1.1 THE TOTAL QUALITY MOVEMENT

Proponents of Six Sigma will be uncomfortably aware that the same enthusiasm and hyperbole now surrounding Six Sigma was observed only a few years ago with "Total Quality Management". Indeed, some of the errors that were made with regard to TQM may also be lying in wait for followers of Six Sigma. However, there are important areas of contrast between the two systems, as follows. Fuzzy v. Clear Goals. The goals of a TQM programme were often expressed in such terms as "exceeding customer requirements", with no means of tracking their achievement. In 6 , goals are set in central, very clearly stated measures (such as numbers of defects and money saved), and the tracking of achievement is an equally central matter. Theoretical v. Practical. TQM practice suggested that there often was only one right way to proceed down the quality path. 6 methods are chosen from a menu of tools, and the criterion of success is whether the project delivers returns, not theory. Focus on Product (Physical) Quality. TQM efforts were mainly directed at the quality of physical production output, and left aside such matters as commercial service. In 6 , service systems are regarded as a richer area even than physical production. Training. TQM training tended to be less well focussed, and often related merely to specific projects. 6 training is more intense and demanding, and more clearly relevant to the DMAIC procedure it is intended to support.

1.2 ISO 9000 The alleged focus on quality in Britain has been sadly diverted over these many years to the achievement of certification in ISO 9000. ISO 9000 is a paper driven system that literally has no relevance whatsoever to quality. The standard is concerned with the documentation of what is done and with providing records so that auditors can verify the matter. It is not directed at customers or financial advantage. Although Government encouragement may explain the mode for ISO 9000 in part, along with the constant presence of avaricious consultants and academics, it is a mystery why UK industry jumped on this bandwagon, thereby wasting so much money and so many opportunities.

1.3 HISTORY OF SIX SIGMA Six Sigma was originally devised by Motorola in 1987, that company seeing it as a simple, consistent way to monitor performance and compare achieved performance with customer requirements. Motorola also saw Six Sigma as a framework within which changes might be initiated so as to improve measured performance to near perfect limits. The technique was adopted as company policy by the then CEO Bob Galvin. It is credited by Galvin as having been instrumental in turning around Motorola's performance and financial fortunes. In the mid 1990s, Allied Signal (which took over Honeywell) adopted Six Sigma and subsequently reported year on year increases in financial performance as a result of doing so. Other big-name US and international companies that have gone down the Six Sigma route include Asea Brown Boveri; Black & Decker; Bombardier; Dupont; Dow Chemical; Federal Express; Johnson & Johnson; Kodak; Polaroid; Sony; and Toshiba. The most famous convert of all is General Electric. The charismatic CEO Jack Welch is on record as stating Six Sigma is the most challenging and potentially rewarding initiative ever undertaken at GE. It is credited by Welch as having pulled the Company "back from the brink" in 1999. Executives at GE state, firstly, that its value is in unifying staff effort around customer-directed processes and "boundaryless" (ie non-departmental) thinking, and, secondly, in forcing staff to focus on metrics, not opinion. In the 1998 GE Annual Report, Welch stated also The First major products designed for Six Sigma are just now coming into the marketplace and beginning to touch some of our customers. These products are drawing unprecedented customer accolades because they were, in essence, designed by the customer, using all of the critical-to-quality (CTQ) performance features the customer wanted in the product and then subjecting these CTQs to the rigorous statistical "Design for Six Sigma" process. Every new GE product and service in he future will be DFSS - Designed for Six Sigma.

