RE: SIX SIGMA

[Rip Stauffer <rstauffer@bluefirepartners.com>]

I'd like to second (and maybe amplify) what John Zavacki pointed out:

"Six Sigma is not the statistical concept of .002 ppm vs. 3.4 ppm, nor is it
a methodology in and of itself.  It is a synthesis of Juran's 'Project by
Project' improvement, Kaizen, and solid industrial engineering methods
brought together with the System of Profound Knowledge at it's core.  It
depends on measureables attained to keep it alive in the eyes of the top
level management, to keep it interesting to the people affected by it, and
to sustain the 'priestly caste' of the black belt."

If you try to understand Six Sigma by traditional statistical terms, you
will probably go down the road that I originally traveled. Although it is
based on that statistical concept, it is NOT, (as John points out) the same
thing. Striving for six sigma quality means that you are striving for an
aggregate equivalent to six sigma. 

What does this mean? Well, take Wheeler's Tokai Rika control chart. If you
do the math, you will find that the workers at Tokai Rika achieved
capability which, if characterized by traditional methods, would give you a
Cpk of close to 2.5. 17000 lighter sockets per day for months and months,
probably no defects at all. SPC will get you there, if you use it.

However, suppose you are producing the dashboard for a car, or the interior,
or even the car as a whole. How many opportunities are there for mistakes or
defects in the production of that dashboard, interior, or whole car? It's
not enough to say, "All our components are manufactured to Cpk's of 2 or
better (although that would be great)." There are other opportunities for
problems, such as assembly mistakes. Poke Yoke can help prevent those, but
may not eliminate all of them. 

The Six Sigma metric (and I do wish they had come up with some other name,
but we seem stuck with this one, if only through inertia) requires study to
identify all the opportunities for producing a defect, given some entity of
interest (component, assembly, subassembly, etc.) Suppose I have a
subassembly that has, through the process of producing it, 50 opportunities
for a defect. If 20000 of those subassemblies are produced, I have 1 million
opportunities to produce defects.

This is where the confusion comes in. You estimate sigma levels based on the
number of produced defects you estimate (by whatever means--it will vary
depending on the type of defect). I know that's backwards, and we don't live
on the mathematical plane. No serious statistician would claim that it
represents reality. But it can make for a convenient, rule-of-thumb type
metric for communicating current quality levels to upper management. As John
pointed out, "the AIM is to improve constantly and forever the system in
which we work."

As far as the 1.5 sigma shift is concerned, Motorola supposedly had a ton of
data to support that. I have, however, read one paper that suggested that a
closer look at the research would tell another story. This author pointed
out that Motorola had only used one rule for signal detection in their
control charts. He looked at them using the Western Electric Zone tests, and
was able to quickly detect much smaller shifts than 1.5 sigma. This
certainly jives with what I have learned about Shewhart charts. Wheeler did
considerable research to demonstrate the power of Shewhart charts with 4
detection rules to detect pretty small shifts (see "Advanced Topics in
Statistical Process Control").

Do you NEED the Six Sigma "umbrella?" Depends. If you are successful with
implementing the Deming method without ever mentioning Six Sigma (except in
its real statistical sense), then great! Certainly all the tools are the
same, and a lot of the methods. Dr. Deming's philosophy is the most powerful
management idea I know. For better or worse, though, I don't see many
companies looking for it (how would they know?)

At the very least, Six Sigma is a foot in the door, and it is raising
quality awareness. I attended a recent Six Sigma conference in Miami, and
heard speakers from Motorola, GE Medical, and a number of other companies
who had a few years' Six Sigma experience. One thing that struck me was that
almost all the more "mature" companies had learned a lot about quality. I
made a list of the 14 points in my notebook and checked them off as the
speakers mentioned them. Perhaps all the companies hadn't learned all 14,
but it seemed that a couple had, and almost all of them seemed to be on
their way. The important thing was, they were learning. They were getting
there the long way, maybe, but they were getting there. 

In the wrong hands, Six Sigma is "Zero Defects on steroids" and could be
very dangerous. It currently has some questionable people in its Vanguard,
and there is a pretty large controversy within ASQ over its endorsement of
those questionable people. But I agree with John's assessment, that it "is a
marketable application of the SoPK and it needs DEN membership to support
it." There are also some good people, Deming advocates and disciples, trying
to work to embrace and improve it. 

Just my two cents.

Rip

Rip Stauffer, Senior Consultant
BlueFire Partners
1300 Fifth St. Towers, 150 So. Fifth St.
Minneapolis, MN 55402
612-344-1027
rstauffer@bluefirepartners.com
==================================================================