Your SlideShare is downloading. ×
Wc Statistics
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Wc Statistics

575

Published on

Published in: Technology, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
575
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
4
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Revised 7/14/96 How do World Class Organizations Use Statistics? (A Study of Best Practices) William C. Parr Head, Department of Statistics University of Tennessee 332 Stokely Management Center Knoxville, Tennessee 37996 - 0532 Phone: 423-974-1631 Fax: 423-974-2490 E-mail: wparr@utk.edu Scope and Purpose In this paper, I outline the dimensions of an ideal “implementation” of statistical methods for improving organizational performance. I do not attempt to give a detailed discussion of all points. Neither do I discuss how to deal with the obstacles organizations typically run into in moving toward this ideal state. Instead, I focus on painting a rough picture of the ideal state. Obviously, individual company culture and characteristics may require adaptation of some specifics. A Caveat, a Note, a Disclaimer, and an Invitation One caveat: No single company I have seen or worked with fully exhibits all the ideal characteristics. I say this, having spoken in detail with representatives of over 200 companies in the last four years. For brevity, I often refer to any organization with these ideal characteristics as a world class organization (WCO). A note: to protect the privacy of organizations with which I have spoken, I have avoided naming any specific organizations in this paper. For the same reason, I avoid specific © 1994, William C. Parr 1
  • 2. discussions of proprietary software, or of specific organization’s choices of process variables for emphasis. A disclaimer: The author formerly worked for several years in the semiconductor industry, at the Harris Corporation. We did not then apply all that I discuss here. Sometimes this was because of lack of knowledge or persuasive power of the author. Sometimes this was because the tools and technologies were not all available in the 1980s. Some advice is based on my continued learning after leaving full time employment in the semiconductor industry. Some is based on seeing what other organizations did well, in dimensions where the author’s efforts were less successful. However, the bulk of the advice is based on the author’s personal experiences with the suggested practices, coupled with conversations over a period of more than ten years with several of the largest and most successful companies in a variety of industries. An invitation: The author encourages comments on the content of this paper. If you find the paper useful, or have a strong reaction to something in the paper, let me know by phone, fax, or E-mail at the addresses shown above. I will put anyone sending me their feedback on my distribution list for future versions of the paper as it evolves, and acknowledge their contributions (if they indicate they are willing to be recognized). Those who have contributed their comments and insights thus far include Richard Lynch (Harris Semiconductor), Lin Wong (Harris Semiconductor), Jim Zalnoski (Harris Semiconductor), Stu Weisbrod (Harris Semiconductor), Amanuel Gobena (National Semiconductor), Andy Palm (James River) and Rex Bryce (Brigham Young University, formerly with Intel Semiconductor and Signetics). They have the thanks of the author. Obviously, they do not necessarily agree with all that I say in this paper. © 1994, William C. Parr 2
  • 3. Table of Contents Scope and Purpose 1 A Caveat, a Note, a Disclaimer, and an Invitation 1 How Can You Use This Paper? 4 Purpose of Statistical Methods 4 Control or Learning? 4 Roles of Corporate Management 6 Roles of Research and Development Management and Organizations 8 Role of Marketing and Other Customer Relations Organizations 10 Role of Production Control (Scheduling) and Forecasting Personnel 11 Role of Plant Management 12 Knowledge Base at Process Level 13 Nature of Variables Studied 14 Role of Statisticians 15 Leader for Statistical Methodology 17 Organizing for Improvement in Use of Statistical Methods 18 Use of Statistical Consultants 19 Role of Technology 19 Education and Training 20 Control Chart Methods 21 Feedback Mechanisms 22 Illusions and Distractions 22 Summary and a Look to the Future 24 References 26 © 1994, William C. Parr 3
  • 4. How Can You Use This Paper? One way to use this paper is to read it carefully and annotate it with your comments about where you agree or disagree. Also, note your judgments about how your organization stacks up against the model. Then (after of course faxing the author a copy of your marked up version!) discuss your impressions with some of your colleagues. A natural next step would be jointly to prioritize those areas in which you judge your organization to be most seriously deficient compared with a WCO. Then, make action plans on how you might reduce the deficiency. Purpose of Statistical Methods The purpose of statistical methods is to enable the people and the organization to manage variation better: To identify sources of variation, and continually reduce process variation, to improve value provided to the customers of that organization. This means improving the ability of all key processes of the organization to serve the customer. Leaders in statistical methodology must ensure that all clearly understand this prime purpose. The organization’s leadership must clearly communicate that process improvement is the prime purpose of statistical methods — not maintenance, control, or developing effective “alarm systems” which notify process managers when processes suffer a turn for the worse. (These three other purposes, suggested by many as the purpose of statistical methods, represent minor areas for statistical work, not the prime purpose.) Control or Learning? Many organizations (and no WCOs) view statistical methods as primarily useful for control purposes. WCOs recognize the difference between control and learning & improving. Control — the situation when we want to put in place an economical means of compensating for variation we either do not know how to reduce, or cannot afford to reduce. We are (perhaps) identifying special causes and responding to them, but the primary focus of the efforts is not aggressive improvement, but maintenance of the © 1994, William C. Parr 4