2. THREE CASE STUDIES 2.1 COVENTRY PRESSINGS Reading Point 1 Coventry Pressings Limited stamps sheet steel to make doors for the Horizon model of Talbot Auto Ltd in Birming-ham. Pressed steel, however, is always likely to suffer defects, such as pits, dents, bruises and dings. Such defects in the pressed metal might be due to the specification of the steel itself or might be caused by the pressing process (eg dirt or debris in the press itself). Recently, Horizon doors have been exhibiting some 18 dings per door, against a target maximum of 2 or 3. This has resulted in major rework costs and, in some cases, in dings getting through to the final vehicle. Coventry Pressings have now initiated a Six Sigma project to eliminate the dings problem. In order to do so, they have prepared and issued a Project Charter which lays out the following: (1) the financial case for solving the problem; (2) the problem statement itself (ie dings on doors); (3) the project scope (in this case, no new equipment is to be purchased); and (4) the target completion dates for the Six Sigma milestones. In order to complete the Charter, it has been necessary also to obtain from the customer of the stamped door a statement of his requirements. (Here, it was found that there was a difference between requirements of the Talbot Painting Shop and of the Talbot Assembly Department, since dings are more easily visible after painting than before it.) Note also that the Six Sigma project champion (the sales & marketing manager) requires that the investigation be confined to Coventry Pressings' own operation and is not to include those of the steel supplier or Horizon distributors. The Six Sigma team now document the cutting and stamping process by the construction of a flowchart. The flowchart follows the conventions of a traditional six sigma tool called a SIPOC diagram. The SIPOC diagram will be explained in Session 2; for now, it is a simple diagram showing the relationship between suppliers, processes and outputs. The various process steps involved are as follows overleaf: receive steel sheeting; unbundle; cut steel; check dies; check press; feed cut steel to press; press steel; remove doors from press; inspect doors; load doors onto racks for despatch. (End of "Coventry Reading Point 1") Reading Point 2 In examining the sub-processes of cutting and stamping, team members make many guesses as to the cause of the dings by means of brainstorming and by the use of fishbone diagrams. However, the hypotheses formulated as to cause at this stage cannot be tested until the team has gathered data (the "Measure" step of a Six Sigma investigation). Data to be gathered include the periodicity of ding production (ie whether they are produced cyclically, or at the end of shifts, etc.); the operating conditions of the press; the chemical and physical properties of the steel etc.. Some of the data enable many hypotheses as to cause to be discarded. Other data require further expert analysis to be undertaken at the "Analyse" stage of Six Sigma, using one or other of a variety of tools, such as statistical regression or DOE (design of experiments). (As with so many studies of this nature, it is likely that one or two causes only will be found to give rise to the problem being investigated.) When the cause or causes of the problem have been determined, the team will move on to the "Improve" stage of Six Sigma and devise changes to the process procedures which will have the effect of eliminating the cause ... and the dings. Finally, the team will put in practice the six sigma "Control" stage, to ensure that the changes and improvements made become permanent.

2.2 MELODY MICROSYSTEMS Reading Point 1 Melody Microsystems specialise in selling and delivering customised computer systems to individuals, small businesses and corporate businesses in the West Midlands. Most sales are obtained through telesales at its distribution centre in Birmingham, but some orders are placed via the Internet, where customers determine their own hardware specification.. Although sales volumes have increased substantially, customer dissatisfaction has also increased, along with customer returns of wrongly despatched equipment. A Six Sigma project has now been initiated and a Project Charter issued to investigate and correct this situation. The project team has been asked to limit its scope to sales involving the Birmingham distribution centre. Furthermore, the Project Champion requires the team to confine its investigation to the sales and despatch part of Melody's operation only (ie only SIPOC steps (1) to (3), as described below). The team have constructed a Six Sigma SIPOC map involving 7 process steps: (1) Customers place orders with Telesales in Birmingham; (2) Telesales send invoicing information to Accounts, and forward customer order information to the Order Verification (OV) section; (3) An Order Verification clerk verifies the order and sends it to Order Picking; (4) Order Picking picks the parts and sends them to Assembly section; (5) Assembly section assemble the system and send it to Despatch; (6) Despatch sends the final system to the customer; (7) Orders which are wrong or where parts are missing are returned and handled by the Returns section. Further investigation in this initial stage (the "Define" stage of Six Sigma) reveals that the phone orders taken by Telesales are passed to OV clerks so that they can be verified for compatibility and completeness. If an OV clerk finds a problem, he either contacts the customer or Telesales, or perhaps checks the database. More often, however, he refers to the Melody web site rather than the database, since the web site is more up to date. The team also find that Telesales clerks receive a bonus depending on the volume of business they obtain. And finally they find that during recent sales expansion, management transferred a number of OV clerks to Telesales itself. Next, the team moves to the Six Sigma "Measure" stage, and collects data on daily sales, number of orders received each day, and daily returns. In doing so, it finds that there has been a large rise in returns in the last 5 or 6 months. Further investigation is made to find whether there exists a relationship between customer type (individual, SME, corporate) and order type (complete system v. components only). The team also finds that Telesales, under pressure of time, do not always verify customer despatch data or system compatibility as orders are being taken, leaving these matters to the OV clerks for correction. They also find that orders often pile up in the OV section, with OV clerks verifying orders so as to get the easy ones out first. (End of "Melody Reading Point 1") Melody Reading Point 2 In the "Analyse" phase of the project, the team constructs run charts relating to sales and returns. They also construct a Fishbone diagram. Ideas as to cause generated by the team at this stage include: (a) That the transfer of OV clerks into Telesales have lead to wrong orders being taken, due to the inexperience of the new staff; (b) That there may be a simple correlation between the number of returns and the sheer volume of business (a scatter graph is drawn to investigate this - some correlation is indeed detected); After yet further analysis, and a return to measuring data, the following causes are arrived at by the team: (1) Small businesses and new customers do not use the web site to do ordering - they rely on Telesales clerks. Telesales are encouraged and rewarded to handle orders quickly, partly due to increase in business. Consequently, Telesales clerks leave all verification and correction to OV clerks. However, Telesales refer to the database if checking proves necessary, not to the web site. (2) Telesales do not use a standard script to get down all required customer information. (3) The orders that OV clerks are required to verify are based on Telesales orders which themselves are based on the database. Further, OV clerks are being overwhelmed by the volume of checking necessary and the number of call-backs they need to do. As a consequence, to speed things up, they are also referring to the database rather than to the more up-to-date web site. In effect, they are simply passing the Telesales orders along rather than verifying them thoroughly from the latest data. (4) Data maintenance staff update the web site as changes occur, on line, but update the database only weekly . In the "Improve" phase of Six Sigma, the team recommend that database updates are made on-line so the database is always in sync with the web site, and that Telesales work to a standard script.

