Dr.
Walter A. Shewhart, the father of
Statistical Process Control, 11 March 1891 - 11 March 1967
"We are not concerned with the functional form of the universe, but merely with the assumption that a universe exists." - Statistical Method from the Viewpoint of Quality Control, Dr. Shewhart
Shewhart was, in many ways, a fairly
unremarkable fellow. He is hardly well known outside of the field of statistics, despite the sheer
beauty and
power of his work. However, his theories defined an entire branch of statistical thought which have affected the entire world.
Having studied
physics at the
University of Illinois and
University of California, the
1920s found
Shewhart working for the
Bell Telephone Laboratories on the
Western Electric Company's manufacturing problems. Every signal that the process was not completely uniform was acted upon, which just made things
worse. From his experience with this factory process, Shewhart developed the basics of
SPC - assignable causes and normal causes,
control charts, and the
Western Elecric Zone Tests.
Shewhart distinguished between common causes, the natural variability of any process, and assignable causes, hiccups in the process which are economically reasonable to seek out and eliminate.
Western Electric's mistake was that common causes were being treated as assignable causes - money and effort was wasted adjusting the process. They needed some way to decide whether a change in their product was a normal, common variation that they could ignore, or a sign of a deeper problem.
Shewhart developed ways of thinking about data which led to ways of treating data which allowed process managers to make this determination. By applying these methods to a process constantly (not just when problems appeared), it is possible to reign a process into a state of control and keep it there. Properly done, this is not only very feasible
economically, it can result in great reductions in overall costs while simultaneously improving the product.
As shown by the quote above, the
robust nature of Shewhart's work is truly amazing. His control charts are capable of taking measurements from a process which is already out of control and does not necessarily even come close to a
normal distribution and still effectively function. Even if the assumptions he based his theories on are violated, they still work! Imagine if
Descartes had said "I think, therefore I am... but even if I'm not, it wouldn't matter."
Unfortunately, Shewhart's ideas did not catch on in his native country very quickly. His colleague,
W. Edwards Deming, eventually despaired of getting SPC to work in America, and took his work to
Japan in 1949 as part of a post-war reconstruction program. So great was his success, that it is widely acknowledged that Shewhart's work was a direct cause Japan's emergence as an economic powerhouse. When applied at all levels within a company, Shewhart and Deming's techniques were revolutionary.
Shewhart also did work in
operationalism and edited the
Wiley Series in Probability and Mathematical Statistics for quite some time, but his major contribution to the world was clearly
statistical process control, for which he received numerous awards and recognitions.
By all accounts, Shewhart was a very polite and
agreeable man, seeking only to further human understanding, and hoping that his ideas would prove useful and be built upon by future generations. Considering how vital SPC is to modern manufacturing and the number of mathematicians who have followed in his footsteps, his hopes were clearly fulfilled.
Sources:
Cutler, http://www.sigma-engineering.co.uk/light/shewhartbiog.htm
Wheeler and Chambers, Understanding Statistical Process Control