  • 5. current level of performance, and rapid detection of any deterioration from that level of performance. Learning & improving — the situation when we need to learn the causes of variation, so that we can change the process and improve performance. In this situation, we are more aggressive. Subgrouping plans can change frequently. In this situation, the organization is on the steep part of the learning curve for the given process. Robustifying — a long word for what some view as a third stage, but this author views as merely a very important component of the “learning” stage. In this situation, we are trying to understand how to reduce the impact of variations in inputs and process variables on the variables we care about (usually, outcome variables, or measurements farther down the line). For instance, we might be trying to reduce the effects of temperature profile variation in a diffusion tube on key parameters of the wafers processed in that tube. If we are trying to robustify the process, then we are not so much concerned with flattening the temperature profile as we are concerned with dampening its effect. Three Stages for Work on a Process (and Three Purposes for Use of Statistical Methods) • Controlling ---- Maintaining current process, with good alarm systems to warn us in case of process degradation • Learning --- Studying sources of variation, ever mindful of looking for ways to reduce variation • Robustifying --- Determining ways to reduce the impact of natural process variation Display 1 © 1994, William C. Parr 5
  • 6. WCOs clearly understand the difference between situations calling for “control” and situations calling for “learning & improving.” They direct their actions in ways appropriate to the situation. They do not substitute use of statistical control charts as sophisticated alarm systems for an ongoing learning process targeted at both continual and radical improvements in process performance. For the same reason that “When you get rid of the things you don’t want, you don’t necessarily have the things you do want,” the pursuit, detection, working to root cause, and root cause removal tied to assignable causes as defined by real time, on–line control charts cannot, standing alone, lead to the improvements in performance which can result from study of the common cause variation to understand how to reduce that variation. “In control” is faint praise for a process. Predictable could be predictably bad. Roles of Corporate Management Corporate executives have several key roles to play in the use of statistical technology. First and of greatest significance is that they need to personally supply leadership in these efforts, not technically but by emphasis. Hence, they need to have a solid grounding in the tools of variation. Executives need to understand how studying variation relates not only to operations and development, but more broadly to their actions at the executive level. This understanding will guide their actions, and enable them to ask the intelligent questions that spur the entire organization on to effective use of statistical methods. The quality of the behaviors of the executives is the single most significant driver of success of an organization at applying statistical methods. When the book to bill ratio (ratio of new orders to shipments) goes up from 1.05 to 1.1, is a celebration called, or is there a thoughtful conversation about natural variation? Does the natural variation from one month to the next month average .1? If so, a change of .05 from one month to the next is not only not cause for celebration, it is a change of magnitude less than typically encountered! No reason to assume it is other than common cause variability. 1 The same discussion can take place for factory yields, on time delivery performance, or any other key outcome measure of interest. 1This area: the personal leadership provided by the executives, is the single area where © 1994, William C. Parr 6
  • 7. Second, corporate management need to insist that the senior statistician of the organization has the access to them described under “Role of Statisticians” below. Third, corporate management need to ensure that appropriate technology (software and access to hardware) are available to support off-line statistical analysis needs, and that statistical process control methods are appropriately linked in with the CAM system. Fourth, corporate management need to provide adequate access to statistical talent for all within the organization. Typical means of access include consultation with the senior statisticians, consultation with site statisticians at each major site, availability of consultants, and proper training in statistical methodology. WCOs consistently meet these three conditions. Many WCOs have already made the difficult transition to linking on-line SPC to their CAM systems (recognizing the distinction between control and learning discussed above). The highest levels of management of the WCO are likely to find themselves to be invited speakers at technical conferences on applied statistics and quality, at which they provide lucid arguments based on theory and on their experiences for advancing statistical applications within industry. WCO organizations and their executives are most strongly differentiated from those in other organizations. © 1994, William C. Parr 7