2.3 AUDIOLINK Reading Point 1 Audiolink of Walsall manufacture micro cassette and digital "autotalk" devices which are used for recording speech and dictation, and which are then able to generate hard copy text. A problem has arisen in which despatches sent to customers do not match customers' specified requirements as to micro cassette or digital. Returns from customers are costly and may threaten Audiolink's competitive position. A Six Sigma project team has been formed, directed at this problem. At the "Design" step, the team has met to consider its Project Charter and define the problem and goals. These concern both defective shipments and the timeliness of deliveries. Next, in the "Measure" phase of Six Sigma, the team investigates the "voice of the customer", and finds far less concern with speed of delivery than Audiolink had previously assumed. Team members also collect data on the number and type of despatch defects; discrepancies between order data and recorded shipment details; the cycle times of major phases in the process; inventory levels; and despatch activity per day. The team draws up a SIPOC map including process steps as follows: take order; pick and/or procure parts; assemble order; deliver order. The data show the cycle time average is 11.6 days (with a maximum of 17.0 days), that most of the problems are related to wrong hardware and that there is little discrepancy between the order data recorded and the order shipment documentation. Discrepancies are in what is physically shipped. As a result, the charter problem statement was amended. Next, the team turned to the "Analyse" phase of Six Sigma, beginning with a brainstorming session from which five "causal hypotheses" were obtained. (These were merely guesses, to be be examined through data analysis in due course). It was felt that the two most promising hypotheses were: (1) that errors were being made in rushing out despatches to meet deadlines, and (2) that assembly staff, including part time workers, were not adequately trained and were mixing up micro cassette and digital devices. (End of "AudioLink Reading Point 1") AudioLink Reading Point 2 Further investigation revealed that procurement kept a months stock of small cheap items, but ordered more complex parts as they were needed for specific orders. They also found that, when deadlines were approaching, despatch staff helped assembly staff in picking The team rejected a number of hypotheses at this point, and collected more data relating to hardware incompatibility. They also used a statistical tool (ANOVA) to see whether hardware incompatibility was more serious at some times than at others. Analysis determined that the most common problems, accounting for 60% of all bad deliveries, were incompatible connectors and adaptors, and that connectors and adaptors were both carried as stocked items, not delivered from the supplier on a Just-in-Time basis like the more expensive, complex components. Further data analysis also showed that the problem being investigated had only arisen when the digital autotalks had come out. One problem is that micro cassette and digital connectors are packed in similar types of plastic bags, which are translucent. It had not been noticed that the wrong connectors (micro cassettes) were being sent with the digital machines, since the Return section's goal was speedy re-assembly, with returns simply put back into inventory. The team moved to the "Improve" stage, soliciting ideas through brainstorming and a company-wide appeal for ideas. What was decided was (1) that no connectors or adaptors would be carried in stock - all parts for an order would be obtained on a JIT basis from suppliers; (2) that packaging colours would be improved, especially as these related to the micro cassette and digital connectors, and (3) that the Audiolink bonus system would be altered to pay bonus only for on-time deliveries that were correct and that it would embrace assembly staff as well as despatch.

3.WHAT IS "SIX SIGMA" ?

