June 2002

Vol. 4, No. 6

Six Sigma and More: 
David Schwinn reflects on Deming’s insistence on on-job training.

by David R. Schwinn

Considering the far-reaching impact of W. Edwards Deming’s approach to management, it is instructive to return to Deming’s 14 Points for Managers and what he called “deadly diseases” in organizations. “Institute training on the job,” Deming’s Point #6 is the focus of this column.
In the mid-1980’s Mike Cleary and I attended a conference for statisticians organized by George Washington University. It was titled something like “Statistics for Statisticians” and taught by W. Edwards Deming. The room was filled, mostly with college instructors in statistics. I was overwhelmed at the end of the conference by the large number of participants who approached Deming, confessing that they had spent their whole careers leading their students astray.
What was it that brought those instructors to that conclusion? During the conference, Deming had stepped back, examined underlying theories of statistics and offered a “new” branch of statistics. In his presentation, he did not introduce new information. He simply framed what they already knew in a profound, new way.
Deming reexamined the traditional mainstream approaches, descriptive and inferential statistics. Descriptive statistics helps you understand a group of items if you can count, measure, and/or operationally define all of the items in the group. If you want to understand the group by only counting, measuring, and/or operationally defining a sample of the group, then you must use inferential statistics. Both of these approaches help you understand what you have. The methods can also help you understand how the group compares to preset standards. They can help you understand how one group compares to another. They can even help you make a decision about what to do with any of the groups about which you have this knowledge. Descriptive and inferential statistics are extremely powerful tools for helping you understand and make decisions about what you have.
Unfortunately, they have very little value in helping you with what you don’t have. And understanding what you don’t have and making decisions intended to affect the future is the job of most managers and, for that matter, many people who are not in traditional management positions. Most statisticians understand this. Dr. Deming helped the seminar participants understand that by focusing on these approaches in the classroom, most of THEIR students did not comprehend the limitations of what they were learning. Their students went out to the world of work and made decisions using descriptive and inferential statistics erroneously.
Dr. Deming suggested that a much more powerful approach was to use analytic statistics, a phrase, I believe, he invented. Analytic statistics is the use of control charts. Dr. Deming reviewed that analytic statistics permit you to examine how a system behaves over time. If you understand how a system behaves over time and it is statistically stable, you can expect that it will continue that behavior unless a significant change is made. You even have a superficial understanding of when and what changes can be considered significant. Without that knowledge, however, a person’s ability to know if they will make or have made significant change is grossly impeded. The bottom line is that, without the use of analytic statistics, you cannot know, in a modern scientific sense, if what you did or did not do made any difference. Unfortunately, that lack of knowledge still does not keep the majority of leaders and managers from claiming that knowledge and acting as if they have it. That is the understanding that devastated the students in that conference.
Now, what has all that got to do with training and Six Sigma? I believe that Six Sigma embraces the need for training in much the same way that Dr. Deming embraced the need for training except in the area of analytic statistics. I understand that Six Sigma embraces analytic statistics but I believe that the difference in the power for improving systems among the three statistical approaches is not well understood. It is useful to remember that Dr. Deming thought that some of the descriptive and inferential statistical tools, such as F-tests and t-tests, which are commonly taught as part of Six Sigma, were a waste of time for those interested in improving systems.
As always, I welcome other thoughts on training on the job and Six Sigma. I'm at

Copyright 2002 PQ Systems.

Please direct questions or problems regarding this web site to the Webmaster.