RE: Specification Story

[Steven_S_Prevette@rl.gov]

My question is how do you calculate process capability without Specification
Limits?  There seem to be two camps here - the Deming camp saying that
capability is defined by the control limit spread and another camp (I don't
know what to call them) that says you compare the control limits to the
specification limits to assess how capable you are of meeting the specs.  We
need an operational definition of "process capability".  Spec limits can be
useful if used in context of a control chart and stable variation.  Why do
we have to have one or the other - why not both together?
-------

I will attempt to reply for the "Deming Camp"

1.  Yes, there are cases where you have been given a specification.  This
may be a "fact of life" (see page 41 of TNE) or a customer demand, such as
in a contract. 

2.  If the process is not stable, not predictable, any Cpk calculation is
statistically meaningless.  If we are routinely getting results outside of
the control limits, it does no good to compare those control limits to the
specification.  Actually, the control chart must come before the Cpk
calculation.  To do Cpk, you must have a control chart first.  So, yes, you
have both together.

3.  Dr. Deming states on page 226 of TNE "This simple illustration should
put to rest forever use of measures of dispersion like Cpk, as it has no
meaning in terms of loss.  Moreover, it can be decreased to any value merely
by widening the specifications."  The simple illustration referred to
centered vs.. off-center performance.  Dr. Deming uses a Taguchi loss
function on page 225 as a providing much better information.  It may be
economically valid to work to reduce variation and raise Cpk, even beyond
six sigma, but you have to have some loss information in order to know.  Or
we assume that we should continue to reduce variation until we have some
proof it is not economically viable to continue to reduce it (in which case
there are likely many other items that could use variation reduction).

4.  Why do we need the term "process capability", much less a definition of
it?  We can blindly work to meet specifications, or meet an arbitrary Cpk,
or we can understand the economic losses of variation about the nominal
value and work to reduce those losses.  Of course we must meet the
specifications we have been provided, or we won't stay in business.  But we
could still simply meet specifications, and fail.  Many examples exist, from
the simplistic buggy-whip or carburetor scenario, to being completely
outclassed by another supplier who does much more than meet specification.

Steve Prevette
Site Technical Authority for Statistical Trending
Environment, Safety and Health
Fluor Hanford, A Fluor Global Services Company
ASQ Certified Quality Engineer
steven_s_prevette@rl.gov
509-373-9371