  • 8. Roles for Corporate Management • Provide personal leadership by example • Provide routine access to their calendar by lead statistician • Ensure presence of appropriate computer technology • Ensure all have proper access to statistical talent Display 2 Roles of Research and Development Management and Organizations The use of statistical methods is not confined to operations. In fact, some of the greatest gains from use of statistical methods come outside the context of operations. The use of design of experiments (DOX) and statistical process control (SPC) are imbedded in the research and development organizations of WCOs. In WCOs, senior R & D managers conducting process or product development reviews place a consistent and strong emphasis on the sound use of statistical methodology, asking intelligent questions. They not only provide access to training, but ensure that all employees are appropriately educated in the proper use of statistical methods. Process “transfer” or new product “transfer,” in a WCO, is not a point in time but a process, enabled by substantial use of statistical methods to confirm the viability of the operating window, the robustness of the product / process, and the consistency of results in development with those in early and ongoing production runs. In a WCO, major process or product developments conducted through a team structure (the norm in WCOs) are supported by either a statistician on each team, or a statistician assigned substantive support duties to each team. © 1994, William C. Parr 8
  • 9. In a WCO, development engineers possess a clear and complete knowledge of process capabilities for key process variables. This knowledge is updated regularly in WCOs (not continually, as might be thought initially would be desirable, and is currently feasible with modern information technology) as appropriate. In a WCO, development engineers clearly understand the impact of producibility of product and process design on providing customer value and enhancing organizational performance. In one case, a lead engineer in development in a semiconductor company (a military supplier) told this author: “You don’t understand --- If our yields go up, we can't sustain our overhead. We’re on a cost plus basis.” Clearly, with this mentality, a sustained and vigorous commitment to continual improvement using statistical methods was not likely to be forthcoming from this lead engineer. © 1994, William C. Parr 9
  • 10. Roles for Research and Development • Use DOX and SPC in process and product development • Use statistical methods for process and new product transfer • Seek out knowledge of current process capability • Design for producibility using statistical ideas and methods Display 3 Role of Marketing and Other Customer Relations Organizations The marketing organization, and all others that have strong customer relations roles, have definite roles to play. One significant role for marketing and customer relations in a WCO is the obtaining, management, and use of information from customer inputs (surveys, customer focus groups or in-depth interviews, . . . ). In this role, the marketing or customer relations organizations must rely on the statisticians for support in the design and analysis of the instruments for surveys, focus groups, or interviews. © 1994, William C. Parr 10
  • 11. Role of Marketing and Other Customer Relations Organizations • Use statistical tools in study of customer value • Interpret customer trends in the light of statistical ideas Display 4 Role of Production Control (Scheduling) and Forecasting Personnel Many operations organizations have a group with the name “Production Control” or “Scheduling.” Also, many organizations have personnel (often within the marketing organization) who produce forecasts. Those with scheduling or forecasting responsibilities need to have a clear understanding of statistical methods for study of variation, as well as (depending on the context) more sophisticated time series – based methods for forecasting. In WCOs, scheduling and forecasting is done with the aid of good statistical tools, whether supplied by special help from a statistician or statistical group, or by training to equip the scheduling and forecasting personnel to do the work with a minimum of consultation. © 1994, William C. Parr 11
  • 12. Role of Production Control (Forecasting) and Forecasting Personnel • Understand impact of variation on flow of product • Use statistical methods for forecasting • Consult with statistical experts on sophisticated forecasting applications Display 5 Role of Plant Management In an operations context, the “rubber hits the road” with plant management. The plant management sets the tone, within the plant, for the organization’s disposition to using statistical methods. (In this paper, I refer to the plant manager and, if appropriate, senior manufacturing engineering manager and senior production manager as, collectively, plant management.) This author has seen a side by side case of two plants in the same organization, provided with the same “resources”: consultants, training, knowledge, higher level plant commitment, etc. . . . but one plant being dramatically successful at improving via use of statistical methods, and the other being a demonstrable failure. The difference was that provided by the plant management (leadership) teams for the two plants. (I am thinking of one specific case. Readers of this paper can doubtless contribute others.) Plant management must ensure that regular meetings occur at which there is a review of the ongoing progress using statistical methods. Plant management must not merely attend these meetings, but is, in WCOs, in a visible role as a key customer of that meeting. Typical practice of WCOs is to have a weekly meeting at the level of a plant or factory unit, to review ongoing improvement work and products using statistical methods, and to periodically review the on-line control charts to see if the group has insights that have not © 1994, William C. Parr 12
