and performed in the off-season period.
5.2.3 Reduce process times: Automation
5.3 Expected impact for the bottleneck
By subordinating processes to the bottleneck, utilization of other department will rise in favor of the bottleneck. The throughput at paint preparation will be higher and variation in process times and quality will be reduced.
5.3.1 Impact by reallocation workers
Since the utilization at paint preparation is 61 percent higher than construction and 22 percent higher than lay-up this might benefit in total one full time employee which will be close to a 15 percent increase in labor capacity.
5.3.2 Impact of process time reduction
It is hard to estimate the impact of the suggested ideas. Nevertheless, as top coaches say in rowing: one small improvement does not help you win. It is improving all small improvements that are within your reach to adapt. Process times will be reduced, but more important, variation will be reduced which is the basis for capacity growth.
6. Conclusion
The throughput capacity of Hudson Boatworks is dependent on the capacity of its bottleneck department, which is paint preparation. Data has shown that this department is already over utilized. To decrease this utilization, process times have to be reduced or labor capacity has to be increased. By following the steps of the theory of constraints, first the bottleneck needs to be exploited. This can be done by improving quality which will decrease the amount of repairs, and therefore will decrease the process times.
Idle workers upstream of paint preparation should be reallocated at paint preparation. Although this way of subordinating workers from other departments will increase the utilization at these workstations, it will also increase the throughput of paint preparation. By changing the order release system towards paint preparation a maximum queue system can be used as an indicator for when to help at paint preparation.
Process times can also be reduced by improving the product quality at the input of paint preparation. Bad product quality caused by unsecure work at upstream departments is being reworked by workers of paint preparation. The rework time has a bad influence on the process times of paint preparation. The quality level might rise by cross-training workers at upstream departments. By working at paint preparation, the workers might face the quality they deliver which can work as a self learning circle. Product quality will increase and this will reduce process times for paint preparation. The research question: “What is the capacity of the plant and how can this capacity be expanded,
without creating a growth in lead times extending lead times, when demand is rising with 5% per year?” can be answered as next:
The current throughput capacities are lower than the demand. Utilization of the bottleneck department is 110% which is possible due to overtime. By reducing the variation in process times and quality this utilization can be reduced. Furthermore, by cross-training workers, labor capacity will be added, which will decrease the utilization. When process times and quality are controlled, labor can be added to this bottleneck.
