Pareto Analysis Doesn't Have To Be "Supercharged"
While I agree with much of what William J. Stevenson says in "Supercharging Your Pareto Analysis" (October 2000, p. 51), it bothers me that quality professionals continue to imply that a Pareto analysis can be anything other than what Stevenson calls "supercharged." As Stevenson points out, "A Pareto analysis is intended to distinguish the 'vital few' factors from the 'trivial many' factors ... ." While the few and many are obvious when nothing but the frequency of occurrence of a variable of interest is portrayed on a chart, what identifies the vital and trivial?
Despite what the article implies, rearranging a histogram doesn't make it a Pareto chart. It's still just a histogram, albeit one arranged in descending order. Stevenson identifies two conditions when frequency of occurrence provides sufficient information, but I disagree that we should allow anyone to call the result a Pareto analysis. When the two identified conditions aren't met, Pareto is needed--not supercharged Pareto, just Pareto.
We in the quality profession have too often permitted halfhearted implementation of quality tools. We need to make it known when a tool isn't being used properly, and we need to show how to use the tool properly in the first place. We especially need those who teach such tools to teach them completely. Not only can a Pareto chart based on frequencies alone be misleading, it isn't even a Pareto chart.
ROBERT M. CRAVEN
The reality is that Pareto charts are often constructed using frequencies. In some cases that is okay, and in others it isn't. The purpose of my article was to address and clarify this issue by answering these three questions:
1. When is it okay to use frequencies?
2. When isn't it okay to use frequencies?
3. How do you proceed when it isn't okay to use frequencies?
WILLIAM J. STEVENSON
Canadian Salary Information Helpful to Reader
I just want to congratulate Quality Progress on the 2000 salary survey ("2000 Quality Progress Salary Survey," November 2000, p. 31). As a quality consultant, I found it interesting to study the gaps in salaries not only in Canada, but also between the United States and Canada.
Last year my boss asked me to gather salary surveys for consultants and develop a salary and benefit plan for my company. I had a copy of ASQ's 1999 salary survey, but I couldn't use it because the gaps between the two countries were unknown to me. Instead I had to rely on a study from Ordre des Ingénieurs du Québec, which reports engineering salaries in all fields. I was glad to receive this year's ASQ survey because it focused on the quality profession and included Canadian salaries.
Common Sense Must Supplement Quality
I would like to answer the questions at the end of the introductory section of the article "Tire Failures, SUV Rollovers Put Quality on Trial" (December 2000, p. 31). I have been in the quality field for more than 20 years. I don't work for a large corporation, but our customer base does include some of the largest global off-road construction and agriculture conglomerates in the world. My feeling is that, yes, I suppose there is a role for Six Sigma in all this, but there are more fundamental issues to be dealt with first.
Bridgestone/Firestone and Ford hold, or held, a public trust. They are entrusted to produce safe, quality products that consumers should be taking for granted by now. Should I have to worry that my vehicle may suddenly roll over, no matter what the cause? Should I be concerned that my tires have been produced with proper care throughout the entire process?
Common sense and an inherent feeling of responsibility to the consumer must supplement quality. Until these fundamental values are brought back to manufacturing and the service industry, there will be a feeling that we need to have sophisticated quality systems to replace simple human value systems.
Automotive Industry Is What's on Trial
The December 2000 issue of Quality Progress features an article titled "Tire Failures, SUV Rollovers Put Quality on Trial" (p. 31). Quality is not what's on trial; rather it's an industry that puts itself above ASQ and the quality professional by stating, "Prior automotive experience required."
I believe that this same enlightened industry has recalled more product than it manufactured in 1999. Perhaps its inbreeding reflects a requirement for familiarity in conducting recalls.
The definition of quality goes deeper than buzz words such as "audit" and "management standards." The promises of the current panacea, Six Sigma, as a standalone approach will only follow statistical process control and zero defects into the repertory of the quality professional. It is unfortunate that quality must be put on trial for the shortcomings of one industry.
EDWARD R. MARKOVIC
What Matters Most: Quality Or Shipping Schedules?
I read the article "Tire Failures, SUV Rollovers Put Quality on Trial" in the December 2000 issue of Quality Progress (p. 31) and was happy to see ASQ providing some insight into the situation.
The responsibility falls on both manufacturers (Ford and Firestone). The inadequacies at Firestone lie in design testing and shelf life. To elaborate, test to failure should have been conducted under more extensive tire pressures and load conditions. This would have increased the odds of discovering the defect(s) and given the auto manufacturers more accurate information regarding tire pressures.
In addition, there were reports from the Decatur, IL, plant workers that materials exceeding their assigned shelf life were used. Shelf life faults can be attributed to inadequate in-house audits, poor material review board (MRB) decisions or management override of the MRB.
Based on the information I have regarding Ford's role, I see two design concerns:
1. Improper weight displacement.
2. Wheel base width.
This implies inadequate quality assurance input during the design and testing stages.
However, the major cause of the problem, quality vs. shipping schedules, can be attributed to both manufacturers. I have been in quality assurance for 25 years and have worked for many major companies. In all cases, the schedule came before quality. If this problem could be solved, then quality related issues would decline greatly.