  • 13. been uncovered in day to day operation. In WCOs, typically a plant / factory would have a meeting with a broader scope either monthly or quarterly, to review projects in less detail, to encourage the efforts, and to uncover obstacles and roadblocks to improvement that may exist. Role of Plant Management • Set tone for attitude toward use of statistical concepts and methods • Ensure regular meetings for review of ongoing progress • Supply personal leadership by example of quality of questions asked Display 6 Knowledge Base at Process Level WCOs determine an appropriate level of knowledge about each of their key processes, and ensure that they have this knowledge. Specifically, units in WCOs determine key processes and key process variables that must be managed. For these key process variables, WCOs conduct studies of the existing measurement processes, and determine the adequacy of these measurement processes for their needs. WCOs know the capability of their processes on the dimensions of these key process variables. WCOs know the appropriate operating window for each key process. © 1994, William C. Parr 13
  • 14. Knowledge Base at Process Level • Key processes • Key process variables • Status of measurement processes for key process variables • Current process capability for key process variables • Variables that affect the mean of key process and outcome variables • Variables which affect the variability of key process and outcome variables Display 7 Nature of Variables Studied In a WCO, the focus is not primarily on “outcome variables” such as yields, quality levels, . . . Instead (although these variables are admittedly important to the customer and to the organization), the focus is on variables in the process which they know that led indicators of movement in these variables, or key cause variables. A clear sign of an organization at a low level of maturity is the predominance of attributes data control charts. As an organization matures, causal factors (process variables) are identified which can be managed to affect the attribute variables (usually outcome variables such as yields). Then, variables control charts increasingly dominate attributes control charts, as maturity increases. © 1994, William C. Parr 14
  • 15. A second sign of maturity in an organization’s use of statistical methods is that the variables measured tend to be earlier in the process, as a general trend. This is also a result of moving from outcomes to process variables. Role of Statisticians Not surprisingly, statisticians are often poorly used. Their placement within the organization, and the roles they take on, both can be key aspects of an organization’s strategy for managing variation. All too often, companies have statisticians focused primarily on the “lowest” levels of the organization, and either ignoring or not having access to the executive group. This is a serious mistake! Those face a dilemma attempting to move forward in the use of statistical methods. It is totally unacceptable to attempt to provide leadership in use of statistical methods by relying solely on those with short course training, no matter how bright, well-intentioned, and otherwise experienced. However, typical training of M.S. and Ph.D. statisticians leaves them initially poorly prepared for work in industry (a few pleasant exceptions do exist — see below!). As a result, the discerning company hires degreed statisticians with demonstrated experience in industry. (Smaller companies will often need to hire fresh graduates, and supplement their knowledge and experience with the advice of outside consultants who themselves have demonstrated successful experience at full time work within organizations.) This author has never seen a WCO in which statisticians spend, on average, more than 25% of their time delivering training. The other demands on their time include consulting with those using statistical methods, advising management, developing new training modules, working with customers, . . . These easily consume, in an effective organization, the other 75% of the available time. For statisticians to spend an average of over 25% of their time on training is a clear leading indicator of lack of success in penetrating the organization with statistical thinking and methods. In a WCO, the statistician is viewed as a legitimate scientist, with substantive inputs which must be sought out for the benefit of all areas of the business. The WCO provides © 1994, William C. Parr 15
  • 16. for the continued professional growth of its statisticians. Activities which will be sponsored including participation by statisticians in meetings of statistical societies (to keep current on new developments in theory and methodology), and may even include allowing statisticians to serve as adjunct professors at local universities, as part of a process of building relationships with those universities. A WCO has a career path for statisticians, of equal validity with its “technical ladder” which is available to those who pursue advancement in the company through technical excellence. Often statisticians will be placed on this technical ladder. In any event, a career path must be created, to make it credible that statisticians will continue to be employed long–term in the organization. The WCO provides internships for students of statistics. This enables them to better understand universities with which they may have a relationship, as well as to examine the student up lose for an extended period of time before considering a permanent job offer. Statisticians in a WCO work to determine what kind of data to share with customers. This includes interpreting customer requests for data, as well as assistance in explaining that data. © 1994, William C. Parr 16
