Integrating Six-Sigma Methods and Lean Principles to Reduce Variation and Waste in Delivery Performance to the Customer (Production System) By E. Dan Douglas Submitted to the System Design and Management Program in Partial Fulfillment of the Requirements for the Degree of Master of Science in Engineering and Management BARKER At The MASSACHUSETTS INSTITUTE OF TECHNOLOGY Massachusetts Institute of Technology APR 17 2003 LI I LIBRARIES February 2003 2003 E. Dan Douglas. All rights reserved The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part. Signature of Author: E. Dan Douglas System Design and Management Program January 2003 Certified by: Prof. Thomas Roemer Thesis Supervisor MIT Sloan School of Management; Operations Management Accepted by: Steven D. Eppinger Co-Director, LFM/SDM GM LFM P~so of Management Science and Engineering Systems Accepted by: Paul A. Lagace Co-Director, LFM/SDM Professor of Aeronautics & Astronautics and Engineering Systems 1 2 Integrating Six-Sigma Methods and Lean Principles to Reduce Variation and Waste in Delivery Performance to the Customer (Production System) By E. Dan Douglas Submitted to the System Design and Management Program on December 13, 2002 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Engineering and Management Abstract Shortening order-to-delivery (OTD) times is a strategic business goal for companies in many industries and the automotive industry in particular. Advantages of shorter delivery times include lower inventory levels, less obsolescence, and the ability to respond more quickly to changing markets. As a consequence, many companies are in the process to reduce average OTD times and employ sophisticated measurement systems to determine the average delivery time from customer order to customer delivery. While reducing OTD times may lead to considerable efficiency improvements within an enterprise, the customers may still benefit very little from such improvements. As a consequence such strategies often fail to exploit the key strategic advantage of increasing customer satisfaction. The customer's focus is not so much on average delivery times, but on variation around the average delivery times. In many scenarios, low variation is much more crucial to the customer than a low average OTD time. Yet many measurement systems today have an almost exclusive focus on averages. This is also the case in the system being studied here. In the case of automotive original equipment manufacturers (OEMs), only some customers receive the vehicle when they expect the vehicle due to the current system architectures and the system dynamics of these architectures and processes. Some customers receive vehicles early, while other customers receive vehicles late. The customers experience the 3 variation from the average and are not satisfied with current delivery performance. Delivery variation must be reduced to better meet individual customer needs. The production system in automotive manufacturing presents a unique opportunity to study both the source and impact of delivery time variation and compare it to average delivery performance. The thesis investigates delivery variation performance in the production system of a major OEM. The relationship of the production system to the OTD system is discussed. The production system is decomposed into three subsystems: order, build, and test, inspect, rework (TIR). The TIR subsystem was determined to be the largest contributor to delivery time variation in the production system. Data was collected and analyzed at the system level and the subsystem level to enable a comparison between the average delivery performance and the delivery variation performance. The TIR system has the most delivery variation, but the best average delivery performance. The significance is that the TIR subsystem was not seen as a source of customer dissatisfaction for delivery of vehicles from the production system. Good performance in average vehicle build time was assumed to yield good delivery performance and customer satisfaction. The possible causes for this variation are outlined for the TIR subsystem. The TIR subsystem is also the production subsystem with the most waste from the Lean enterprise perspective. The waste in the TIR subsystem causes variation in delivery. This variation then causes waste in the relation between the enterprise and customer. There is a dynamic in which the internal waste causes increased waste outside the production system. The cost for this waste is incurred by the customer and enterprise and is discussed as part of this work. Thesis Supervisor: Prof. Thomas Roemer Title: MIT Sloan School of Management; Operations Management 4 Acknowledgements The Massachusetts Institute of Technology (MIT) System Design and Management (SDM) Program provided me with the opportunity to improve my knowledge and skill in business management, leadership, engineering, and system design over the past two years. The completion of program requirements along with a full work schedule required many long days and short nights. The opportunity was made possible by the support of those around me. I would first like to recognize Sharon, my wife, who supported me and sacrificed her own needs during this time. Sharon worked to make sure the time I spent with her and Zoe, my daughter, was quality time, not time spent taking care of the many distractions of everyday life. Sharon has supported me throughout and has been flexible to the needs of my school and work schedules. Thank you, Sharon, for taking care of Zoe and me. My daughter Zoe turned four this fall. I was out of town for four months last fall when Zoe turned three. Zoe I appreciate your patience with dad while I have been away. I see how upset it makes you when dad does not come home at night. Hopefully the long periods away will be few now that school has ended. We have more time to laugh and play and learn. I would also like to thank Thomas Roemer, my thesis supervisor, for the time we spent together working on my thesis material and other subjects of interest to me. Thomas's operations expertise, previous works, and interest in the automobile industry allowed insight into the subject of this thesis. Thomas provided me with ideas and insight that made a difference in the way I think about the subject matter. The work I completed in the SDM program was made easier by the support of my plant management, product design management and peers. I never questioned their dedication to make me successful. They provided me with financial support to complete the SDM program and with individual support when there was more work to be completed than hours in the day. I would like to especially recognize the persons in the 5 IT department that services my facility for their support of my crazy requests and for developing tools to help gather and analyze data for this thesis. The IT team has provided outstanding customer service and team members have become friends I look forward to seeing each day. Thank you. Last, I would like to thank my peers in the SDM program. You give new meaning to the phrase "work hard, play hard". It is good to know you can have so much fun and receive so much support from those who are going through similar challenges as yourself. Thank you for helping me to keep things in perspective. 6 Biography E. Dan Douglas is a Certified Six-Sigma Master Black Belt and a Product Engineering Manager at a major automobile manufacturing company. He is a leader in the operating team for two manufacturing and assembly facilities within the company. Dan has held various positions within the company for the past 10 years including design engineer, development engineer, engineering supervisor, Six-Sigma black belt, manufacturing plant resident engineer, and program manager. Dan has worked on several different vehicle systems and subsystems at the company. Dan worked for the DANA Corporation for five years prior to joining his current company. Dan held engineering positions in several different departments while he was at DANA. Dan graduated from Kettering University in 1993 as a Bachelor of Science in Mechanical Engineering. 7 8 Table of Contents A bs tra c t ........................................................................................................................... 3 Acknowledgem ents ..................................................................................................... 5 Biography ........................................................................................................................ 7 Table of Contents ........................................................................................................ 9 Table of Figures........................................................................................................ 11 Introduction ......................................................................................... 13 1.1. M otivation............................................................................................ 13 1.2. Research............................................................................................... 14 1.3. Thesis outline ....................................................................................... 15 Average perform ance and perform ance variation ................................. 17 1. 2. 2.1. The difference between average performance and performance variation: A sim ple case........................................................................................ 2.2. 17 Expected delivery performance and delivery performance variation........18 Delivery variation background............................................................. 21 3.1. Three segments of the order-to-delivery (OTD) system ....................... 21 3.2. State of the current production system ................................................. 23 3.3. Critical to delivery (CTD)...................................................................... 25 3.3.1. W hat is delivery tim e?......................................................................... 26 3.3.2. Delivery tim e variation......................................................................... 27 3. 3.4. Who are the customers and what do the customers want?..................30 3.5. The cost of delivery variation................................................................ 34 3.6. System costs........................................................................................ 35 3.7. M otivation to drive change in the system ............................................ 42 M ethods .............................................................................................. 45 Six-Sigm a and DMAIC ......................................................................... 45 4.1.1. Define phase....................................................................................... 47 4.1.2. M easure phase ..................................................................................... 48 4.1.3. Analyze phase..................................................................................... 50 4.1.4. Im prove phase ..................................................................................... 51 4. 4.1. 9 4.1.5. 4.2. Controlphase........................................................................................52 Lean principles ..................................................................................... 54 4.2.1. Elimination of waste .............................................................................. 54 4.2.2. Value stream mapping ......................................................................... 57 4.3. 5. Integrating Lean and Six-Sigma to find sources of variation and waste ... 58 Automotive production system analysis ............................................... 59 5.1. Automobile OEM production system description.................................. 59 5.2. Process maps and system decomposition ........................................... 60 5.2.1. Order-to-Delivery process 61 5.2.2. Production process map ....................................................................... 5.3. p ............................................................ 61 Delivery time variation measurement system..................63 5.3.1. Measurement system operational definition......................................... 63 5.3.2. Measurement system process map......................................................66 5.4. Current performance ........................................................................... 67 5.5. Cost associated with the current performance ..................................... 70 5.6. Analysis of variation in the current system ........................................... 72 TIR subsystem possible sources of variation...................80 5.7. 5.7.1. TIR subsystem decomposition............................81 5.7.2. TRsubsystemprocessmap............................................................. 5.7.3. Cause and effect diagram for the TIR subsystem................84 5 .7 .4 . 5.8. 5.8.1. 5.7.4. C&E marix......................................86 C &E matrix ............................................................................................... 86 Analysis of variation in the TIR subsystem....................90 Measurement system for key possible sources of variation in the TIR subsystem ........................................................................................... 5.8.2. 83 . 92 Target reduction in variation in the TIR subsystem .............................. 92 Conclusions and recommendations ..................................................... 93 6 .1. C onclusions......................................................................................... . 93 6.2. Recommendations...................................................................................96 6. G lo s s a ry ...................................................................................................................... 10 1 R efe re n ces .................................................................................................................. 10 3 E n d n ote s ..................................................................................................................... 1 05 10 Table of Figures Figure 1 Order-to-Delivery system at the first level of decomposition..................22 Figure 2 OTD System with production system at the second level of decomposition . . 25 ........................................................................................................... 28 Figure 3 Variation to Customer Expectation ........................................................ Figure 4 Variation measure in the production system..........................................30 Figure 5 Customer and enterprise costs............................................................. 35 Figure 6 Example of an initial OTD delivery variation plot ................................... 49 Figure 7 Example of an improved OTD variation plot as compared to the Measure P ha se ................................................................................................ . . 53 Figure 8 System dynamic waste / variation loop................................................. 56 Figure 9 Overview of the automotive Order-to-Delivery system .......................... Figure 10 High level process map of the overall automotive production system from 61 the time the plant receives the order until the order is shipped............. 62 Figure 11 Days of delivery variation (SPAN) for 95% of the volume......................65 Figure 12 Measurement system map .................................................................... Figure 13 Variation from manufacturing order date to actual ship date for one day of 66 production system orders (short term variation for the entire production process) 8 days .................................................................................. Figure 14 . . 