RE: Statistical Control and Purpose

[David Kerridge <d.kerridge@lineone.net>]

Although it is some way from the original thread, the new discussion
of whether statistical control is "natural" is very interesting. It came up
in some investigations I did for Deming's NYU research seminars.

Statistical control seems to arise in two different ways. The first is when
things become as chaotic as they can within some physical limits. For
example, if we throw dice, we will get a process that is under statistical
control if we shake them thoroughly enough. If we don't shake them
enough we get results that are *less* chaotic than those of the well-
shaken dice.

Many atomic processes, as Walter Shewhart observed, are under
statistical control. If we look at any example of statistical control in
physics we find it is when the entropy of a process has reached the
limits set by the system - it is as *large* as it can get.

The sort of statistical control that we aim for is almost the opposite
of this. It is when we have reduced the chaos in a system until we
have made the variation as *small* as it can get. Again, we are up
against a limit, set by physical constraints of some kind, though they
are harder to see.

Very likely some processes in some organisations may be
as chaotic as they can get. Chaos cannot grow indefinitely if
for example, there are feedback loops, creating a system of
permanent tampering - as with rule two of the funnel. This
drives variation underground rather than removing it. And in
general, many management practices limit variation rather
than removing its causes.

But statistical control of this kind is most undesirable. Once we
start improving the process, and dismantling the artificial limits
on chaos, special causes will become visible, and can be eliminated.

Eventually we will achieve the desirable form of statistical control,
when chaos is a little as possible, with the given system.

The only connection between these two forms of statistical
control is that they arise when there is a limit, upper or lower,
to chaos, and that chaos is, as it were, pressed against the limit.
========================================================================