  • 17. Leader for Statistical Methodology Every large company which aspires to WCO status must have a leader for their application of statistical methodology who has direct and frequent access to top management, and who has statistical expertise evidenced by a graduate degree in statistics and multiple years of experience in industry. This person must have the clear respect of the leadership team of the organization. This person should participate in the major management activities of the organization, including being a regular participant in major meetings of the president and staff. In WCOs, although a statistician is seldom a part of the senior staff, they consistently participate in major meetings of the president and staff. This lead statistician should have a close relationship with at least one university which has a clear focus on practical use of statistics. Possible universities which have distinguished themselves in this direction in recent years include the University of Tennessee and Rochester Institute of Technology. These universities are only a partial list, and many good statisticians exist within academia (although those with a true interest in industrial applications are in the substantial minority). This lead statistician should participate in professional societies such as the American Statistical Association and the American Society for Quality Control, to keep abreast of new developments in the field. This involvement can include attendance at society meetings and other activities as appropriate. © 1994, William C. Parr 17
  • 18. Leader for Statistical Methodology • Has routine substantive access to senior executives • Participates in key management meetings • Has strong relationship to at least one university Department of Statistics • Participates in professional statistical societies • Provides advice and guidance to entire organization in use of statistical methods Display 7 Organizing for Improvement in Use of Statistical Methods Organizational structure can have a major impact on the effectiveness of use of statistical methods. Consistent experience across many companies shows the desirability of each major site having at least one lead statistical person, who reports to the site manager. This person should report to the site manager, to make it possible to fill the roles for the site statistician discussed above. A general, more broad principle: Statisticians should be located to maximize interaction with those whom they serve. Specifically, it is almost always a mistake to isolate large statistical groups in corporate headquarters. It is far better for an organization to have at most a small group at corporate, and the rest of the statisticians distributed throughout the organization, tied closely to their customers. © 1994, William C. Parr 18
  • 19. Location is to be interpreted in a three way sense here: i) statisticians should have offices among those they serve, ii) statisticians should be a part of any computer (E-mail, data sharing) networks used by those they serve, and iii) the statisticians and those to whom they report must view their success as having its primary expression in the success of those they serve. Use of Statistical Consultants Statistical consultants have often been a curse in industry. In many cases, they have primarily offered training, often focused on minor variations on control charts and designed experiment techniques, advertised to make improvements in the “efficiency” of the methods. However, statistical consultants can provide a valuable resource. WCOs often bring in statistical consultants periodically to provide advice to their own internal statistical resources, and to their management. As “outsiders” they can often speak to senior executives about matters which are awkward for full time employees to address. In addition, statistical consultants can bring a fresh point of view to examining an organization’s use of statistical methods. Often, company statisticians can become, due to work pressures, consumed by the work of today, and less than optimally aware of emerging trends or capabilities in their industry or others. As organizations evolve into new structural forms, an increasing trend is for some WCOs (none yet in the semiconductor industry have moved very far in this direction, yet) to use consultants to provide added competence of capacity for consulting and training. This can be particularly useful when the consultant has special expertise not possessed by in-house statistical staff, or when the consultant lives local to the corporate site being supported, so the support can be supplied on a regular basis, over an extended period of time. Role of Technology Effective use of technology can yield high leverage for implementation of statistical methods. © 1994, William C. Parr 19
  • 20. Many leading organizations, particularly those in the semiconductor and chemical process industries, have either moved or are now moving to intensive use of the computer for real-time applications of statistical process control. Leading companies, particularly in the semiconductor and chemical process industries, make powerful statistical packages available to their employees, in ways where on-line data can be easily downloaded from CAM systems for analysis purposes. WCOs have integrated actual data on process capability into their design packages, to permit product designers to make choices using producibility as a real criterion. A challenge faced by organizations which have removed paper control charts and moved to on–line charts lies in the possible removal of the real–time graphical interaction of the operator with the control chart which is provided by manual plotting. New means must be found by such organizations to retain the learning events associated with examining the control chart, while retaining the gains due to removing paper charts, and the resultant improvements in accuracy and efficiency. Education and Training Education is a key dimension of a successful statistical effort. WCOs understand the difference between education and training, and provide both, as appropriate. Key areas for education include creating a mindset in the organization which includes variation in its world view. Deterministic views of the world must be removed. Training is a key dimension of a successful statistical effort. WCOs are moving their training from long courses focused on general methods (Course on SPC, Course on DOX, . . . ) to shorter courses focused on specific competencies (How to set up control charts for a new process, how to use control charts in process development, how to quickly set up and quality new equipment, how to identify and prioritize improvement opportunities, how to determine process capability, . . . ). These courses, as a result, are more interactive, and often need to be taught to smaller groups of people. © 1994, William C. Parr 20