68 Overall production system variation for 95% of the total population over 100 days of production orders......................................................................69 Figure 15 Overall delivery average (expectation) for 95% of the population over 100 days of production orders ...................................................................... 70 Figure 16 Relative inventory cost for delivery variation......................................... 71 Figure 17 Variation from manufacturing order date to start of production date for one day of production (short term variation for the order subsystem of the production system ) ............................................................................... 11 73 Figure 18 Variation from start of build date to end of build date for one day of production (short term variation for the build subsystem of the production system ).................................................................................................. Figure 19 Variation for the test, inspect, and rework subsystem prior to the actual ship da te ..................................................................................................... Figure 20 . 74 . . 75 Delivery variation performance for the entire system and each of the three subsystems at the first level of decomposition...................................... 76 Figure 21 Average delivery performance for the production system and each of the three subsystems at the first level of decomposition ............................ 77 Figure 22 Correlation between order and build subsystem delivery variation........78 Figure 23 Correlation between order and TIR subsystem delivery variation..........79 Figure 24 Correlation between build and TIR subsystem delivery variation .......... 80 Figure 25 Process map for the TIR subsystem (some steps aggregated to simplify the process m ap)....................................................................................... . . 84 Figure 26 Cause and effect diagram for the TIR subsystem..................................86 Figure 27 TIR subsystem cause and effects matrix with the top 6 possible causes h ig h lig hte d ............................................................................................. . 89 Figure 28 Problems per 100 vehicles built measured in a major inspection area......91 Figure 29 System dynamic waste / variation loop................................................. 95 Figure 30 Expected long term performance of delivery variation in the production system with and without waste in the system ............................................ Figure 31 96 Example of a TIR subsystem delivery performance chart......................97 12 1. Introduction 1.1. Motivation Shortening order-to-delivery (OTD) times is a strategic business goal for companies in many industries and the automotive industry in particular. Advantages of shorter delivery times include lower inventory levels, less obsolescence, and the ability to respond more quickly to changing markets. As a consequence, many companies are in the process to reduce average OTD times and employ sophisticated measurement systems to determine the average delivery time from customer order to customer delivery. While reducing OTD times may lead to considerable efficiency improvements within an enterprise, customers may still benefit very little from such improvements. As a consequence such strategies often fail to exploit the key strategic advantage of increasing customer satisfaction. For example, a restaurant may have very short average delivery times between customer orders and delivery, indicating that crucial resources, such as available space are efficiently employed. The sole focus on average delivery times however cannot distinguish between a system where allindividual dishes arrive in a short time and a system where the dessert arrives before the appetizer. In other words, the customer's focus is not so much on average delivery times, but on variation around the average delivery times. In many scenarios, low variation is much more crucial to the customer than a low average OTD. Yet many measurement systems today have an almost exclusive focus on averages. 13 This is also the case in the system being studied here. In the case of automotive original equipment manufacturers (OEMs), only some customers receive the vehicle when expected due to the current system architectures and the system dynamics of these architectures and processes. Some customers receive vehicles early, while other customers receive vehicles late. The customers experience the variation from the average and are not satisfied with current delivery performance. Delivery variation must be reduced to better meet individual customer needs. The OTD system has a few key subsystems at the first level of decomposition of the system architecture. These subsystems are the order system, the production system, and the distribution system. The production system in particular has been optimized to provide the lowest average vehicle build time, but delivery variation in the overall production subsystem is not emphasized. The production system in automotive manufacturing presents a unique opportunity to study both the source and impact of delivery time variation and compare it to average delivery performance. 1.2. Research The research follows the Six-Sigma DMAIC (Define, Measure, Analyze, Improve, and Control) methodology to quantify average delivery performance and delivery variation performance in the production system. The define, measure, and analyze (DMA) phases will be used to guide collection and analysis of the data from the production system of an automotive OEM. Lean enterprise principles and tools are integrated with Six-Sigma methods to identify the sources of variation and help define waste elimination that will improve delivery variation performance. 14 The improve and control (IC) phases are not part of this thesis. These two phases are highly dependent on resources for implementation that are outside the control of the author. Recommendations will be made about how the improve and control phases could be completed at the end of the thesis. 1.3. Thesis outline The thesis begins with a discussion of average performance and variation performance in delivery of products and services in section two. Section three covers general background of the order-to-delivery (OTD) system using several industries for illustration. The expectations and costs for the key stakeholders are also discussed in section three. Section four contains an outline of the analysis method using the SixSigma DMAIC process and Lean principles and tools. Section five contains the specific data for the automotive OEM production system being used as a case in this research. The conclusions and recommendations in section six summarize the significant results of the data analysis and outline the significant opportunities for improvement. 15 16 2. Average performance and performance variation 2.1. The difference between average performance and performance variation: A simple case. The important difference between average performance and variation performance is easier to understand when illustrated using a familiar experience. A common situation used to explain the difference is the example of a group of students in a classroom. The room is kept at 50 degrees (F) for the first four hours of the day. Most students find this too cold and uncomfortable. The students go to lunch and while they are at lunch, the heating system turns on and heats the room to 90 degrees. The students return from lunch and spend the next four hours in 90 degree heat. The students are uncomfortable in the heat all afternoon. The first four hours at 50 degrees and the second four hours at 90 degrees gives an average temperature of 70 degrees. The students were uncomfortable throughout the day in a classroom that had an average temperature of 70 degrees. 70 degrees is normally a very comfortable temperature for the environment, except the students did not experience the average temperature. The students experienced the variation around the average temperature of 70 degrees in the classroom. The variation in this example was -20 degrees in the morning and +20 degrees in the afternoon. Students in the classroom expect a temperature of 70 degrees. A variation of one or two degrees on either side is probably acceptable to the students. Beyond a few 17 degrees of temperature variation the students will start to experience more discomfort with greater variation from the 70 degree average. Having little temperature variation in this case is important to the students. Variation is a consideration in systems that complete a process more than once. Temperature variation in a class room was considered in the previous example. Other cases involve machining parts (variation of critical dimensions), preparing food (variation of ingredients), and sorting luggage going to different airliners (variation of quality) for example. These systems are all subject to variation. There are several cases of variation in delivery systems for products and services too. Two such cases are outlined in section 2.2. 2.2. Expected delivery performance and delivery performance variation Delivery variation occurs when a product or service is not delivered at the expected (or average) delivery time. A product delivered before the expected delivery time for the system is early; deliveries after the expected delivery time are late deliveries. A couple common examples of delivery variation that will help to clarify this concept are presented in this section The first example is that of a customer buying a new car from a car retailer. Customers are told to expect a new car to be delivered in six weeks (for example) when a customer orders a new car. The expected delivery time allows the customer time to sell their old car, prepare a new loan, and prepare to insure the new car. 18 What happens if the car is early? The customer probably has a used car they were targeting to sell in the sixth week, but not a couple weeks early. The customer does not likely have the financing and insurance prepared for an early delivery either. What happens when the car is delivered late? The customer has sold the used car and has to find another way to get around. The loan and insurance terms may have already started, or the terms may change based on market conditions. Either early or late delivery creates a problem for individual customers buying a car. The car retailer gives the six week estimate based on their experience of the average delivery time for all of their customers. Only some of their customers experience this average delivery time. Many of their customers experience either early or late deliveries. The car retailer thinks they are performing well based on the average, but the customers do not. The second example is for a person purchasing a house. When a person purchases a house, the person expects to close (or take ownership) on a predetermined date. Several items must be completed for the closing to take place. The loan, title, and closing documents must all be ready. Homeowners insurance needs to be in place too. The transfer of ownership will be delayed if any of these items are not ready at closing. Some closings are delayed due to one or more items being delivered late. The buyer already has money invested in other parts of the closing package that cost the home buyer when the closing is delayed. 19 On average, the loan, title, closing documents, and insurance are ready for the closing. Why are closings delayed? There is variation in the delivery for each of these items. The person purchasing the home may experience average performance on most of these deliveries, but variation on one item can cause the closing to be delayed. The loan company, title company, mortgage broker, and insurance company all deliver average performance, but the individual home buyers do experience variation around the expected deliveries. No wonder it can be so frustrating to buy a home. The variation in both of these cases comes from the system in place to deliver these products. Each system is composed of several subsystems. It becomes difficult to predict individual delivery performance when one or more of the subsystems are a large source of variation. The goal of this thesis is to analyze the variation in a production delivery system and compare it to the average delivery performance of the production delivery system. Corrective actions can be taken in the system to reduce the delivery variation performance once the source of delivery variation is determined. The customer experiences expected delivery performance when this variation is reduced to an acceptable level. The OTD system, delivery time and delivery variation, and the impact on key stakeholders are discussed further in section three. 20 3. Delivery variation background Delivery variation occurs when a delivery process is repeated in a system as discussed in the preceding section. Section three provides background on the OTD system. The meaning of delivery time and the importance to the customer is explored further. The costs for delivery variation are presented for the two key stakeholders, the customer and the enterprise. Much of the discussion in this section is from the point of view of these two key stakeholders in the system. 3.1. Three segments of the order-to-delivery (OTD) system The OTD system includes all the subsystems it takes to complete the delivery of the product. The span of the OTD system extends from the time when the customer ordered the product until the customer received the product. The OTD system completes its intended function quickly in some cases. The order is usually delivered in a few short minutes in the fast food business for example. The OTD system takes much longer to complete its intended function for some larger more complicated products. The OTD system may take several years to complete the process for an airplane order. Even though the specific systems for hamburgers and airplanes are very different, the systems have similarities at high levels of OTD system decomposition. The customer orders a product, the product is made, and then the product is delivered in both of these systems. A generic OTD system can be decomposed into three different subsystems in each of these enterprises: 21 1. Order system (Customer order to start of production) 2. Production system (Start of production to start of distribution) 3. Distribution system (Start of distribution to delivery to customer) Order-to-Delivery System Order System Figure 1 Production System Distribution System Order-to-Delivery system at the first level of decomposition Decomposition diagrams are a system tool used to break down a complex system to enable understanding of how the system is constructed. The system is placed at the zero level and each time the system is decomposed one level lower, the system is said to be decomposed to the next level. The system can be aggregated by starting at the lowest level and working toward the system or zero level. The first level decomposition for the OTD system is shown in Figure 1. The three major subsystems can be decomposed further to allow easier comprehension of the particular OTD system of interest. The decomposition should be conducted to an elemental level where the persons investigating variation in the system can better understand the complexity of the system they are a part of. What takes place in each of these subsystems? The order system starts with the interaction of the customer and the enterprise (or agent of the enterprise). The customer 22 places an order in this interaction. The enterprise then routes the order through internal channels to prepare the order for production. The process enters the production system when the order is ready for release to the production system. The product is scheduled for production, manufactured, assembled, tested, inspected, and reworked in the production system. The product is shipped and enters the distribution system once the production organization is satisfied with the product. The distribution system delivers the product to specific customers or agents of the enterprise (retailers for example) who will deliver the product to the customer. The customer receives the product they ordered at the start of the OTD process and the delivery is complete for this customer / enterprise interaction. The research in this thesis will focus on the production subsystem in the OTD system. The customers, products, and enterprises involved in the data collection are from an automobile original equipment manufacturer (OEM) and operate in a complex OTD system. Automobile OEMs are enterprises that design, develop, produce, distribute, finance, service cars and trucks for a wide range of customers. Honda, General Motors, Toyota, BMW, Ford, Volkswagen, Fiat, Kia, and Nissan are all examples of automobile OEMs. The production subsystem in this automotive case is very complex and will be decomposed two to three levels to help understand this portion of the overall system. 