Quality Transformation: Statistical thinking for everyone


Popularized in manufacturing, statistical process control must be practiced throughout an organization to assure quality.

I recently presented “Statistical Thinking for OD Professionals.” for the Organization Development Network Annual Conference. It was well received, but more importantly, it reignited in me a passion for the use of control charts in decision-making for everyone who uses numbers, or needs to use them for decisions that guide the future. That reigniting was once again fueled by the November 16, 2016 article in Quality Progress titled, “Understanding Variation – 26 Years Later.” (Nolan, T., Perla, R.J., and Provost, L.).

My passion comes from remembering a few “big, bad” decisions over the years and many smaller mistakes in the misuse or lack of use of appropriate statistical tools. Here are a few:

  • A lack of understanding of Myron Tribus’ Perversity Principle led 44 out of 56 schools in the Atlanta area to change test answers with 11 teachers being convicted of misdoing and, more recently, Wells Fargo’s opening of new accounts for their customers without their customers’ permission or even knowledge. As a reminder, Myron stated:

If you try to improve the performance of a system of people and machines by setting numerical goals and targets for their performance, the system will defeat you and you will pay a price where you did not expect it. (Myron Tribus and Yoshikazu Tsuda, “The Quality Imperative in the New Economic Era,” Quality First, National Institute for Engineering Management & Systems, Alexandria Virginia, 1992).

  • Amazon’s rate, rank, and fire strategy resulted in an average employee tenure of one year. General Electric, well known for that strategy years ago, stopped doing it in the last few years after understanding the negative repercussions of the strategy. These practices obviously indicate a lack of knowledge about common cause variation versus special cause variation.
  • 300,000 veterans supposedly in the U.S. Department of Veterans Affairs system may have died at least in part because of the common understanding of the Perversity Principle: the principle that rewarding or punishing people for performance that lies beyond the system’s capability to perform will generate negative outcomes. Again, “the system will defeat you and you will pay a price where you did not expect it.”

In the “Understanding Variation” article, Nolan, Perla, and Provost noticed that:

  • A The Wall Street Journal headline, for example, announced that “U.S. GDP (Gross National Product) Grew a Disappointing 1.2% in the Second Quarter” as if that were significant. A simple analysis using control charts shows that change, like nearly all the change that The Wall Street Journal and most other financial reporting systems announce, to be common cause variation and of no significant consequence. Their reporting frequently and sadly encourages investors to make buy and sell decisions based on a faulty analysis of the situation.
  • Using a bar graph to indicate nursing facility residents with one or more falls with major injury, the U.S. Department of Health and Human Services concluded in 2011 “when taking this scale of scored values into account…it is easy to see that they are not changing very much from quarter to quarter.” In fact, a control chart analysis indicated a clear worsening in the fatality rate beginning in 2007. Whatever happened in 2007 needs to be reversed if we care about falls. We’ve been paying a price for that error for nearly ten years.
  • The Bureau of Labor Statistics (BLS) showed a color coded map of the U. S. indicating in which states the fatal work injury rate rose, declined, or stayed the same, as if that was useful information. A control chart analysis showed that all states were part of the system’s common cause variation except for North Dakota that had a significantly higher fatality rate. Another missed opportunity…this time for reduced fatalities.

Most of us have learned about Shewhart-developed and Deming-popularized statistical thinking in a manufacturing environment, but, properly done, it can and should be done everywhere. Over the years, I have used these statistical tools to help organizations improve systems characteristics such as:

  • Food waste by weight
  • Overtime
  • Wait time in emergency room
  • Medical error rate
  • Customer or client wait time in person, on the phone or on the internet
  • Accounts receivable
  • Sales
  • Inventory size and/or accuracy
  • Building code violation rate
  • Budget variance
  • Traffic accident rate by number and/or cost
  • Police on the road by number and/or proportion
  • Restroom cleanliness
  • Vehicle downtime
  • Water bill error rate
  • Payroll errors
  • Insurance claim processing time
  • Computer downtime
  • Number of computer clicks or telephone referrals
  • Number of prisoner incidents
  • Employee absenteeism and tardiness
  • Prisoner recidivism
  • Resource usage such as gas, electricity, water and other raw materials
  • Profit

My conclusion is that every day, multitudes of organizations, institutions, governments, and communities miss opportunities to get at least a little better by ignoring special causes of variation and create at least a little more waste, loss, confusion, chaos, frustration, and systems illness by misinterpreting common cause variation that leads to poor decisions.

Nolan, Perla, and Provost suggest that we encourage people and organizations to:

  1. Make data available over time
  2. Provide data in formats that allow for construction of Shewhart charts
  3. Determine whether a process is stable (and react appropriately to improve system performance)
  4. Think carefully and creatively how to stratify data.

That seems pretty straightforward. Go on to use your own favorite statistically-based improvement algorithm such as PDSA, DMAIC, or any of the improvement schemas imbedded in PQ System’s Total Quality Transformation training system. I’m reminded of the innovative ways Bonnie Small and her co-authors showed us how to use control charts (Statistical Quality Control Handbook, Newark, NJ: Western Electric, 1977). Although her examples were all manufacturing-related, the leap beyond manufacturing and quality is not difficult.

I hope you agree we have an opportunity here that is more than worthwhile. Let’s teach how to make statistically-based decisions starting with the use of control charts in our schools and in every other part of the world we touch and care about and, once learned, encourage those folks and institutions to continue to make more intelligent decisions to positively affect our lives and the lives of future generations.

As always, I treasure your comments and questions.

David Schwinn

David Schwinn


  • A serious example of the Perversity Principle may be found in the pharmaceutical industry. Often the variability of QC test methods are such that out of specification test results are too often gotten, a consequence of inadequate test method precision. OOS test results require investigation for root cause, corrective action and a detailed report. But some are reluctant to state inadequate test method precision as a root cause, or they do not know how to discover it.

  • re: I recently presented “Statistical Thinking for OD Professionals.” for the Organization Development Network Annual Conference.”

    How many of these types of presentationsd have you done over the past 30 years? What has been the result?

    Wow on the reference to the AT&T book and Bonnie Small. Agree on the value of the book.

    On the Quality Progress article titled, “Understanding Variation – 26 Years Later:”

    In assessing lessons learned from what has and has not been accomplished over the past 25 years, how about a different question: “Where could we be within the next 25 years?” by applying a better approach?” Call me, operators standing by … 🙂

  • Awesome!! I am using control charts in my work in software and done improvements!! Great way to understand data.

    • Thanks for thoughtful reading of this post and responding to the content! Happy to know that you’re using control charts effectively.