3.1 VARIATION At the moment that a manufacturing process operates - ie at the split second that the process creates the part that is being manufactured - there are many factors which contribute to the potential state of the part under transformation. Examples are the temperature of the machine; pressure; the electrical potential; the speed of the device; chemical properties; and so forth. The particular state of each of these factors is due to chance - ie due to random probability. Since the state of these factors affects the form and properties of the object undergoing manufacture, we can predict that small variations will exist after manufacture in each object's physical properties, one to the next. Such variation among the manufactured items is indeed observed. The variation is random, and we conclude it is due to chance causes influencing the factors contributing to manufacture. Variation is also observed in the output of processes other than manufacture. Consider, for example, a parcel delivery service in a large town. The time to deliver a parcel will depend on the distance the particular parcel has to travel from the depot; the state of the traffic; the time of day; the weather; and such other chance causes as the disposition of the delivery vehicle and even its driver. We might publish a target for delivery time, but we would expect there to be variation of the actual time to deliver due to chance causes we can do little about. Chance variation due to random, inescapable causes is inevitable in all systems and processes, and those observing variation in the output from these systems must understand it and why it occurs.

3.2 FREQUENCY DISTRIBUTIONS To "record" the output from a system or process, it is obviously necessary to measure it using some selected metric such as hours/minutes; kilometres per hour; grams; etc. When the measurements of many outputs are recorded from the system, the various readings are found to be distributed among a range of values. For example, there will be a lowest reading; a highest reading; all readings above a certain value; etc. If we find how many readings are in each class of value (eg under 50; over 200; between 30 and 70; and so on), we can be said to find the frequency of the various readings. A frequency distribution is therefore a more elegant term for how many are in each category of measurement. There are three common ways a frequency distribution might be presented. As a Simple Table As a Histogram NSIX019n The third method is as a continuous curve on a graph, provided we have a formula to allow us to draw the curve. The vertical axis is the frequence of the distribution frequency, usually expressed as a percentage .

3.3 EXPRESSING FREQUENCY DISTRIBUTION A common frequency distribution found in manufacture is a bell curve. The measurements relating to a particular bell curve have the greatest frequency at a central point, and then fewer and fewer occurrences as one moves away from the centre in either direction. A sketch of the bell curve is given below, showing the central point. The drawing of the bell curve on paper is deceptive - as the curve gets further and further away from the centre point, the curve gets nearer and nearer the horizontal line, but never quite reaches it. The formula necessary to draw the bell curve is extremely complex as follows: One of the terms in the formula is µ (mu, Greek letter "m"). This is the centre of the curve on the horizontal axis of the graph. A second vital term is (sigma, Greek letter "s"). This term governs the degree of spread of the bell. If is large, the bell is very spread out. If is small, the bell is tall and narrow. As before, note that in each of the three bells above, the curve does not actually touch or reach the horizontal axis,. Although the curve does not reach the horizontal axis, and in fact continues to infinity on either side of the centre, it gets very close. If we take a distance from the centre to 3 on either side, we encompass 99.73% of the area of the curve. The further and further we go from the centre, either side, the more and more of the area of the curve is covered. For example, if we go from the centre plus or minus 4 standard deviations, we cover 99.9967% of the curve.

3.4 FREQUENCY DISTRIBUTION & SPECIFICATIONS In order to know whether the output from a process or system is within the specification required by the manager or engineer, one must clearly compare the specification range to the range for the process/system as observed in the frequency distribution of its output Suppose the specification range was the same as the observed output range of (average + 3 ) to (average - 3 ). We must remember that an observed range of (average + 3 ) to (average - 3 ) is only 99.73% of the total output from the system. This would mean that 0.27% of the output from the system was outside the specification. By convention, when the specification range is equal to (average + 3 ) to (average - 3 ) of observed output, the process is said to be "just capable". The process capability is defined as: , If the process is just capable, Cp - 1.0. Suppose, however, that the specification range was far narrower than the observed range of (average + 3 ) to (average - 3 ) . In particular, suppose the observed range was twice the specification range. This is illustrated as follows: The capability index is 2.0, and = 2.0. The process is now said to be at 6 sigma. The percentage of the bell curve covered is 99.99967%

