By John Hunter, author of the Curious Cat Management Improvement Blog (since 2004).
Deming Lecture at the 2012 ASA Joint – Quality Improvement: From Autos and Chips to Nano and Bio” by Jeff Wu, Georgia Institute of Technology.
One of the points Dr. Wu mentioned is Shewhart called common causes, chance causes, and special causes, assignable causes. I must say I have always thought assignable causes seems like a better name to me. It makes it more obvious that using a “what caused this one result” strategy when the result was within the control limits (and thus not an indication of an assignable cause result) isn’t a good idea. But for now special cause is still the commonly used terminology (I wouldn’t be amazed if this reverts to Shewhart’s original term at some point).
Dr. Wu also points out
Deming’s major insight was to recognize that Shewhart’s idea [control charts, chance and assignable causes etc.], originally in manufacturing, can also be applied to enterprises
One of my favorite statistical tools/strategies is Design of Experiments (DoE) (which my father, Bill Hunter, worked with a great deal) and Dr. Wu discusses the value of using this strategy.
If the process is in control but with low process capability, use Design of Experiments to further reduce variation.
I discussed Deming and DoE a bit in a previous post: Statistical Techniques Allow Management to do a Better Job (about a paper Dr. Deming wrote).
DoE can also be used for purposes other than reducing variation but his talk focuses on Taguchi’s ideas on using DoE to reduce variation.
I liked this statement about a failure mode experienced by users:
You scream and kick the machine to no avail.
Does this describe any processes in your organization?
Another quote I liked (when looking at results of a simple DoE effort that reduced error from target by 75%):
When you run design of experiments, even very simple things, you really produce magic.
This is one of the keys for management improvement: great results don’t require finding some secret strategy all that is needed is to apply powerful quality tools, concepts and strategies that have been around for decades. And still most organizations continue to ignore the ideas that can produce magic, if you just apply them.
It is true that what Dr. Wu calls “very simple things” are simple for Design of Experiments but getting an understanding of DoE does take a bit of effort. It isn’t one of the easier statistical tools to use, but like control charts it really doesn’t require much to use well. They both can seem complex at first but if you have a good system they can be applied when they are the right tool to use very efficiently by those working with the process with great results. A bit of expertise in setting up the experiments properly is often a bit more useful with DoE but even that is not that difficult (certainly not requiring you to be comfortable with the complex formulas shown in the video).
Related: A Historical Look at Deming’s Career: Lecture by J. Stuart Hunter – Special Cause Signal Isn’t Proof A Special Cause Exists – Using Deming’s Management Methods to Enhance the Application of Taguchi’s Ideas (podcast with Bill Bellows) – Root Cause, Interactions, Robustness and Design of Experiments