3.2. State of the current production system The current production system generally contains four elements at the second level of decomposition of the OTD system: 23 o Order subsystem (Production orders are received) o Build subsystem (The product is manufactured) o TIR subsystem (The product is tested, inspected, and reworked) o Shipping subsystem (The product is shipped) What occurs in each of the subsystems of the production system? The necessary materials are procured and the unit or batch is scheduled for production on a certain date and time when the order is received from the order subsystem in the OTD system. The product is then manufactured and is "completed" with the exception of tests, inspection, and rework. The unit is delivered for shipping to the customer (distribution) when test, inspection, and rework are satisfactory. The second level of decomposition in Figure 2 depicts these four subsystems. Each of these second level subsystems can be decomposed further if necessary to aid in understanding a specific production system. 24 I Order Sy stem M I Order-I-,Delivery System I Production Distribution Systerm System I -iI Production Order Figure 2 Manufacturing Build Test, Inspect, Rework Plant Shipping OTD System with production system at the second level of decomposition Delivery time in the production system is the time it takes from production order to plant shipping in Figure 2. Average delivery time for the production system is a measure of the expected delivery time based historical performance. Performance around this average delivery time indicates how much variation there is in this production system. 3.3. Critical to delivery (CTD) Critical to delivery is a term used to signify items that are critical to the customer and enterprise with respect to delivery. Delivery time (DT) and DT variation are critical to the customer for delivery in the OTD system 25 3.3.1. What is delivery time? Delivery time is the amount of time it takes to deliver the product to the customer after the order is placed in the order system. DT for the hamburger example used before is the time it takes to complete the customer order at the counter, a couple minutes. DT for the airplane example is the time in months from contract approval to delivery of the airplane. The many possible sources of variation in the OTD system are hard to visualize at the first level of decomposition. The challenge of predictably delivering a product to the customer becomes more evident when the system is further decomposed to a second, third, or fourth level. More complex OTD systems tend to have more opportunity for variation in delivery. The increased opportunity comes from the larger quantity of system inputs that have an effect on delivery variation. The following quote describes a familiar realization for someone who has been a part of a complex OTD system. Many companies have trouble delivering products on time.' Performing as expected on deliveries is difficult when we look at how hard it is to meet schedules in the order, production, or distribution subsystems alone. Each subsystem presents unique opportunities to miss a delivery expectation. The impact of one system delivering early or late is amplified by subsequent systems when subsystems are linked as they are in the serial OTD system. 26 Delivery time for the production system is the time from the production order, or production kick-off, until the product leaves the production system for the distribution system. Delivery stability and predictability are desirable, but tough to attain in the production system. The distribution system can use a standard operating system when the production system is predictable and stable. When there is variability in production delivery time, the distribution system must try to make-up for this variation in order to meet expected overall delivery time. Section 3.4 describes why understanding the OTD system and the variation in the OTD system is important for the enterprise's customers. The common customers and what they expect of the OTD system is also part of section 3.4. 3.3.2. Delivery time variation What then is delivery time variation? At the system level, delivery time variation is the difference in delivery time from order to delivery for one unit or batch when compared to other units or batches. For example consider a person buying a new boat. The customer is told at the time of order that the new boat will be delivered in 42 days (or 6 weeks). The first customer actually receives the boat in 42 days and is satisfied with the delivery. The delivery met the first customer's expectations. The next customer orders a boat and is told to expect delivery in 42 days. This customer receives the boat in 72 days. The second customer is not satisfied with delivery because the season ended and they will now pay for the boat all winter before getting to use the boat. The third customer is also told that they should expect to receive the new boat in 42 days. Only this time the boat shows up in just 12 days. Customer three is not satisfied with the 27 delivery performance of the boat enterprise. The customer has another boat they will be making payments on for the expected 42 days. They were planning to sell their current boat toward the end of the expected 42 days, but did not expect delivery in 12 days. There are three customers with average delivery at the expected 42 days, but only one of the three is satisfied with the delivery performance. The boat company (the enterprise) on the other hand believes they have performed quite well with an average on-time delivery of exactly 42 days. Figure 3 shows the delivery variation time scale as it relates to customer expectations. Deliveries to the left of the expected delivery in this figure are early deliveries; those to the right are later than expected. Early and late deliveries are both undesirable to the customer. Product Delivery Variation to Customer Expectation Early Product Delivery Late Product Delivery 0 On-Time Product Delivery Figure 3 Variation to Customer Expectation 28 The concept of delivery variation in the overall OTD system can be carried into each of the three subsystems of the process. The production system needs to meet delivery expectations of the enterprise in order for the OTD system to meet customer expectations for delivery of the product. Figure 4 shows the concept of delivery variation for the production system. Average delivery performance expectations may be met by the population distribution shown in the figure, but variation in the production system creates problems for overall OTD system performance. The purpose of this research is to develop a method to find and eliminate the sources of variation in the production system that impacts the overall OTD system. The following quote illustrates this point: Identifying and correcting the root cause impediments needs to be done in order to achieve significant improvements in order-to-delivery performance. 29 Order System Production Distribution System System Days Production Order Plant Shipping I Figure 4 Variation measure in the production system If production system variation is so easy to see, why do so many enterprises have variation problems? Many enterprises rely on average delivery performance to indicate how they are performing. These enterprises miss the point that customers experience delivery variation around the average, not the average delivery performance. 3.4. Who are the customers and what do the customers want? An enterprise may have several types of customers. The following list includes many of the more common customers of the OTD system: i Retailers 30 o End customer (Retail customers) o Commercial customers o System integrators (enterprises who purchase product to build their own product) A common customer is the retailer who buys a product or products from the enterprise with the intent to resell the product to the end customer. An example of a retailer is a clothing store that buys clothing from the producer for resale to retail customers. The end customer, or retail customer, purchases the product with the intent of using the product. The end customer does not purchase the product with the intent of reselling the product short term, although they may eventually sell the product at some salvage value in the future. We are all retail customers from time-to-time when we buy our groceries, homes, and vacations for example. The commercial customer is a large volume customer who buys the product for use in their commercial business. Many times commercial customers are fleet or large account customers. The automotive business has many examples of fleet customers for cars and trucks, such as rental car companies, taxi companies, and police departments. Automotive OEMs also have large account customers such as universities and municipalities who use cars and trucks to conduct their business. The system integrator is a customer who buys the product with the intent to use the product as part of a larger system they are producing for sale. An example of a system integrator would be an airliner manufacturer who buys engines from an engine builder, installs them on the airplane system, and sells the airplane. 31 The customers in each of these cases are very different, but have a common need for reliable delivery performance from the enterprise they buy products from. Customers simply want what they want, when they want it3. An example of this concept is found in the delivery of food products to customers. The problems in the grocery sector.... In food, the stumbling blocks are more about fulfillment getting customers what they want when they want it. In the area of fresh produce, the highest margin part of any supermarket business, delivery, is time-critical and produce is easily damaged. 4 Having the right produce on the shelf when the customer is looking for the specific item can mean the difference between sale and no sale in this delivery example. Variation in OTD systems can cause the enterprise to lose a potential sale or worse, lose customer loyalty. Jack Welch realized this at GE when he reflected on why Six-Sigma was not as successful as he would have liked. He changed the focus of the Six-Sigma effort at GE when he realized this. Jack Welch discusses this in his book "Jack: Straight From the Gut": It was Piet who came up with the answer to why our customers weren't feeling Six Sigma improvements. Piet's reason was simple: He got all of us to understand that Six Sigma was about one thing - variation! We had all studied it, including me... But we never saw it the way Piet laid it out. He made the connection between averages and variation. It was a breakthrough. We got away from averages and focused on variation by tightening what we call "span". We wanted the customer to get what they wanted when they wanted it. 32 Span measures the variance, from the exact date the customer wants the product, either in days early or days late. Getting span to zero means the customer always gets the products when they ask for them. Jack Welch was able to change the way the customer perceived delivery of GE products through this change in focus using both Six-Sigma and Lean processes and tools. The success at GE is clear in the following excerpt from Jack Welch's autobiography. We used Six Sigma and a customer-oriented perspective including span to guide us. That reduced the delivery span from 15 days to 2. Now customers really felt the improvement because orders arrived closer to their want dates.6 Other companies are seeking to give the customer this predictable delivery service like GE. Wal-Mart has led the retailing charge to give the customer what they want when they want it. Alcoa also has added a focus on the customer similar to WalMart. Alcoa General Manager, Giulo Casello, speaking of a new world headquarters site with added manufacturing facilities: "It gives us the ability to give the customer what they want, when they want it," said Giulio Casello, the division's general manager. "It really allows us to get everyone focused on the same goals."7 Some enterprises have realized that giving customers high delivery service levels is not only an advantage in the marketplace, but it is critical to survival. These 33 companies understand that in the end, customers get what they want and customers will go elsewhere if they are not satisfied. 3.5. The cost of delivery variation Costs associated with delivery variation generally fall into two categories, customer cost and enterprise cost. Customer costs are those incurred by customers that are associated with delivery time variation as the name implies. Enterprise costs are those incurred by the enterprise that is associated with delivery time variation. There are "hard" cost and "soft" cost for both the customers and enterprises. Hard costs are those that are easily quantified in real dollars. An example of a customer hard cost associated with delivery time is the cost for a customer to rent a replacement product in place of a product that is being delivered late. The rental cost is a hard cost; cash is spent on the rental expense. Soft costs are those that are more difficult to assign a dollar figure to. Lost sales due to poor delivery time performance is an example of an enterprise soft cost; customers become frustrated and buy from another enterprise. Delivery variation will create both customer and enterprise costs in the transaction in many instances. Common customer and enterprise costs are tabulated in Figure 5. 