3.5 CALCULATING THE SIGMA LEVEL OF A PROCESS The quick way in Six Sigma to calculate the achieved sigma level of a process (in other words, its capability index), is to find the number of defects, or non-conformancies, produced in a given period or over a given phase of operation, and scale the figure found to a standard measure of defects per million opportunities (DPMO). "Opportunity" means an opportunity for the process actually to produce a defect, such as the production of a unit of output or the performance of an instance of service. When the DPMO figure has been found, a statistical table can be consulted (see p.16) to find the sigma performance of the process. The statistical table is obtained directly from calculations involving the bell shaped curve and its area, as explained above. The scaling up referred to is obviously necessary because one does not wish to observe a million opportunities literally. Two Examples Example 1 Suppose that there are 1000 process operations, and that the number of defects produced is 50. That is, there are 50 defects per 1000 opportunities. To scale this to a number of defects per million opportunities, we divide the 1000 opportunities by 1000, to obtain 1, and then multiply by 1,000,000. If we do so, we must also divide the number of defects by 1000 and multiply by 1,000,000. In doing so, we obtain the figure 50,000 ((50/1000) × 1,000,000). That is, the process yields 50,000 DPMO (50,000 defects per million opportunities). Next, we consult the Table on p.16. Columns 2 and 5 of the Table give values of DPMO from 933,200 down to 3.4. The nearest DPMO value to 50,000 is 52,100 (column 5, second entry from the top). We see that the sigma corresponding to 52,100 is 3.125. We can therefore say that our process has a 3.125 sigma capability. Note however that in Six Sigma one operation of the process may be considered to present more than one opportunity to create a defect or non-conformance. For example, potentially, 3 different flaws might be possible on a manufactured unit, or 3 different types of mistake might be possible on an invoice. In this case, if 1000 objects are made, or 1000 invoices are prepared, the manufacture or preparation could be considered to have presented 3000 opportunities, not just 1000. Example 2 Suppose that 1000 invoices are produced and that it is found that 50 of them are wrong. However, in an assessment of the invoicing procedure, it is estimated that there are 5 opportunities to make a mistake for every invoice prepared. That is, there are 50 defects per 5,000 opportunities. As before, to scale this to a number of defects per million opportunities, we divide the 5000 opportunities by 5000, to obtain 1, and then multiply by 1,000,000. If we do so, we must also divide the number of defects by 5000 and multiply by 1,000,000. In doing so, we obtain the figure 10,000 ((50/5000)× 1,000,000). That is, the process yields 10,000 DPMO (10,000 defects per million opportunities). Next, as before, we consult the Table on p.16. The nearest DPMO value to 10,000 is 8,800 (column 5, eighth entry from the top). We see that the sigma corresponding to 8,800 is 3.875. We can therefore say that our invoicing procedure has a 3.875 sigma capability The general formula for the DPMO rate is as follows: The final point on the sigma figure relates to the so-called "straight through" percentage. If there are, say, 4 stages in a process, and each stage individually produced 10% defects, then if 100 units are processed, the number of conforming units produced that go straight through from beginning to end without any correction at any of the intermediary stages is about 65. This is the "straight through yield". The number of defects is 35. However, in reality after every stage, the 10% of defects may be corrected and sent on. In this case, there would be 100 good units ready for Stage 4, and the final tally would be 90 good units. In six sigma, both the final yield and the straight through yield are recorded and used in process analysis. The better figure to use for six sigma purposes is the straight through yield, but either might be used. What is vital is that there should be consistency. Mini Exercises 1. The Six Sigma team at Audiolink of Walsall estimate that for every sales order they respond to, there are 8 opportunities for an error. In the first quarter 2003, Audiolink despatched 200,000 orders and received 16,000 complaints and returns. From the Table on p.16, what is the approximate sigma rating of Audiolink's order/despatch operation? What would have been the probable number of complaint or returns in the first quarter if Audiolink's process had been at 6.0 sigma? Exercises Continued Overleaf 2. A team from Melody Microsystems has investigated the opportunities for customer delivery errors in the despatch of assembled systems as follows: Invoicing - 1 opportunity; The Packing & Despatch operation - 1 opportunity; and Correctness of the assembled system - 2 opportunities per order. Last year, Melody sent out 100,000 orders, and receives 8,000 invoice complaints and 32,000 actual returns. From the Table on p.16, what is Melody Microsystem's approximate sigma rating on its sales/despatch operation? What number of complaints and returns would the company have received last year if its process had been operating at six sigma? 3. Production one week at Coventry Pressings is 5,000 doors. It is found that 350 doors contain more than 2 dings. (a) What is the process yield of the pressing operation, as a % of output of good quality doors? (b) From the "straight through percentage" figures in the Table on p.16, what is the sigma rating of the process? (c) How many doors out of 5,000 would be defective if the rating was raised to 4.5 sigma?

Course Agenda: