Featured Product
This Week in Quality Digest Live
Management Features
Mike Figliuolo
Sure, you have to be professional, but have a good time anyway
Margaret Graziano
Unlocking the power of organizational culture
Graham Ward
Asserting yourself and setting clear boundaries
Henning Piezunka
Businesses and leaders influence the kinds of ideas they receive without even realizing it
Chris Caldwell
Significant breakthroughs are required, but fully automated facilities are in the future

More Features

Management News
A tool to help detect sinister email
Developing tools to measure and improve trustworthiness
Manufacturers embrace quality management to improve operations, minimize risk
How well are women supported after landing technical positions?
Adds increased focus on governance
Survey shows 85% of top performers rely on it to achieve business objectives
Key takeaways from Marcum’s 2023 National Manufacturing Survey

More News

Davis Balestracci


‘Which of Deming’s 14 Points Should I Start With?’

Answer: none of them

Published: Monday, November 14, 2016 - 14:51

Have you ever heard something like: “I’m committed to Dr. Deming’s approach [or Six Sigma or lean or TPS, it doesn’t matter], but executives don’t seem to listen anymore. All they do is keep interrupting my very clear explanations with, ‘Show me some results, then show me what to do.’ I was shocked that my demonstration of the red bead experiment neither awed nor convinced them; several of them even walked out during it. Which of Dr. Deming’s 14 Points should I start with to get their attention and the results they want?”

My answer would indeed be: None of them—and all of them!

If anyone either continues to ask that question or is confused by my answer, please read this, then heed the following advice from Deming himself. Why? Because you don’t quite get his message.

From Ron McCoy’s book, The Best of Deming (SPC Press, 1994):
• “We are being ruined by best efforts.”
• “Judging people does not help them.”
• “If you stay in this world, you will never learn another one.”
• “Does experience help? No! Not if we are doing the wrong things.”
• “There is nothing more costly than a hack.”

‘Off we go to the Milky Way!’... again

How many of you have to endure quarterly review meetings—the dreaded “account for” results vs. goals, usually arbitrary numerical ones—and the ensuing, “What are you going to do about [insert specific negative variance here]?” Let me suggest how you can easily amuse yourself during these dreadful meetings, then get a ton of respect and credibility from your colleagues after the meeting.

Missed appointments for physical therapy at a medical center were an ongoing costly problem. The national standard was 20 percent, and the department supervisor had been asked at the end of the previous year to stretch and set a “tough” goal. She decided on half of the national standard—10 percent.

Here are some actual data presented at a year-end review, which as you can see, is afflicted with the current toxic plague of “red, yellow, green” stoplight data presentation.

Click here for large image.

On the surface, the year’s performance wasn’t too bad: nine greens, one yellow, and two reds. But someone astutely observed, “Both reds were in the second half of the year, though. After July’s red, there was a nice trend down. Good work! But the trend of the last couple of months and a red December aren’t good signs. Can’t you do what you did in August, September, and October again?”

This red December performance cast a real pall on things. She was grilled about the disturbing trend and why it was so high so late in the year. Doubt began to creep in as to whether her improvement efforts were effective—they had obviously slipped.

This was reinforced when the indicator’s overall yearly average performance of 10 percent was “up” when compared to the previous year’s performance of 9.4 percent (the two boxes in the lower-left corner). Even more pointed questions resulted, and there was discussion about making the goal even “tougher” for next year.

Déjà vu?

The alternative? Simpler than you might think.

The data are right in front of you in the bottom row of the table. It takes only a few minutes to sketch a run chart (a time-ordered plot with a data median drawn in as a reference line):

Even though the data are limited, nothing looks amiss. There is no presence of any trends—i.e., five or six successive increases or decreases. Neither is there a run of eight consecutive points, either all above or all below the median, that would indicate a shift—e.g., a possible improvement if it happens early in the data above the median, or if it happens later in the data below the median.

You could take another five minutes and come up with the following process behavior chart, an Individuals chart. (Note to any nitpickers: I’m intentionally not using a p-chart.)

The process has been stable the entire year, which is common cause: All data points are within limits, and no special cause tests are triggered. Currently, any one month’s performance will randomly fluctuate between five and 15 percent, which encompasses all three traffic lights’ alleged special cause endpoints. Additionally, any one month can differ from its immediate predecessor by as much as 6.3 percent.

In other words, each data point is merely statistical variation on a process perfectly designed to produce, consistently, 10 percent cancellations/no shows.

So the result of all the department supervisor’s hard effort has been...?

Oh, and the math required to create this chart? Basic multiplication and addition as well as the ability to count to eight, subtract two numbers, and sort a list of numbers from lowest to highest.

No belt required.

Do you realize that your process is perfectly designed to get the process results you’re already getting? Unless you know this, any well-intended but ultimately unsuccessful efforts to improve such a stable process is treating common cause as special. “Efforts to improve the process” have now become part of both the process’s inputs and another component of its natural common-cause variation.

This scenario was one of several given to me when a large organization asked me to speak at a lean Six Sigma conference. By then, it was eight months into the next year, and I was able to add these additional data to the previous year, which resulted in the following process behavior chart:

I wasn’t sure what the new goal was (it really doesn’t matter). Regardless, I saw no change from the previous year (the last moving range isn’t a special cause; it’s <6.3). Incorporating these additional data into the calculations hardly changed the common-cause limits. The previous year’s graph with its 12 data points was a good enough initial estimate of the situation. So much for the people who insist you must have 20 to 30 data points for an accurate chart.

Then I got really curious and asked whether they had any more data. They gave me the data for the year prior to that of the first chart above, which resulted in this process behavior chart for all 32 months:

There is no convincing evidence that anything had changed during the past 32 months. After presenting this to a room full of lean and Six Sigma practitioners, I was met with a stunned silence.

What should you do?

Let me first tell you two things not to do

First, there is a widespread, naive assumption that a chart showing only common cause indicates the need for a total process redesign. This is not necessarily true (and usually not).

You must resist any knee-jerk urge either to form a process redesign team or brainstorm a cause-and-effect diagram to answer the question, “Why do we have cancellations or no shows?”

Have you all made my past mistake of facilitating the similar, cause-and-effect-diagram-from-hell that will surely result?

Second, please, please tell me you’ve gotten beyond the pedantic response, “Deming says that common cause means it’s management’s fault, and it’s up to them to fix it.”

I’m ashamed to admit that I have been guilty of this approach—30 years ago. Funny thing, no executive ever said thank you. Besides, Deming never said that. It’s the common phenomenon of more than 30 years of his Funnel Rule No. 4 manifesting from something he originally said. Is it any wonder why leadership might turn a deaf ear to you, especially if the approach gets vague results?

So what should one do after constructing this plot?

Hint: By not working on meeting the goal, you will meet the goal.

Similar irony: By not working specifically on any of the 14 Points while using the appropriate common-cause strategy, you will now be able to work on all of them.

Confused? How about even more confusion: The department head has been meeting the 10-percent goal the entire 32 months.

To be continued.


About The Author

Davis Balestracci’s picture

Davis Balestracci

Davis Balestracci is a past chair of ASQ’s statistics division. He has synthesized W. Edwards Deming’s philosophy as Deming intended—as an approach to leadership—in the second edition of Data Sanity (Medical Group Management Association, 2015), with a foreword by Donald Berwick, M.D. Shipped free or as an ebook, Data Sanity offers a new way of thinking using a common organizational language based in process and understanding variation (data sanity), applied to everyday data and management. It also integrates Balestracci’s 20 years of studying organizational psychology into an “improvement as built in” approach as opposed to most current “quality as bolt-on” programs. Balestracci would love to wake up your conferences with his dynamic style and entertaining insights into the places where process, statistics, organizational culture, and quality meet.


Deming's 14 Points

Right on Davis.  In the UW-Madison College of Engineering, one of my statistics professors knew that I came into the program with over 10 years applying SPC.  He asked, How many data points are needed to estimate limits?  Twenty-five, I quickly informed all of my classmates.  Professor Wu, then pointed out to the class that I was wrong.  The answer is two.  I went on to teach statistics in the College of Engineering and never forgot the lessons of Wu, Bisgaard, Ermer, and the most wonderful George Box who quoted Cole Porter's song lyrics Experiment.  "Experiment. . . And it will lead you to the light."  Simply put, use statistics to explore. 

Data points needed to estimate limits!

The answer by the professor is two.

I do not quite understand this.

Can you pl enlighten me !

Regards and thanks


Estimating Limits

Two data points provide a mean and a standard deviation.  While recognizing that they are a limited sample size, we still have a ballpark estimate of a process.  We move on from there.  Also, consider that with two data points we can conduct Monte Carlo simulations for whatever hypothesized distribution we may want to explore.  Claiming that an absolute number of data points is needed, say 25 or 30, before we can estimate control limits is the stuff statistical immaturity.  Needless to say, however, there are a lot of caveats that experienced statistician/engineers know how to conceptualize and act on.