34 Customer Hard Costs o Storage space o Extra handling and product movement o Product damage and loss o Financing cost o Rental cost o Unscheduled down time (including layoffs) and overtime o Obsolescence cost o Large incoming product inventories o Duplication of costs for early deliveries Figure 5 3.6. Customer Soft Costs o Lost sales o Project delay cost o Poor customer satisfaction Enterprise Hard Costs o Storage space o Extra handling and product movement o Product damage and loss o Financing Finished goods inventory costs o Obsolescence cost o Overtime to make-up for late units o Work-in progress inventory (WIP) o Added capacity to eliminate late deliveries Enterprise Soft Costs o Lost sales o Cost of being unable to launch new products o Added system complexity to expedite orders Customer and enterprise costs System costs The customer and enterprise have similar costs depending on who is responsible for each expense per the agreement made between the two. The cost of delivery variation can be thought of as a system cost. The system cost remains whether the customer or the enterprise pays for an expense. Higher system cost is an indication of inefficiency or waste in the system. The enterprise suffers lower profits or raises the price to the customer if they incur the expense. The customer may look for a different 35 enterprise to work with if they incur too much system cost. The customer will choose to operate in a system that is more efficient and has less waste... System costs for those outlined in Figure 5 are discussed below: Storage space Storage space costs for the system can occur when a new product is delivered early and the customer does not have current storage space for the product that was delivered early. This may seem like a small problem, but think about the case of the automobile fleet customer. Finding space to store a few thousand vehicles is difficult and expensive if the automobiles show up a few weeks early. The storage space cost includes the cost to set-up, insure, and secure the storage space. Storage space rented or purchased to hold extra units can be costly when extra product is warehoused by the enterprise to protect for delivery variation. The cost can be high even if building space is available. Special racks, environmental conditioning, and warehouse staff may be needed to maintain these extra units. This was the case at Porsche in the early 1990s when Porsche was loosing money in their production operations. The production floor for Porsche looked like a spare parts storage . warehouse at the time and led to waste and inefficiency in the system 8 Extra handling and product movement Extra handling and moving cost come from the movement of the product due to the fact that it is not at the right place at the right time. An early shipment of a 53' trailer of goods to a customer many times means the customer will have to move the product out of the way until they are ready to use the product and then move the product to the 36 area it will be used. The people, facilities, and equipment necessary to handle and move the product cause expense for the customer. Extra units in the production system and frequent schedule changes cause extra handling and product movement in the enterprise system. Extra handling and movement can be an expense of the enterprise as it is for the customer. Product damage and loss Damage and loss due to theft, accidents, and unexpected natural occurrences such as storms are possible when inventories are carried to protect against delivery variation. The cost can be high when damage and loss occur. A car dealer that has an unfortunate situation where vehicles are damaged due to a hail storm will incur costs to repair the vehicles and may experience sales losses due to the damage to the vehicles. New vehicle customers tend to shy away from vehicles that were previously damaged and repaired. Opportunities for damage or loss in the production system are presented by extra handling, product movement, and storage. Work in progress (WIP) is dropped, spilled on, contaminated, and run into in workplace accidents. Damaged or lost product cause the enterprise to further vary from the delivery schedule due to the time needed to repair or remake product for the losses. The waste associated with loss and damage is costly for the enterprise. Financing cost Financing costs depend on the contract terms and how early or late the product is. Sometimes the customer pays even when the product is delivered early; other times 37 payments begin for the customer and the product arrives late. A contract that is prepaid and is then delivered late is an example of a customer who is financing a product that they do not posses. The expense can range from a small down payment to the full cost of the product. WIP inventory is many times financed by an enterprise and ties-up assets that would otherwise be used to produce more product or develop new products. Financing costs are also incurred if the facility is changed or added to in order to improve delivery variation performance. Rental cost Rental costs are those the customer pays to rent a product to use in place of the purchased product that is late for delivery. Automobile customers will sometimes have to rent a car to use for the period between selling their old car and receiving their new car. Delivery variation can create this unexpected expense. Unscheduled down time and overtime Unscheduled down time (including layoffs) and overtime may occur when a system integrator experiences early or late deliveries. Part shortages will sometimes force an enterprise who is the customer in this case, to halt their operations until the needed product is available to integrate into the system they are building for sale. The employees will sometimes be laid-off until the product is available from the supplier. Overtime begins when the product finally does arrive and the system integrator has to work extra hours to meet their product delivery schedule. Obsolescence cost 38 Obsolescence cost are realized when a particular product that has been stored in inventory to protect for delivery variation becomes unusable or obsolete. Obsolescence can be caused by changes in technology, competitive products, aging phenomenon, and a shrinking customer base. The photographic industry provides a good example of obsolescence cost. Many camera and lens manufacturers changed the lens mount architecture when they switched from manual to auto focus cameras. Camera shops (customer of the manufacturers) experienced obsolescence costs for the older systems because the technology had changed the architecture of the cameras. Photographers purchased much less of the manual focus technology and the several camera shops experienced obsolescence costs for the product that sat aging in inventory. Large incoming product inventories Large incoming inventories are owned by either the customer or enterprise. The main reason for these incoming inventories is to protect for delivery variation from the enterprise. Large incoming inventories are often used by system integrators to protect against part shortages. Inventory costs may be expected to be less expensive than halting production operations for a part shortage due to delivery variation. The inventory costs are factored into the cost of doing business. Just-in-time (JIT) delivery practices have attempted to reduce these on-site part inventories. Duplication of costs for early deliveries Duplication of cost for early delivery occurs when the current product is being paid for through the expected delivery date of the replacement product and the replacement product arrives early with payments that begin at delivery. The customer pays for both the current and replacement product in this case. A person buying an 39 automobile and selling their current automobile is an example where duplication costs may be incurred. The customer will pay duplicate costs until the older car is sold or disposed of when the replacement car arrives early and the old one is still in the customer's possession. Lost sales Lost sales occur when the product is not at the right place at the right time for the end customer to purchase. Some customers will not wait for a product from a specific enterprise if they know they can go to another enterprise to have their need fulfilled. Once the enterprise looses this customer, they incur expense to find another customer for the product that was late for delivery. The cost of loosing a fleet of product sales can be very high for expensive products like cars, large medical equipment, and airplane engines for example. Project delay cost Project delay costs are those where a major project is delayed due to early or late delivery. Many times project delays affect parts of the project that are not directly tied to the product that experienced delivery variation. For example, a building project can be delayed when the escalator does not arrive on time. The persons who were doing the escalator work are certainly affected, but so are the persons who must wait to close the side of the building that the escalator will come in through and the persons who were to lay tile once the escalator was installed. Project delay costs can become very large with only a slight variation in delivery. Overtime to make-up for late units 40 The first cost that is easily quantified is the cost of overtime worked and paid to make-up for units that are late for delivery. A common action production managers take when deliveries are running behind schedule is to work more hours. Sometimes these hours are not only worked on units that are late. Units are built early along with the late units. The enterprise is paying overtime for early units too in this situation. Many businesses do not have tight enough control on overtime to make sure the overtime is only used for late orders. Work-in progress (WIP) inventory Work in progress (WIP) inventory is also an area where delivery variation causes increased cost for the enterprise. Units are juggled in the schedule to try and meet delivery dates. WIP inventory increase as jobs are juggled in the production system. Sometimes WIP inventory is greater than the actual demand for the product. The surplus is partly caused by problems associated with delivery time variation. Cost of being unable to launch new products The opportunity cost of not being able to design, develop, and produce new products due to the losses of time and money spent on delivery time variation also costs the enterprise. New products receive less attention because scarce resources are busy working on current product delivery problems. Finished goods inventory The enterprise will build a large (sometimes expensive) finished goods inventory in an attempt to always meet customer demand. The inventory also creates problems with storage space costs, obsolescence costs, handling and product movement costs, 41 and damage and loss costs similar to those experienced by the customer. Finished goods inventories are managed by the enterprise to provide the customer with fast, reliable service. Added capacity to eliminate late deliveries A cost the enterprise experiences is the cost of adding capacity to the production system to meet delivery expectations. Managers of production systems that consistently miss delivery targets will many times believe the problem is a lack of capacity and will try to solve their problem by adding capacity. It is the production system architecture and the process that is the problem, not the size or capacity of the system. 3.7. Motivation to drive change in the system The motivation to reduce delivery time variation for the customer and the enterprise is illustrated through this discussion of costs and expectations. Who bears the increased cost is really irrelevant because the system cost will likely be higher and the system will likely be less efficient when there is more delivery variation. The hard part now is to understand how change can be made in the production system. An influential change that can be made is to base the objectives and rewards of those managing the production system partially on a delivery variation target. Managers will resist change if their internal objectives do not align with customer expectations. Managers reach high levels in their organization because they have delivered on key objectives they were measured against in the past. Managers deliver on these 42 measured objectives for personal benefit even when the measurements do not drive behavior that benefits the customer and enterprise: For instance, if the success of the shop floor is measured simply by efficiency, utilization and/or standard hours of output, you can be sure that parts will be produced even when they are not needed. The result: too much inventory of unneeded material and possibly shortages of what is needed.9 Reduced inventory and improved customer service level are mutual goals. The system costs for the customer and enterprise are reduced when both work to reduce the inventory in the system and eliminate other system costs incurred due to early or late delivery. In the past, it was thought that large inventories need to be in place to meet high customer service levels. Now the enterprise metric to drive low delivery variation and low inventories aligns with the customer expectation for better delivery performance. 43 44 4. Methods The data in this research is collected and analyzed using the Six-Sigma DMAIC methodology as a framework and tools from both Six-Sigma and Lean to perform the actual analysis in section five. The purpose of section four is to give the reader a background in the methods that will be used in section five to analyze the automotive OEM case data. 4.1. Six-Sigma and DMAIC Six-Sigma is a methodical process that uses several different tools to improve processes and systems. The goal of the Six-Sigma process is to make a measurable improvement to a process, system, or product and to maintain the improvement long term. Some of the tools used in the Six-Sigma process are familiar to those in different industries where methodical problem solving tools have been used in the workplace. Many of the statistical tools used in Six-Sigma are new to those who are trained in the Six-Sigma process. Persons who are trained in the use of these processes and tools are given a designation that indicates their ability to apply both the method and tools. The three levels of certification are for Green Belts, Black Belts, and Master Black Belts. The Green Belt is the first level in Six-Sigma ability. Green Belts receive general training that allows them to apply simple tools to their day-to-day job.10 Black Belts have more in depth training and apply the Six-Sigma tools and processes to more difficult problems as their primary function." Master Black Belts complete additional training in Six-Sigma 45 methods and tools after they have already been certified as a black belt. In addition, Master Black Belts deploy training for the enterprise in Six-Sigma, act as a champion for Black Belt and Green Belt Projects, and define projects with key enterprise leaders.1 2 The Six Sigma breakthrough strategy can be applied at different levels in an organization and will yield different results. Six-Sigma tools and processes can be used at the business level, the operations level, or the process level. A project to reduce the variation in delivery time in the production process takes place at the operations level. Jack Welch used Six-Sigma at GE at all three levels and talks specifically about the operations and process levels in his autobiography: Plant managers can use Six Sigma to reduce waste, improve product consistency, solve equipment problems, or create capacity.1 4 The Six-Sigma process is broken down into steps known in the Six-Sigma community as DMAIC. The Acronym stands for: u D: Define L M: Measure u A: Analyze Li 1: Improve L C: Control The DMAIC process takes a problem all the way from problem definition through improvement and control of the process. The DMAIC process is described in the next five sections. 