Quality Efforts Need Industrial Rationalization
I want to respond to Susan Daniels' final comments in the introductory section of "Tire Failures, SUV Rollovers Put Quality on Trial" (December 2000, p. 31). First, Daniels states "standards are just the first step on the road to quality." This implies that QS-9000 or ISO 9000 certified companies do not have effective quality systems. Only an uninformed person would imply that Firestone's reliability problem exists because it is beginning the long, long road to quality. ISO 9000 and QS-9000 are extensions of decades of effective quality management systems.
Second, can Six Sigma help? Poor use of statistical process control (SPC), for example, is endemic in today's industries, and the use of control charts comprised of nonrational subgroups is widespread. There is nothing in the essence of Six Sigma that indicates that Master Black Belts will have the proper training in SPC and design of experiments to properly apply such tools. To be an SPC expert requires a university-level engineering course and years of experience using control charts.
Eighty to 120 hours of Black Belt training are woefully inadequate to qualify the recipients for quality engineering or quality management positions. In fact, I consider it unbefitting for ASQ to imply such. Any loss of life due to such assertions should bear consequences.
Furthermore, there is no such thing as zero defects. Planes still fall out of the sky, and tires still fail. All zero defects accomplishes is the allocation of precious resources to focus on the trivial many instead of the vital few. Here again, if Firestone was focusing on zero defects, it would surely have missed the cause of the tire failures.
As for continuous improvement, it makes sense, but only in the context of how the Japanese developed their industrial and economic system. For the Japanese, continuous improvement was not a means to an end. It was an end to a means. Continuous improvement was the outcome of what the Japanese began in 1934: industrial rationalization. Industrial rationalization is what's missing in our quality efforts.
Cross Plains, WI
Audits Should Examine Nonconformity Records
Dennis Arter's article "Internal, External Quality Audits Couldn't Have Prevented Recall" ("Tire Failures, SUV Rollovers Put Quality on Trial," December 2000, p. 31) was off base from my perspective. I have performed many supplier quality audits for PacifiCorp and was president of the Western Utility Association for Supplier Quality (WUASQ), and I can't imagine a supplier quality audit that doesn't include examination of nonconformity corrective action records. How a supplier deals with nonconformities as reported by other customers is indicative of how the supplier will deal with your issues.
Most suppliers have been open with us in revealing their nonconformities and corrective actions. This is a vital audit point for us, and any company not willing to share this information will not be considered as a supplier. The fact that the supplier had nonconformities is not important to us; rather what's important is how the nonconformities were resolved.
WUASQ sponsored supplier audits have been loosely based on the new ISO 9000 standard since the first draft was released. Paragraphs 8.3, 8.4 and 8.5 of ISO 9001:2000 or elements 4.1, 4.13 and 4.14 of ISO 9001:1994 should be critical points in any ISO 9001 based audit. How can a supplier be certified without an examination of these points?
Arter's mention of the obstacles to good audit practice was revealing. It's amazing that any industry would set up internal and external rules that consider potential legal blunders more important than customers' lives.
Salt Lake City, UT
I guess I failed to state my reasons early enough in the article. They were listed following the first paragraph.
No conformance or corrective action system can function without data. What I tried to say, albeit poorly, was that the data were not available to those who might have been in a position to see the patterns emerging.
Some Managers Only Worry About Bottom Line
I get so angry every time I hear about quality related fatalities. I have worked in the quality profession for 15 years and have seen many sneaky acts performed by upper management at companies I previously worked for. It looks like the bottom line is the only thing that some managers are worried about. I've actually seen managers hide a customer's product so the customer wouldn't see and want to audit it. Upper management just views quality as overhead with no real value, but those of us in the quality field know that is not true.
I wonder when upper management will finally realize that building quality into the production system will increase profits and provide customer satisfaction. I have been through several ISO 9000 inspections, and each time an inspector came, management made a complete turnaround regarding quality habits at the company. ISO 9000 inspectors need to be more in-depth when questioning personnel.
Error in December "Statistics Roundtable"
I found an error in the December 2000 "Statistics Roundtable" (J. Stuart Hunter, "Metrics for Uncertainty," p. 72) that I thought I should bring to your attention. In the second column, near the top of the first page, the author states that "log10Odds = bels (10 bels is a decibel)." This is an error; 10 bels would be a dekabel. A decibel is 0.1 bel.
San Luis Obispo, CA
DeWayne Sharp is correct. If we define a bel as log10(Odds), then 10 of these would certainly be a dekabel. What is termed a "bel" in the article is a centibel or one hundredth of a true bel. It is embarrassing to be caught using incorrect units in an article on measurement. However, whether we quantify log(Odds) in a unit called a centibel, decibel, bel, dekabel, bit or napier, the quantities in the selected unit still add and subtract. That is the essential point of the article.
Interestingly, the correct technical definition of a decibel is still 10log10(Odds). But why take a simple expression such as log10(Odds), multiply it by 10 and call the result, 10log10(Odds), a decibel? The answer rests in engineering convenience. As the probability scale moves from 0.5 (zero bels) to 0.9, we gain 9.54 decibels or approximately 1 bel. As the probability scale changes from 0.9 to 0.99, we move up to 19.95 decibels or almost 2 bels. Moving along the probability scale from 0.99 to 0.999 takes us up to 3 bels, and at 0.9999 we reach 4 bels. Appending nines to probability is basically equivalent to increasing the number of bels. The beauty of the bels scale is that it adds and subtracts.