  • 21. Training in WCOs is offered closely in time with the need for application. That is, in keeping with the philosophy of JIT, WCOs do not offer long training courses “just in case” the information is needed. Instead, shorter courses (see above) are offered, tied to specific needed competencies. Training in statistical methods should be imbedded in strategies for improvement. To divorce statistical methods from the managerial issues related to their application, or from the “softer tools” which are a part of process management and improvement, is a catastrophic error! WCOs do not provide training in statistical methods in large batches of 40 - 80 hours of training. Instead, they offer smaller modules, focused on specific needed competencies, as discussed above. In a WCO, training is arranged and provided for in a timely manner. Outside resources (consultants, universities, and others) are used to avoid the need to reinvent the wheel, and to allow statisticians within the organization to focus the majority of their time on other activities. In a WCO, training itself is viewed as a process which can be improved. As a result, it is continually reexamined to seek out opportunities for improvement. WCOs recognize that training is not a cure-all. Instead, it merely removes the obstacle of lack of knowledge, and can help slightly with issues of motivation. The key differentiation of the WCO, as discussed above, lies in the behavior of the senior executives of the WCO, and the clearly indicated expectations voiced by those executives for sound management in the presence of variation. Control Chart Methods A common peril, avoided by WCOs, is the proliferation of ad hoc methods for computing control limits. Limits which ignore the methods of rational subgrouping and the theories originally developed by Walter Shewhart and since advanced by others will, in many cases, totally fail to distinguish between special and common causes. Particularly prominent among such ad hoc methods, which must be avoided, is the practice of computing general summary statistics (sample variances over data obtained from a process over a period of time) and using these quantities to form the basis of the © 1994, William C. Parr 21
  • 22. control limit calculations. Clearly, any method which, for instance, takes 30 subgroups of size 3, and then computes the sample standard deviation of these 90 numbers as a quantity of any relevance in control limit computation, is to be avoided. Feedback Mechanisms Frequent feedback is essential for initiating, sustaining, and enhancing the profitable use of statistical methods. Typical feedback methods in WCOs include: weekly unit meetings to discuss applications of statistical methods, quarterly “celebration meetings” in which successful applications are discussed, circulation of write-ups of successful efforts, and incorporation of the need for competence in managing variation into employee development plans. Other, less common but highly impactful methods employed in some WCOs, include: executives asking key, focused questions about the capability of existing and proposed equipment at capital review meetings, intelligently using statistical terms and concepts. Another high impact executive action is leveraging an understanding of the theory of variation to ask insightful questions about subgrouping strategies being employed. A third is to ask insightful questions about the validity of assumptions that “assignable causes” have been determined, including “Do we know what we did which caused these improvements?” Illusions and Distractions Many illusions and distractions exist with which organizations have been able to distract themselves from true progress in use of statistical work. Most WCOs have experience with falling in with one or more of these distractions. However, WCOs are able to emerge from these mistakes and move forward. We itemize some of the most common distractions below. Counting control charts, and viewing a large number as a sign of progress, is a common distraction for organizations early in their experiences with statistical methods. This is truly a case of IBNR (Interesting But Not Relevant) in action! © 1994, William C. Parr 22
  • 23. Another distraction is avoiding using simple control charts and two level fractional factorial designs, but continual pursuit of so-called advanced methods (CUSUM, EWMA, more sophisticated designs, . . . ) to the exclusion of the simpler methods. The simpler methods are the ones which should be in common use. The more sophisticated are (in some cases) appropriate, but have much more limited domains of application. They should be viewed, at least in the early stages of an organization’s experiences with them, as specialists’ tools, with the simpler methods being those which should be in the competence and common practice of all. Still another distraction is exaltation of the control chart and its use without a context mapping from customer value to key process variables to prioritization of what should be worked, long before the first control chart. Much basic work needs to be done to understand where to focus, and this work can