46 4.1.1. Define phase The define phase of a Six-Sigma project is where the team working on the problems or opportunity answer a few simple questions: u What is the problem or opportunity? Li What is the defect? L Who are key stakeholders?1 5 o What is the goal? The team will use a few documents to help answer these questions at the start of every project. A business case is used to help understand how the problem or opportunity impacts the business financially1 6 . The team will document how much the problem currently costs, what they expect to gain by improving upon the problem, and what they expect to spend fixing the problem. Once the business case is complete and the team decides with the approval of the appropriate enterprise leader, the team will develop a charter that includes the business case, a statement of the problem, and a statement of the goal1 7 . The team will develop a high-level process map and a list of what is important to the customer1 8 from the charter. The list of what is important to the customer may focus on quality, cost, function, delivery, or other characteristics of the enterprise / customer relationship. The charter for a delivery variation project would include a statement about delivery time variation, how it impacts the customer and the enterprise, and what the associated costs are. 47 4.1.2. Measure phase The measure phase uses the stakeholder analysis, charter, and high level process map from the define phase to aid in selecting what is critical to the customer. The critical characteristic list may include CTQ (critical to quality), CTC (critical to cost), and CTD (critical to delivery) characteristics' 9 . The critical characteristics in this research of delivery time variation in a production system are CTD and CTC; critical to both delivery and cost. The next step is to contemplate how best to measure and then develop a measurement for these characteristics. Sometimes there are existing measurement systems to measure these characteristics, other times the team has to develop a way to measure the critical characteristics. Performance standards for the measurement should be validated if they already exist, or developed if they did not20 . Performance standards can be thought of as a specification for the measurement the team is using. The specification is likely to have a target with some tolerance band around the target. The performance standard for delivery variation will be the allowable days of variation for the particular OTD system (these standards will vary depending on the customer and the product). The measurement system is then validated2 1 for the measurements that are chosen to check the quality of the data used to determine the performance of the current process. The analysis that is used will depend on the type of data being collected (variable, ordinal, or attribute) and the operational definition of the measurement system. The measurement system analysis (MSA) will check how well 48 the measurement system is able to discriminate the variation in the production system as compared to the variation in the measurement system. The risk of not validating the measurement system is that the team may not be able to quantify the difference if the measurement system contains too much error. Baseline data on current system performance can be collected once a measurement system is in place 2 2 . The baseline data defines the number of days of variation for a certain percentage (P) of the population in the case of the production system and delivery variation. An example of this performance measure is shown in Figure 6. Initial Delivery Time Variation Capability 0 0.9 - -5 0.80 x o0.7- E0.6- 0 0.5 * 0.41 0.3SPAN E 0.2U0 0.1 00 0 0 5 0* 15 10 Days 20 14.5 days (P = 9Eno) Figure 6 Example of an initial OTD delivery variation plot 49 25 30 Figure 6 is a graph of delivery dates for a large number of units from a production system. The definition of this measurement system requires that delivery variation is determined when the top and bottom 2.5% of the population are eliminated from the measure. The 14.5 days of delivery variation in this example are measured after the elimination of the top and bottom 2.5%. Said differently, 95% of the population has a delivery variation of 14.5 days as measured by the span. The baseline data will help the team understand the problem better. The team can develop a more focused problem statement with the knowledge they gain from collecting the baseline data2 3 . The team will also evaluate whether the current data . pinpoints problem locations or occurrences24 4.1.3. Analyze phase The analyze phase of the Six-Sigma process is where the function Y = f(X) is explored in detail. The Y in the equation represents the output and the f(X) represents the function of the inputs. Another way to say Y = f(X) is the output is a function of the inputs. The team develops a list of possible X's, and then works to build a relationship between the important inputs (important Xs) and the output (Y). The idea is to find the sources (inputs) of variation that are causing variation in the output 25 . Another way of thinking of the analyze phase is to identify root cause(s) and then to confirm them with data26 . Once the important Xs or causes are known (with statistical validation), the system can be changed and the effect evaluated in the improve phase. 50 The possible causes or inputs in the production system of the OTD System include: ... acquiring materials, scheduling production and other information are often big contributors to overly long order-to-delivery cycle times.2 7 Data mining such as regressions and variation analysis are often performed to help define the critical few inputs to the system that control the output. Multiple outputs should be evaluated at this point in the analysis and again in the improve phase. Each input may affect one or more outputs. It is important to evaluate other outputs with data or engineering judgment to ensure the customer and enterprise do not get an unexpected result in another output when changing one of the inputs. Other outputs that are commonly important in the production system are cost and quality. 4.1.4. Improve phase The improve phase starts once the source of variation is known. The team will develop and test changes to the system or process to validate the desired change in the output. The team tests and implements solutions that address root cause in this phase2 8 . The team will also develop new operating tolerances for the improved system 29 . The team will asses the variation in the improved system to see if the new performance meets stakeholder expectations. 51 In the case of the OTD system, we will look to see if the delivery time variation has been reduced. We will do this at the system level and at the subsystem level for all subsystems that were changed. 4.1.5. Control phase The control phase starts by measuring the new process performance (once the measurement system has been revalidated with the new process parameters) 30. The new process performance is evaluated to see if the performance improvement meets project performance targets (and the project is closed) or if additional changes will need to be made to meet the performance targets of the project. 52 Improved Delivery Time Variation Capability 0 4 0. 0.90.8- 0.0.7S0.60 0.5a 0.4- 0.3E 0.2- SPAN - %U0 .1 * 0.0 0 5 15 10 Days Figure 7 20 25 5.7 days Example of an improved OTD variation plot as compared to the Measure Phase Delivery variation after improvement to key inputs is shown in Figure 7 for the example started in the measure phase. The variation for 95% of the population was reduced from 14.5 days to 5.7 days by finding and eliminating sources of delivery variation in the production system. The methods and tools used to improve this example are the same as those employed in the actual automotive OEM case used in this research. Process controls and a monitoring system are then put in place to make sure the gains that were achieved by the team are maintained3 1 . A key portion of the process 53 control plan is to make sure the process owner buys into and owns the maintenance of the change. The project results are documented, learning is shared with other organizations that might benefit, and recommendations are made for future work in the project area. 4.2. Lean principles Using Lean principles in combination with the Six-Sigma process as discussed in the preceding sections provides a systematic method to find and eliminate waste associated with delivery variation. Waste elimination and value stream mapping are respectively the Lean principle and tool combined with the Six-Sigma method to provide better delivery variation performance and improved customer satisfaction. 4.2.1. Elimination of waste Waste elimination is at the very center of Lean enterprise principles. The elimination of waste in the system allows the enterprise to perform its intended function with a minimum of energy and resources. The Lean enterprise has an advantage over it competition because it has sought and eliminated waste from its system. The concept is highlighted in the book Manufacturing Operations and Supply Chain Management - The Lean Approach3 2 The understanding of what waste is and how to remove it is fundamental to the creation of a Lean enterprise or supply chain; however, many managers and companies often do not understand or realize the importance of the concept. 54 Types of waste are often categorized or classified to simplify the understanding. A popular waste classification for a production system was developed by Toyota. Toyota classifies wastes in each of their production systems in the following seven categories: Seven wastes identified in the Toyota Production System (TPS) 33 : " Overproduction " Waiting " Transportation o Inappropriate processing " Inventory " Motion " Defects Waste can be identified and eliminated. An example of transportation waste in the TPS classification would be when incoming parts are stored and then moved to the operation. Extra movement of parts occurs. The elimination of the waste would be to find a way to move the necessary parts directly from the incoming supply truck to the operation. A system dynamic occurs between waste and variation when we consider product delivery 34. Waste in the production system causes variation in delivery from the production system. The variation in production delivery then causes additional waste for the customer or enterprise. The external waste for the customer or enterprise 55 sometimes comes full circle and causes waste in the production system. This dynamic is depicted in the causal loop diagram in Figure 8. Production 00000 Production System system Waste Delivery Variation External I Customer or Enterprise Waste Figure 8 System dynamic waste / variation loop The system dynamic becomes clearer with an example. An automobile production system contains multiple reworks for quality defects (waste - defects) that cause variation in delivery of a vehicle to a customer that buys hundreds of vehicles each year. The customer has to adjust (waste - waiting) their planned use because the vehicle is delivered late. The customer asks for the next vehicle to be delivered earlier 56 than actually needed to allow time for late delivery. The customer is gaming the system. The enterprise experiences waste in their production system when they pull one vehicle from the build schedule (waste - inappropriate processing) and replace it with this new customer order. The vehicle that is pulled is now subject to delivery variation due to waste. The waste / variation loop is real and costly for both the customer and the enterprise. Section 4.2.2 discusses a valuable Lean tool called a value stream map to help identify the waste in a system. The mapping tool with the principle of identifying and eliminating waste are two powerful yet simple elements of a Lean enterprise. 4.2.2. Value stream mapping Value stream mapping is a simple to use Lean tool employed to help identify the process flow and resulting waste in the system. A values stream map is simply a map that depicts the current process including any rework or handling for the current production system. The detailed map is completed by a team of people involved with the current process so as to reflect the reality of the current system. Elements of the system are next labeled as value added, non value added, or non value added, but necessary for the enterprise. Value added steps are then identified as those that add value from the customer point of view 35 . These may be items that add form or function to the product or are steps that the customer is willing to pay for. Non-value added, but necessary steps are also identified. These are steps that are not of direct value to the customer, but necessary for the enterprise. An example of 57 no-value added, but necessary are steps completed to fulfill regulatory requirements. The last category identified is the steps that are truly non-value added. The non-value added steps of interest in the production system concerning delivery are those that cause variation in the delivery system or delay expected delivery. Non-value added steps of this type are targeted for elimination to improve delivery performance. 4.3. Integrating Lean and Six-Sigma to find sources of variation and waste The combination of Lean principles and tools and Six-Sigma methods is useful in quantifying variation in delivery performance. System decomposition diagrams are utilized in constructing the process map for a given system (a production system in this case). Six-Sigma measurement systems are then set-up to measure the delivery performance at key points along subsystems boundaries. The data from these measurements enables analysis that indicates the source of variation. Waste is often found to be this source of variation. The combined method outlined here will be used in section 5 for an automotive OEM production system. The purpose of the case is to show how subsystem mapping and measurement systems (or the integration of Lean and Six-Sigma) can lead to a strategic delivery performance advantage for the enterprise and better customer satisfaction for delivery of products. 58 5. Automotive production system analysis Some of the key steps in the data analysis as well as the actual data for the automobile OEM production system being studied are presented in section five. The data has been normalized due to the confidential nature of the data in the competitive automotive industry. The production system is described using system decomposition diagrams and process maps. The system is analyzed using actual normalized data from an automotive production system. 5.1. Automobile OEM production system description The scale of an automobile OEM production system is helpful to understand in looking at the specific case. The magnitude of the production system being studied in the case is typical of several OEMs and is represented by the following: * 200,000 to 300,000 vehicles per year are produced * Vehicles sell mostly in the $20,000 to $30,000 range * The manufacturing facility encloses one to two million square feet * The work force numbers 2,000 to 3,000 * 1,000 vehicles might be produced on an average day * Thousands of parts are assembled to create the finished product L Annual sales are in the $4 billion to $9 billion range depending on sales volume and the average selling price u Profits may range from $80 million to $600 million depending on sales and cost U Customers include dealerships (retailers), retail customers (end-users), fleets (commercial), and system integrators 59 The opportunity to improve profits by eliminating waste is expected to be very large for the production system described above. Waste elimination occurs through delivery variation reduction in this case. A small change in per vehicle cost performance can make a big change in profitability for the OEM. Saving four to five dollars on each vehicle sold can translate into a million dollars in increased profit. The cost to the customers can be significant as well. A customer who depends on timely delivery of vehicles can incur tens of thousands of dollars in cost when their operations are interrupted due to delivery variation. 5.2. Process maps and system decomposition Process maps are used to help the researcher and the research team better understand the system they are working on3 6 . Process maps have a rational boundary for which items not included will be considered outside influences on the system 37. A boundary is chosen to include items which are likely active elements of the system and exclude elements that are outside the system being considered. An example of an element inside the production system is the assembly line. A railway subsystem that delivers vehicles would be outside this production system boundary, but would be inside a rational boundary for the entire OTD system. The process map is completed by the most experienced and knowledgeable persons working in the area of interest, in this case the production system. The maps used for this production system are of the current state, but maps can also be used to depict the future state as well. 60 5.2.1. Order-to-Delivery process map The first process map developed and used is the high-level process map for the OTD system shown in Figure 9. The system decomposition diagram developed in section three was used to aid construction of the process map. The process starts with the customer want and customer order. The process goes through order processing and into the production system from there. Once the product completes the production process, it enters distribution and the overall cycle is completed when the customer receives the product. Customer Customer WantOrder from Comapny Order Processing CustomerI Plant Order Distribution +Plant Fulfilled Fuliled ~ Shipping Manufacturing Plant Order Manufacturing ________ Figure 9 Overview of the automotive Order-to-Delivery system 5.2.2. Production process map Although the overall process map helps us understand how the production system fits into the OTD process it does not provide an understanding of the production system. More detailed process maps have been developed for the production system at 61 further levels of decomposition. The high level production process map for the automotive production system being studied is shown in Figure 10. Manufacturing Assembly Line 1 Production Order Manufacturing Assembly Line 2 Figure 10 Test, Inspection, Rework Plant Shipping High level process map of the overall automotive production system from the time the plant receives the order until the order is shipped The process starts by receiving the order in production. The product then goes through the build process before entering the test, inspect, and rework (TIR) subsystem. The product is ready for distribution once the product completes the TIR processes. The high-level process map leads to a natural segmentation at subsystem boundaries for measuring system performance. The first task will be to determine what part of the system the variation comes from. Three segments were measured to better understand this variation: o Order (variation from production order date to the start of the build) o Build (variation from the start to the end of the build) o TIR (variation from the end of build, through test, inspection, and rework, until the product is shipped) 62 A measurement system was needed to help quantify the system variation and the variation in these three steps once these measurement points were determined. 5.3. Delivery time variation measurement system The measurement of delivery time is necessary to allow analysis that leads to the elimination of the waste that causes the variation. The measurement of delivery time implies that the person collecting the data will gather data for individual units or batches at the start and again at the end of the process. The date the order is received and the date the order is shipped are these endpoints in the automotive production system. The data that is collected in this case is the date, time, and individual identification number (or specific ID) of the unit being ordered and produced. Each unit is measured for delivery performance in the automotive production system. 5.3.1. Measurement system operational definition Collecting the date, time, and specific ID for each unit in this case is completed with the help of information technology and barcode technology. The date information for the order is recorded and stored on a central database when it is received from the upstream OTD order system. The vehicle builds start at some point after this order date. The unit is built and the barcode is read at the end of the process. The vehicle specific ID is read from the barcode when the unit is shipped. The barcode for each unit is read by a barcode scanner. The data is then sent to and stored in a central database with a common or central clock. 63 These two measures provide enough information to determine the production delivery time for each unit. Several units for a particular order date can then be combined to determine the delivery variation for orders placed on a particular day. The delivery variation for a day is depicted in Figure 11. The population is rank ordered from the shortest to the longest delivery time. The variation for 95% of the population is then determined as the difference between the value for 97.5% of the population and 2.5% of the population as shown in the figure. Why use 95% of the population instead of 100%? The correct population threshold will be determined based on an understanding of the current enterprise and the goal for delivery variation reduction. In a business where every unit must be delivered on time, 100% would be used. 90% or 95% is used in a business where there are five or ten in 100 exceptions that enterprise leaders are willing to accept. Using 100% in these cases would tend to skew the data to exaggerate the magnitude of the delivery variation. The population threshold level is something that the team will have to set and was chosen to be 95% in this automotive OEM case. 64 Product Delivery Variation 95% 1.0 p (Cumulative %of the Population) Days of 2.5% 2.5% Delivery Variation (SPAN) 0.0 Product Delivery (Days) Early Product Delivery Figure 11 Late Product Delivery Days of delivery variation (SPAN) for 95% of the volume The team was also able to measure important subsystems of the production system with the same type of measurement system and the same data analysis as shown in Figure 11. Measure the overall production system process performance and compare that to the performance of the subsystems at the next lower level to determine which subsystem has the most opportunity for improvement. Working on improvements without this type of analysis may mean the team works hard to improve the system by . working in the area with the least opportunity for improvement 38 65 5.3.2. Measurement system process map The process map for the measurement system at this 2 nd level of decomposition in the production system (the production system is at the 1st level of decomposition in the OTD system) is shown in Figure 12. Manufacturing Assembly Line 1 Production O r d erP Manufacturing Assembly Line 2 Order to Build Build Test, Inspection, Rework l n Shipping Test, Inspection, Rework Production Process Figure 12 Measurement system map The time measurement for each unit from order to shipping (the overall production system) is shown in the figure with the three subsystem measurements depicted above it. The first measure is from the order until the barcode scanner at the start of production reads and records data for the unit. The second subsystem is from the start of production barcode scanner to the end of production scanner. The last 66 subsystem is from the barcode scanner at the end of production to the time the unit is shipped. The measure shows how long the TIR subsystem took to complete for the unit. Each of these subsystems can be further decomposed to a more elemental level. The subsystem with the most variation is chosen for further decomposition. Working on the subsystem with the most variation will allow resources to improve the production system performance by focusing on a smaller subsystem. 5.4. Current performance Current performance for the production system is determined for each day of orders. Figure 13 depicts the measurement for a single day of production. The average delivery time is the average for all 1000 units ordered on this day. The variation for 95% of the population is measured to be eight days as shown in the figure. 67 Variation from Manufacturing Plant Order Date to Actual Ship Date (one ay - 1000 units) 0 0 1 - 0 0 0 - - - - P - - - - - 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 - 0 0 000 0 5 10 5 ( 0d s (P = 95%) Figure 13 I I I I 20 25 30 35 40 Days Variation from manufacturing order date to actual ship date for one day of production system orders (short term variation for the entire production process) 8 days The delivery time average and variation for the overall production system were then collected over 100 days (around 100,000 vehicles). Variation and average values were determined for each day of orders and are plotted in Figure 14 and Figure 15 respectively. Both the average and variation in the system change from time-to-time depending on how well the system is running. How well the system is running depends on the performance of the three subsystems. These overall performance plots do show that there is variation in the system, but do not lend insight to the problems that cause 68 the variation. The problems will start to surface as the analysis goes deeper into the subsystems. Delivery Variation 120 |-Overall VariationI - 100 U, - 80 60 - 0 Cu Cu .4-a 4-a 40 - Cu U, 20 IVA 0 0 20 40 60 80 100 Order Day Figure 14 Overall production system variation for 95% of the total population over 100 days of production orders 69 Average Delivery 120 -Overall 100 _ Average - 80 CU - 60 G) 40 - CU) - 20 0 0 20 40 60 80 100 Order Day Figure 15 Overall delivery average (expectation) for 95% of the population over 100 days of production orders Three spikes in the data in Figure 14 are from periods with an assignable cause. The cause of the three largest spikes is known (assignable) and will not be discussed here. These spikes can be removed from the data when analysis of common cause variation (normal to the system) 5.5. Cost associated with the current performance Variability in the current production system leads to system costs for both the customer and enterprise. A conservative estimate of the cost of delivery variation for the 70 enterprise is to look at the cost of inventory in the production system. Inventory level data is compared to the amount of delivery variation to get an idea of the increase in inventory cost for larger delivery variations. The data is normalized by giving the lowest inventory order date a value of one and then plotting relative inventory levels for the other order dates. The relationship of cost and delivery variation for this system is shown in Figure 16. There appears to be a relationship between inventory cost and delivery variation in the graph. More data points would be needed for days with high variation before we can determine if the relationship is statistically valid. This cost estimate is very conservative compared to the actual costs considered in section 3.5 on the costs of delivery variation. Relative Cost of Delivery Variation 5 4- 12- s . 0 0 10 20 30 40 Delivery Variation Figure 16 Relative inventory cost for delivery variation 71 50 60 The non-value added cost from the Lean manufacturing perspective is much higher than the inventory cost alone. Other costs associated with delivery variation include labor, material, facility, and obsolescence costs. The magnitude of these costs can be in the millions to tens of millions (dollars) for an automobile OEM production system. The actual numbers for these costs are hard to determine because many of the costs are hidden in the normal operating cost for the production system. 5.6. Analysis of variation in the current system We analyze variation in the order, build, and TIR subsystems to determine what causes the variation in the production system. The order, build, and TIR subsystems are analyzed for each day using the method outlined in the operational definition for the measurement system. The examples that follow for each subsystem are made from data collected on the same order date. The variation in the production order system was determined for each day of orders as depicted in Figure 17 for 95% of the population. The variation for the order subsystem on this day was 1.7 days. The order variation was determined for all 100 days in the data set. The average order time was also determined for the data set. 72 Variation from Manufacturing Plant Order to Build Start Date (one day t 1000 units) - *0 * o 0 1.0 0.90.8- - - 0.7 0.6P0.50.4 0.3 0.20.1 - A 0.0 0 Figure 17 2 41 1.7 days (P = 95%' I____________ 3 8 10 12 14 Days Variation from manufacturing order date to start of production date for one day of production (short term variation for the order subsystem of the production system) The variation for the build subsystem was likewise determined and is depicted for the same day in Figure 18. The average and variation in the build subsystem was analyzed for all 100 order days too. 73 Variation from Build Start Date to Build End Date (one day - 1000 units) 1.0 0.9 VO 0 P O 0 0 0 0 O 0 - - 0 0.8 - 0.70.6P 0.50.40.30.20.1 0.0 r- 0 Figure 18 ----- lu I 2 IIII 8 4 6 2.6 days (P = 95%) 10 12 14 16 Days Variation from start of build date to end of build date for one day of production (short term variation for the build subsystem of the production system) The variation in the TIR subsystem was also determined for the 100 days of data. An example of the variation analysis for 95% of the population is shown in Figure 19 for the same day as the order and build subsystems. 74 Variation from Start of Test, Inspection, and Rework Date to Ship Date (one day - 1000 units) 0 00 00000 p 0 10.90.80.70.60.50.40.3 0.2 0.1 0 5 10 15 25 30 35 Days (P = 95%) Figure 19 20 Variation for the test, inspect, and rework subsystem prior to the actual ship date The variation for all 100 days was then plotted as a run chart. The run chart in Figure 20 shows how the variation in all three subsystems compares to the overall variation in the production system. Relative variation was determined by giving the lowest variation a value of one unit on the chart. The run chart shows that the test, inspect, rework subsystem is consistently the subsystem with the largest variation. The amount of TIR subsystem variation normally approaches that of the overall system. The TIR subsystem will be the focus for efforts to reduce delivery variation in the production system. Working to reduce delivery variation in the TIR subsystem will enable the biggest delivery variation performance improvement in the production system. 