be done with simple tools, before even the first control chart. A control chart for a process for which we have not made a flow chart is truly a case of the cart being put in front of the horse! Some organizations obsess on training. They keep their statisticians busy half or more of their time in front of the classroom, with a resultant lack of time for supporting existing statistical practice in the company, advising management on how they can move statistical practice forward, and leading the advance in practice throughout the company. Some organizations have their statisticians become analysts. That is, the statisticians focus on analyzing data — yield data, data from the CAM system, . . . All this in the hope of finding the statistical needle in the haystack of variation which will lead the organization forward to prosperity. Instead, the statisticians should be working to enable and create competence in others, and to assist others. In the words of Deming, following Shewhart (Out of the Crisis, page 471) “Industry in America needs thousands of statistically minded engineers, chemists, physicists, doctors of medicine, purchasing agents, managers. Fortunately, anyone in these fields can learn to use in many problems simple but powerful methods of statistics, and can understand the statistical principles behind them, without becoming a statistician. Guidance from a theoretical statistician is necessary, however. Without such guidance, wrong and costly practices take root, and some problems of production and distribution may be overlooked entirely.” © 1994, William C. Parr 23
  • 24. Another distraction for some organizations has been to turn the statistical staff into general internal consultants on “Quality” or “TQM.” It must be recognized that statisticians have no certain advantage in these areas. Although their tools are a key part of the toolbox, they are not the whole. Summary and a Look to the Future In this paper I have attempted to paint a picture of some of the practices which are most closely linked to successful use of statistical methods, and which are in the practice of many World Class Organizations already. If I were forced to pick three of the broad categories to emphasize in conclusion, they would be 1) the discussion related to the leader for statistical methods, and 2) the discussion related to the role of corporate management, and 3) the discussion related to the role of plant management. These three are clearly the driving areas for emphasis for an organization desiring to be World Class in their use of statistical methods for improvement. The other categories are, largely, consequences of these three. Any organization wishing to move forward in their effectiveness would do well to pay close attention to these three categories, and strive to strengthen themselves in these areas. To fixate on the other categories, without emphasizing these three, would be a misdirection of effort and resources. What is likely to change over time? These three categories will only change by virtue of becoming more important, as organizations become more lean, and clear direction becomes increasingly vital for those organizations. Training and development in statistical methods will continue its present movement to short, modular, competence based units. Fewer and fewer organizations will maintain long (40 hour or more) training programs. Statisticians will more and more find themselves moving out of corporate headquarters (except for a small nucleus of competence, including the leader for statistical methods) and collocating with those whom they serve. © 1994, William C. Parr 24
  • 25. Statisticians will continue to find themselves expected to be experts in areas broader than statistics: modern management methods, quality, reengineering, and ISO 9000 being only a sampling of these areas. The future for statistics is bright. It can and will be bright for statisticians, also, if they adapt to the new realities. Wise organizations will find such statisticians, and develop others. Those of us who work both in universities and with industry look forward to working with such organizations as they move forward. © 1994, William C. Parr 25
  • 26. References Deming, W. Edwards (1986). Out of the Crisis. Especially Chapter 16, “Organization for Improvement of Quality and Productivity. MIT CAES. This is the best source currently in print, on the subject of this paper. Readers of the current paper who are not familiar with Deming’s thoughts on this subject should secure a copy of Out of the Crisis and read Chapter 16 in conjunction with this paper. Parr, William C. (1994). How do Word Class Organizations Use Statistical Thinking and Methods? Invited opening presentation to 1994 Fall Technical Conference, Birmingham, Alabama. Overheads from this talk and an associated paper focusing on the semiconductor industry are available on request. Parr, William C. and Hild, Cheryl (1994). Maintaining Focus of the Organization. To appear in Quality Progress. A discussion of some of the ways organizations distract themselves from moving forward. Many of the distractions have direct import for the subject of this paper. Snee, Ronald D. (1991). Can Statisticians Meet the Challenge of Total Quality? Quality Progress, pages 60 - 64. Focuses on the role of the statistician, looking at attributes of statisticians who will be able to contribute successfully in a WCO. Although some of these attributes are appropriate for any statistician in a WCO, some of the them are likely to only be requirements for the leader for statistical methodology in the organization. Wheeler, Donald (1993). Understanding Variation: The Key to Managing Chaos. SPC Press, Inc. A splendid treatment of the ideas of statistical thinking, written targeteted at the plant manager. An excellent source for helping plant managers begin to understand how variation affects them and the work they do. © 1994, William C. Parr 26

×