75 Delivery Variation 120 --Overall Variation ------ Order Variation --- Build Variation - - TIR Variation 100 - - .2 80 60 40 - 40 > CU - 20 0 20 40 60 80 100 Order Day Figure 20 Delivery variation performance for the entire system and each of the three subsystems at the first level of decomposition Periodic shapes are observed in the order and build subsystem variation run charts. The periodic nature of the delivery variation chart for the order subsystem comes from the method used to process orders. The periodic nature of the delivery variation for the build system comes from the length of the build process and where certain vehicles are during non-production periods like weekends. The run chart for average delivery performance does not indicate that the TIR subsystem is a problem for delivery within the enterprise. Figure 21 is the run chart for the 100 days of data used in this analysis. More resources have traditionally been 76 focused on the order and build subsystems due to the enterprise focus on average delivery performance. The TIR subsystem normally has the smallest average delivery time impact; the exact opposite conclusion from what we saw when studying variation at the subsystem level. Average (Expected) Delivery 50 -Overall 45- ----- Order Average 40 --- Build Average - - TIR Average Average - a>35 - 20 -25 CU-20 - 15 0 0 20 40 60 80 100 Order Day Figure 21 Average delivery performance for the production system and each of the three subsystems at the first level of decomposition The study of variation at the subsystem level has led us to understand that most of the variation is coming from the TIR subsystem. The build and order subsystem are more interesting to the enterprise for their average performance impact, but individual customers care about the variation performance in the TIR subsystem. 77 The subsystem variation data was also used to see if subsystem variation is related to variation found in prior subsystems. The 100 days of data was used to check for correlation between the order and build, order and TIR, and build and TIR subsystems. No relationship was found in this analysis as depicted in Figure 22 - Figure 24. Correlation Between Order Variation and Build Variation 7 1 6 5 R2 = 0.1716 0 CU 4 .r CU 3 -0 -- -. S e 2 1 0 -r 0 2 4 6 Order Variation Figure 22 Correlation between order and build subsystem delivery variation 78 8 Correlation Between Order Variation and TIR Variation 60n - 50 * - c 40 0 4-0 - 30 R 0.0054 01. - 20 0 2 0 10 i II 0 2 4 6 Order Variation Figure 23 Correlation between order and TIR subsystem delivery variation 79 8 Correlation Between Order Variation and Build Variation - 60 * a - 40 . 0 50 0 0 0 30 - CU R2 = 0.0005 - . 20 .0 *0 0 00 0 . .0 S I. 0 U 0 1 . 0jW ' -. - 10 * .. -. 2 3 0 4 5 6 7 Build Variation Figure 24 5.7. Correlation between build and TIR subsystem delivery variation TIR subsystem possible sources of variation The variation analysis at the production subsystem level indicates that the source of much of the delivery variation in the production system is in the TIR subsystem. The next step is to perform variation analysis on the TIR subsystem. The steps in conducting the analysis start much the same as the production subsystem analysis. The TIR subsystem is first decomposed and mapped to enable an understanding of complexity in the subsystem and to show possible sources of variation. 80 5.7.1. TIR subsystem decomposition The TIR subsystem consists of several smaller elements at the next level of decomposition. The smaller elements fall into four categories: u Vehicle Test L3 Vehicle Inspection L Vehicle Rework L Vehicle Inventory There are five different vehicle tests in the subsystem that range from simple subsystem functional tests to full vehicle operational tests. The time to complete the tests is standardized and is fairly short. Every vehicle completes all five tests and is either passed to the next point in the process or sent for further evaluation or repair (rework). Examples of these tests that are common among OEMs are dynamometer roll system and functional tests, water leak tests, and electrical system tests. The 12 different inspections in the system range from static craftsmanship evaluations to dynamic evaluations on the road. Some of the 12 inspections are standard for every vehicle. Other inspections are conducted on an audit basis, so only some of the vehicles produced each day go through the evaluation. The vehicles that do go through the non-standard evaluations tend to stay in the system longer and therefore may be a source of variation in the TIR subsystem. The vehicles that pass each inspection are sent to the next step in the process. The vehicles that do not pass are sent for rework. Examples of inspections that are common among OEMs are body gap and flushness inspections, suspension alignment audits, paint finish inspections, and 81 fluid level inspections for engine oil, transmission fluid, cooling fluid, steering fluid and brake fluid. Seven different rework areas are present in the TIR subsystem. Each rework area consists of operators with specialized skills and tools to perform certain types of repairs and rework. The numbers of repair persons and spots for vehicles to be reworked depends on the service area. Almost all the service areas have an inventory that feeds the service area. The inventory and variation in the extent of the repair in the rework area are a possible source of variation. Rework completed in an OEM production facility will include paint blemish repairs, damaged wire repairs, powertrain repairs, and fluid level repairs to name a few. The 13 inventory areas in the TIR subsystem feed both standard and nonstandard test, inspection, and rework areas. The inventory is handled on a first-infirst-out (FIFO) or last-in-first-out (LIFO) basis depending on the arrangement of the storage area and the work practices of the person delivering a vehicle to inventory or taking a vehicle from inventory. Inventory can be a source of variation in the TIR subsystem, especially in areas that practice LIFO inventory control. LIFO causes the first vehicle in to be the last vehicle out. There is delivery variation between the first and last vehicles brought into inventory. LIFO inventory control practices depend on the arrangement of the storage area and the work practice of the employees in the area. Sometimes the conversion from LIFO to FIFO can be as simple as teaching the workforce about the importance of FIFO vehicle handling. Other times the change can be expensive because facility changes 82 must occur to enable FIFO inventory control. The cost of variation must be weighed against the cost of changing the facility if this is the case. The decomposition is at the third level of the OTD system and the first level of the TIR subsystem. Grouping the elements for the TIR subsystem in this way gives consistency of level in the analysis. The analysis is most easily conducted when the elements we are studying are similar in the subsystem. We want to compare one inventory area to others, or to test, inspection, or rework areas. The decomposition helps us avoid studying elements at this level with elements one level up or down in the decomposition. The analysis might yield misleading results if we were to compare an inventory area's performance to an individual work element conducted in a repair area. 5.7.2. TIR subsystem process map The next step is to map the elements from the decomposition diagram to better understand the subsystem and the interactions between elements. The process map for the TIR subsystem is shown in Figure 25. The map in Figure 25 shows the relationship between the four categories of elements. Inspections three through 12, inventories six through 13, and rework six and seven are shown as one process block to simplify the map for illustration purposes in this thesis. The purpose of the map is to show how complex the TIR subsystem is between the completion of the build and the shipping of the vehicle. The process map is used to help identify possible sources of variation in the TIR subsystem. 83 Build cormplete Inerbxy 2 -b Invntory 3 Test 1 Test 2 Test 3 Rewkork 2 Irertory - Revrk i inventcry 4 Irspect bn I F rocess map a ggregated at this point Test,4 ]nspectin 3-12f Irentbry 6-13/ Irvenbory 4 - Irspectkon 2 Rework 6-7 Wend to Dis ribution System Figure 25 Inrerery 5 Test 5 Rekrk 3 Process map for the TIR subsystem (some steps aggregated to simplify the process map) 5.7.3. Cause and effect diagram for the TIR subsystem A cause and effect diagram is a tool used to help gather ideas about what possible causes might contribute to a certain effect. The diagram is constructed by placing the effect with groupings of possible causes pointing to the effect. There are standardized groupings of causes such as manpower, methods, and machines, but 84 other groupings can be used if they make more sense in the system effect being studied. With the process map as a guide, a cause and effect diagram can be constructed to list the several possible sources of variation in the TIR subsystem. Each of the elements in the TIR subsystem decomposition are listed here as possible sources. The groupings for the possible causes follow the element types from the TIR subsystem decomposition. Other possible sources of variation can be added by the team as the team develops new ideas about the causes of the delivery variation. A generic cause and effect diagram for the system of interest is pictured in Figure 26. 85 Cause and effect diagram for TIR sub-system Test Rew ork Rework 1 Test 1 Rework 2 Test 2 Rework3 Rework4 Test 3 Rework 6 Test 4 Rework 6 Rework?7 Test S 4 entory13 hvertory12 Inantoryl11 hspeodon 12 hspeodon 11 hspecion 10 h rpedon 0 hspeoiion 8 hspection 7 hspection 6 hspecion 5 hspection 4 hspection 3 Vehicle Deliver/ Variation ertrv hertory 1 hentory hktory hnrtory hxertor5 n nmntry4 hwetry3 tnertcor2 tuertorvl hspecdion2 ispelon 1 Inventory Figure 26 ispect Cause and effect diagram for the TIR subsystem 5.7.4. C&E matrix Searching for the largest sources of variation from the 37 possible sources listed in the cause and effect diagram would take a long time and a lot of resources. There is a Six-Sigma tool called a cause and effect matrix that will help narrow the search to only the most likely sources. Studying a few likely sources is a better use of scarce resources and takes less time to complete. Studying sources that are not likely would 86 be a form of waste and would contradict the very Lean principle the analysis is working to eliminate. The cause and effect matrix works by listing the possible causes and related effects and weighting each cell in the matrix to determine the most likely causes to include in the analysis. The possible causes are called inputs and the effects are called outputs. The nomenclature comes from the fact that we are trying to understand the transfer function between the inputs and the outputs. This is often written as: Y = f(x) or the output is a function of the input Notice we are working with effects (or outputs) in the cause and effect matrix. This is because we want to make sure we understand all the important effects that may be impacted by changing one or more of the possible causes (as discussed in the section on Six-Sigma methods). The outputs are placed along the top of the matrix and each output is given a score from one to ten based on the relative importance of the output (a score of ten being most important). The inputs come from working with the decomposition diagram, process map, and cause and effect diagram. Each input is then given a score for the likelihood that the input would affect the output at each intersection in the matrix. The cells are multiplied and added across the input row to give a total score for the input. The inputs with the highest score become the inputs of most interest for the initial study as seen in Figure 27 The scores for each cell are based on the expertise and opinions of those on the team. It might not seem very scientific, but usually yields a list of likely causes that will be studied first. The short list of causes is easier for the team to work with than a longer 87 list that may range from 30 to 100 inputs. The next set of possible output variation causes can be selected for study if the first set is not determined to be the majority of the cause when the initial analysis is complete. The cause and effect matrix (C&E matrix) for the TIR subsystem is shown in Figure 27. The outputs for the C&E matrix include delivery variation, total cost, and quality. Changing an input in the subsystem is likely to affect all three of these outputs. The output weighting in the C&E matrix for the TIR subsystem is completed from the customer's perspective. Quality is more important than delivery which is more important than cost for automobiles as judged by the person or team conducting the analysis. The inputs for the TIR subsystem C&E matrix are taken directly from the C&E diagram. Each input is given a score based on the likelihood that it affects the given output of quality, cost, and delivery. The values in the cells are multiplied by the output weighting and summed across the input row to give a total score for each input. The scores are then arranged so the largest score is at the top of the list and the smallest score is at the bottom. The top six likely causes of variation are then selected for analysis. The tops six in the TIR subsystem are as follows: Rework 3, Rework 1, Inspect 4, Inspect 5, Inspect 12, and Inventory 8. These inputs now become the focus of the analysis for which measurement systems will be developed, changes will be made, and improvement will be verified for. The measurement systems for these elements are not available and the analysis of these elements is not part of this thesis. The steps to develop measurement systems and conduct analysis at this level are the subject of the next section of this thesis. 88 Cause and Effects Matrix for the TIR Sub-System Foutputs (Y's) 10 9 4) Figure 27 a 10 9 9 8 7 10 8 10 4 6 9 4 8 8 3 7 8 2 6 3 8 2 3 8 9 9 9 5 2 8 8 2 1 1 1 2 2 0 0 0 Possible X's Rework 3 Rework 1 Inspect 4 Inspect 5 Inspect 12 Inventory 8 Rework 2 Inventory 7 Rework 4 Inspect 9 Inventory 9 Inspect 3 Inventory 11 Inventory 12 Test4 Inventory 13 Inventory 10 Inspect 1 Rework 5 Inspect 2 Inventory 6 Inspect 6 Inspect 11 Rework 7 Inventory 2 Inventory 3 Inventory 4 Rework 6 Test1 Inventory 1 Inventory 5 Inspect 10 Test3 Inspect 7 Inspect8 Test2 Test5 7 0 C 9 8 10 10 9 3 5 3 6 8 3 8 4 4 9 4 3 9 4 8 3 9 8 2 1 1 1 3 8 1 1 7 8 8 8 7 7 I- 0l 0 - Weighting for Outputs 6 5 2 1 2 6 5 5 7 1 3 2 2 2 1 3 3 2 4 2 2 1 1 3 3 3 3 5 1 3 3 2 1 1 1 1 1 222 196 195 179, 167 162 157 155 145 141 132 130 126 126 124 124 123 122 122 121 116 115 114 113 112 112 112 110 105 103 103 102 96 96 96 95 95 } Top possible causes TIR subsystem cause and effects matrix with the top 6 possible causes highlighted 89 5.8. Analysis of variation in the TIR subsystem Actual variation analysis was not possible as part of this thesis. The method for conducting this analysis is the subject of this section. The analysis consists of developing and validating a measurement system that will provide trusted data for the six areas listed in the preceding section. The next step is to collect and analyze data to determine which of the six elements has a significant amount of variation. The source of variation for each element can be analyzed and likely includes resource limitations, wasteful processes, problems with worker skill levels, and variation in the quantity and type of problems generated in upstream systems. Improvements are then developed and tested to ensure the desired reduction in delivery variation. Training would be administered or new workers would be employed as a possible improvement if worker skill level were found to be the issue in a rework area. Data for the number of daily problems found in the largest inspection area were compared to the delivery variation in the TIR subsystem (in the absence of delivery variation data for the specific inspection area) to see if problems created in the build subsystem cause delivery variation in the TIR subsystem. The number of problems per 100 vehicles was collected over a 50 day period and compared to TIR delivery variation for the same time period. The graph for the problem rate is shown in Figure 28. 90 Problems Identified In A Major Inspection Area - 180 - 160 - co 140 CD - 2 120 U, - o 100 S80 - C E 60 0- 40 - - -D -0 - 20 0 0 10 30 20 40 50 Build Day Figure 28 Problems per 100 vehicles built measured in a major inspection area The problem rate appears fairly stable over the period as compared with delivery variation in the TIR subsystem. The standard deviation for problems during the period was 13 % of the average. The standard deviation for delivery variation during the period was 72% of the average. We do not conclude that the delivery variation in the TIR subsystem is caused by the daily change in problems found in this large inspection area. We would not rule this out in other areas and would need to collect data to determine if there is a relationship in other areas. 91 5.8.1. Measurement system for key possible sources of variation in the TIR subsystem The measurement system for the elements of interest needs to be able to quantify the amount of variation in time for different vehicles that are processed through the rework area. The system also needs to identify the number of vehicles that are processed through the area. The simple solution is to provide a handheld barcode scanner for the vehicle to be scanned when they enter and leave the rework area. The amount of variation caused in each area can be determined given this information. 5.8.2. Target reduction in variation in the TIR subsystem The data collected by the handheld barcode scanner can then be analyzed using the same combination of Lean principles and tools and Six-Sigma methods as was done in this thesis for the order, build, and TIR subsystems. Compare the variation caused by each element to the total variation for the same population of vehicles in the entire TIR subsystem in this case. If the variation is large relative to the overall TIR delivery variation, then the element becomes an area where changes are made or new systems or processes are tried to reduce the variation in the TIR subsystem. 92 6. Conclusions and recommendations 6.1. Conclusions Average delivery time in the production system is important to the enterprise. Lower average delivery times generally lead to lower cost to the enterprise. Sophisticated measurement systems are in place at automotive OEMs to measure average delivery performance. The average performance of the build subsystem is the primary focus of the OEM since it is the subsystem where materials and other resources are employed for the longest time. Average performance of the build subsystem continues to be important to automotive OEMs for its impact on enterprise cost. The measurement systems that are in place to measure average delivery are also capable of capturing data that allows analysis of delivery variation performance. Delivery variation is especially important to individual customers as well as the enterprise. Each customer uses their own delivery experience as the basis for judging the delivery performance of the enterprise. Customers expect to get what they want, when they want it. System costs for both the customer and enterprise increase with delivery variation. The test, inspect, rework (TIR) subsystem was determined to be the largest source of delivery variation in this automotive OEM production system. Variation in the TIR subsystem does not appear to be stable throughout the measurement period used in this thesis. The TIR subsystem has traditionally received less attention with regards to delivery than the build subsystem. This is because TIR subsystem average delivery 93 performance is relatively short compared to build subsystem delivery performance. Average delivery and delivery variation lead the enterprise to work on two different subsystems in this automotive OEM case study. TIR delivery variation performance can be improved with a systematic approach to find and eliminate variation and waste. The systematic approach recommended in the thesis uses Six-Sigma methodology in combination with Lean principles and tools. The combined approach leads to analysis of the most important inputs to the subsystem. The system is decomposed to a level where the relationship between the inputs and output is understood using this systematic method. The sources of variation within the TIR subsystem are still to be determined. The relation of variation between subsystems did not show correlation. Many times the inputs that need to be improved are sources of waste. This is expected to be the case in the TIR subsystem. Improving variation in the TIR subsystem will improve delivery performance in the production system and ultimately delivery to the customer in the OTD system. The system dynamic loop can be improved by reducing delivery variation in TIR subsystem inputs. The system dynamic waste / variation loop is pictured in Figure 29. 94 Production Production System system Waste Delivery Variation External Customer or Enterprise Figure 29 System dynamic waste / variation loop Delivery variation will be improved in the production system when waste in the TIR subsystem is reduced. The improvement in delivery variation will cause a reduction in waste that is external to the production system. The external waste reduction is a benefit to both the customer and the enterprise. The external reduction in waste associated with delivery variation will help reduce the gaming that goes on in the OTD system that causes the production system to incur waste. This waste elimination comes in the form of less expediting and fewer schedule changes in the production system. Both the customer and the enterprise benefit from the reduction of waste that causes delivery variation in the TIR subsystem. 95 The long term variation performance in the production system depends upon how often the system dynamic loop is repeated and whether waste is increasing or decreasing in the system. The general shape of the performance curve for delivery variation is pictured in Figure 3039. Expected Performance of Production System Delivery Variation (based on system dynamics) w/o waste reduction E c co a)0 o' reduction Time Figure 30 6.2. Expected long term performance of delivery variation in the production system with and without waste in the system Recommendations The most likely sources of variation in the test, inspect, and rework (TIR) subsystem should be analyzed and improved to reduce delivery variation in the 96 production system. A chart can be used to report the average delivery performance and the delivery variation performance in the TIR subsystem and allow management to better see the variation in the subsystem. The chart provides visual cues that the system is either in or out of control. An example of such a chart is found in Figure 31 15 TIR Sub-System AverageNariation Performance Management Chart - Example: TIR sub-system delivery became unstable and corrective actions were taken --- 97.5% Average - -2.5% 10C', 5 01 Figure 31 11 21 31 41 51 61 71 Day Entered TIR Sub-System 81 91 Example of a TIR subsystem delivery performance chart The chart shows that delivery variation increased mid-way through the period used in the example. The team took action to return control to delivery variation performance and verified the improvement using the chart. Maintaining similar charts of performance for the most important inputs will make it easier to determine which input 97 went out of control to cause this change in delivery variation performance in the TIR subsystem. Delivery variation performance will need to be made an objective of the leaders who are responsible for the production system in the enterprise. The focus on reducing variation in and stabilizing the system will be lost unless performance targets are set in a person's objectives and the person is rewarded for meeting or exceeding the objectives. The performance to target becomes measurable through the use of the variation management chart described above. The same type of measure and objective is currently used to achieve build subsystem average delivery performance. Three areas for future study emerged from the research conducted to complement this thesis: " Cost to the customer " Delivery variation in other industries " Measurement systems The actual cost of delivery time variation to the customer is an area where little has been published. The knowledge gained from a study of these costs would allow an enterprise to determine the amount of resources to use in reducing delivery variation and waste. The initial work might be to study costs for a single industry by using several enterprise and customer case studies, such as the automotive industry. Future work could compare costs across different industries. 98 Delivery variation in other production systems would be of interest to see where variation originates in other production systems. Systems with both large and small amounts of variation are of interest and may lead to an understanding of how some production system architectures naturally lead to better delivery variation performance. The results of the study would lead to system architecting principles that yield a competitive delivery performance advantage for the enterprise. Measurement systems that allow for ease of measurement in the production system might also be an area for future research. Measurement systems that are inexpensive and easy to use for several different types of products would make it possible for more production systems to measure and understand both average delivery performance and delivery variation performance. 99 100 Glossary OEM: Original equipment manufacturer OTD: Order to delivery CTQ: Critical to quality characteristic CTC: Critical to cost characteristics CTD: Critical to delivery characteristics Delivery time DT: DMAIC: 6-Sigma define, measure, analyze, improve, and control project phases Production Order Date: The date the manufacturing plant receives the order from the internal order process from which the order will be produced Off Line Date: The date the product is complete with the exception of test, inspection, and repair Ship Date: The date the product is released to the distribution system and is sold by the manufacturer Span: Measure of variation in product delivery dates compared to customer expectation (sum of days early and days late) MSA: Measurement system analysis (similar to gauge R&R) Specific ID: ID for a specific unit of product that differentiates the single unit from all others. An example would be a serial number on a computer or other product Time Stamp: Date and time that a specific unit passed a certain point in the process. Used in calculating the overall process variation or the variation contributed by sub processes Cause & Effect Diagram: Diagram used to map possible causes of a problem Cause & Effect Matrix: Matrix that ranks most likely causes based on there importance to selected outputs TIR: The test, inspect, rework subsystem in the production system used in the case study System Decomposition: The breakdown of a system structure into subsystems that are smaller than the prior level. Decomposition ends when the system is broken down to the elemental level Expected Value: The average value for a population (of units in this case) 101 102 References Copeland, Dave. Alcoa showcases chemicals facility. Pittsburgh Tribune-Review, Thursday, June 27, 2002 Carlile, Paul R. "15.309 Organizational Processes" MIT Sloan School, Cambridge. IAP and Spring 2001 Crawley, Edward F. "16.882 System Architecture" MIT, Cambridge. Fall 2001 Donovan, Michael R. The Order to Delivery Cycle: Quick Response is Essential. R. Michael Donovan Co., 1998 Eppinger, Steven D. and Ulrich, Karl T. Product Design and Development. New York: McGraw-Hill, 2000 George, Michael L. Lean Six Sigma, Combining Six Sigma Quality with Lean Speed. New York: McGraw-Hill, 2002 Goldratt, Eliyahu M. and Cox, Jeff. The Goal, A Process of Ongoing Improvement. Great Barrington, MA: The North River Press, 1992 Harry, Mikel J. and Schroeder, Richard. Six siqma : The Breakthrough Management Strategy Revolutionizing the World's Top Corporations. New York: Doubleday, 2000 Hines, Jim. "15.983 System Dynamics Foundations" MIT, Cambridge. Summer 2002 Jordan, Jr., James A., and Michel, Frederick J. The Lean Company, Making The Right Choices. Dearborn, MI: Society of Manufacturing Engineers. 2001 Maier, Mark W. and Rechtin, Eberhardt. The Art of System Architecting. New York: CRC Press, 2000 Mudd, Tom. Back in High Gear. Industry Week, February 21, 2000 Nightingale, Deborah. "16.852J Integrating the Lean Enterprise" MIT, Cambridge. Fall 2001 Pine, Joseph B. Mass Customization: The New Frontier in Business Competition. Boston: Harvard Business School Press, 1999 Pyzdek, Thomas. The Six Sigma Handbook, A complete Guide for Greenbelts, Blackbelts, & Managers at All Levels. New York: McGraw-Hill, 2001 Rath & Strong. Six Sigma Pocket Guide. Lexington, MA: Aon Consulting Worldwide, 2002 103 Roemer, Thomas A. "15.769 Operations Management" MIT Sloan School, Cambridge. Summer 2001 Rosenfield, Donald B. "15.761 Manufacturing Strategy" MIT Sloan School, Cambridge. Fall 2001 Sheehy, Paul. "Six-Sigma Black Belt Training" Six-Sigma Academy, Phoenix. Spring 2000 Taylor, David and Brunt, David. Manufacturing Operations and Supply Chain Management, The Lean Approach. London: Thomson Learning, 2001 Voyle, Susanna. TESCO: Food e-Tailers in Search of a Recipe. The Financial Times, February 15, 2000 Welch, Jack, and Byrne, John A. Jack: Straight From the Gut. New York: Warner Books, 2001. Wheeler, Donald J., and Poling, Sheila R. Building Continual Improvement, A Guide for Businesses. Knoxville, TN: SPC Press, 1998 Womack, James P. and Jones, Daniel T. Lean Thinking. New York: Simon and Schuster. 1996 Womack, James P., Jones, Daniel T., and Roos, Daniel. The Machine That Changed The World. New York: HarperCollins, 1991 104 Endnotes 1 Donovan, Michael R. The Order to Delivery Cycle: Quick Response is Essential. R. Michael Donovan Co., 1998 2 ibid 3 Pine, Joseph B. Mass Customization: The New Frontier in Business Competition. Boston: Harvard Business School Press, 1999 4 Voyle, Susanna. TESCO: Food e-tailers in search of a recipe. The Financial Times, February 15, 2000 5 Welch, Jack, and Byrne, John A. Jack: Straight From the Gut. New York: Warner Books, 2001. Pg. 337 6 ibid., 337 7 Copeland, Dave. Alcoa showcases chemicals facility. Pittsburgh Tribune-Review, Thursday, June 27, 2002 8 Mudd, Tom. Back in High Gear. Industry Week, February 21, 2000 9 Donovan. The Order to Delivery Cycle 10 Harry, Mikel J. and Schroeder, Richard. Six Sigma : The Breakthrough Management Strategy Revolutionizing the World's Top Corporations. New York: Doubleday, 2000. Pg. 195 - 196 11 ibid., 195 12 13 14 15 16 17 ibid., 193 ibid., 108 -109 Welch. Jack: Straight From the Gut. Pg. 339 Nightingale, Deborah. "16.852J Integrating the Lean Enterprise" MIT, Cambridge. Fall 2001 Rath & Strong. Six Sigma Pocket Guide. Lexington, MA: Aon Consulting Worldwide, 2002, Pg. 5 ibid., 5 18 ibid., 5 19 Harry. Six Sigma. Pg. 129 20 ibid., 129 21 ibid., 129 22 Rath. Six Sigma Pocket Guide. Pg. 6 23 ibid., 6 24 ibid., 6 25 Harry. Six sigma. Pg. 129 26 Rath. Six Sigma Pocket Guide. Pg 6 27 Donovan. The Order to Delivery Cycle. 28 Rath. Six Sigma Pocket Guide. Pg. 6 29 Harry. Six Sigma: Pg. 129 30 ibid., 129 31 Rath. Six Sigma Pocket Guide. Pg. 6 32 Taylor, David and Brunt, David. Manufacturing Operations and Supply Chain Management, The Lean Approach. London: Thomson Learning, 2001. Pg. 79. 33 Ibid., 81 34 Hines, Jim. "15.983 System Dynamics Foundations" MIT, Cambridge. Summer 2002 35 George, Michael L. Lean Six Sigma, Combining Six Sigma Quality with Lean Speed. New York: McGraw-Hill, 2002. Pg. 51-56. 36 Nightingale. "16.852J Integrating the Lean Enterprise" 37 Crawley, Edward F. "16.882 System Architecture" MIT, Cambridge. Fall 2001 38 Nightingale. "16.852J Integrating the Lean Enterprise" 39 Hines. "15.983 System Dynamics" 105