Six Sigma Psychology - Part 2
More Six Sigma Stupidity
- Dr Tony Burns, M. McLean
Terence McKenna’s discussion on the
importance of logical thinking (1) reminded me of my last article, “The
Psychology of Six Sigma” (2). McKenna points out that if people were able to
think logically they would not be concerned by nonsense such as inter-galactic
pro bono proctologists. In fact, about 6 million Americans actually claim they
were dematerialized and beamed into the alien mother ship where they were
probed in every orifice by naked little green men. According to a National
Geographic study (3), 77% of Americans believe them, or at least they believe
the little green men have been among us.
Terence states: “Now it’s time to refine
our mathematical skills and not be afraid to denounce pernicious forms of
foolishness.” It is ironic that quality should be at the heart of logical
thinking but the converse has been true. The key to logical thinking is The
Scientific Method, which has been sadly absent from considerations of Six Sigma
™ *. Terence may be more comedian than scientist but he makes some very good
points.
The Psychology of Six Sigma
On average, Quality Digest articles (a
check of the last 50) have had 0.3 responders giving one line of comments. “The
Psychology of Six Sigma” elicited 60 times as many responders as the average
and well over 400 times as much response content. Happily, there were far more
responders who were supportive rather than neutral and negative. It is
wonderful to see people starting to think. It is even great to see a couple of
responders being so moved to attack me personally. At least they have woken up.
I felt it worth a follow up.
I’ve often wondered how quality managers
have approached their CEO’s. Perhaps it goes something like this: “Yes, we know
that the ‘six sigma’ of Six Sigma is nonsense. We have decided to turn our
backs on the world’s leading process statisticians, such as Dr Shewhart, Dr
Deming, Dr Taguchi, Dr Wheeler, with PhD’s in statistics. Instead we plan to
worship the pretentious ramblings of a man with no relevant qualification or
experience. We have no idea why, but they say it is ‘breakthrough’.” Perhaps we
should be checking that the body of the person next to us is ‘one owner only’?
Six Sigma Crash
Thankfully, according to Google Trends,
interest in Six Sigma has crashed 70% in the past decade and its home has been
pushed out to 3rd world countries such as India.
Of course many still believe in the ‘six
sigma’ of Six Sigma. Others will scoff and say that Six Sigma is much more than
just ‘six sigma’. Six sigma is that drifting, shifting, correction, adjustment
thingy that all processes are supposed to suffer. One responder stated that the
Six Sigma originator pointed out “… he had never intended for 1.5 to become a
production metric.“ Well! Throw away those six sigma tables right now! Trash
those thousands of six sigma calculators!
The Six Sigma founder goes on: “…he had
used 1.5 as a "worst-case-scenario SWAG". A little obfuscation there
to make it sound important but undeterred, we head for our acronym finder. 95
hits. Surely it can’t be: “Still Wondering And Guessing”. Maybe it is “Sold
Without A Guarantee”? “Stupid Wild-Ass Guess” sounds a lot more likely.
Worst Case Six Sigma
Let’s move along to the
"worst-case-scenario” part. Is +/- 1.5 sigma really some sort of worst
case for a drift? Harry really does claim so here (4) “The 1.5 sigma shift
factor’s … origin can be mathematically derived as an equivalent quantity
representing or otherwise reflecting the worst case error.” Astounding! Harry
is actually claiming that processes can’t drift by more than 1.5 sigma. What a
relief managers must feel. No matter what the process does, simply broaden the
spec limits a tad and she’ll be apples. Let’s just think about this for a
moment. Harry supposedly theoretically derives the worst thing that can
happen to your process and gives you a theoretical way to overcome it. Doesn’t
that fill you with confidence, particularly when the shallow theory is such
nonsense (4,5,6,7,8)?
In reality there is no limit to how bad
it can get. Dr Wheeler gives a beautiful story of what should be one of the
world’s best controlled processes, NB10. (9) NB10 is a National Bureau of
Standards weight standard that was weighed weekly. It showed special cause
variations of up to 8 sigma!
Out of interest, the International
Prototype Kilogram has also shown large mass excursions over time (10). It has
prompted the Avogadro Project (10b) to develop a new, calculated weight
standard, based on a multi-million dollar spherical crystal of pure silicon 28.
Pity help service industries where “Complicated
non-routine high stress task” has 1 error in 4 or “Routine task with
care needed” has over 1 error in 100 operations. (22) If the most precise
process in the world has errors greater than the supposed “Six Sigma worst
case”, what chance do they have? No doubt they already see the ridiculous 3.4
dpmo as a bad joke. Fortunately, if service industries learn what quality
really means, they can throw Six Sigma into the trash can where it deserves to
be.
Sustained Invisibility
Next we have from a responder: “What
many people don't understand is that the use of that table … assumes that you
have a sustained 1.5 sigma shift that somehow remains undetected.”
There’s no definition of what “sustained” is supposed to be but we find the
definition as “continuing for an extended period or without interruption.”
Looking up “extended period” doesn’t help but we might assume some sort of
unintended permanent-ish step change to the process, putting the process
totally out of control. We turn our backs on the process so the change “remains
undetected” but somehow a six sigma table is supposed to fix everything.
Crazily, this also implies that no longer is the nonsense +/- 1.5 sigma but
just + 1.5 sigma. Now seriously, can anyone keep a straight face and claim this
is evidence for use of 1.5 sigma?
One responder implied that originally
six sigma simply meant Cp of 2. Pity it didn’t stop at that before the
ridiculous 1.5 sigma drifts and shifts crept in.
Short Term Stupidity
One responder raises the old “short
term” vs “long term”: “As for the Sigma shift, the main concern was how much
product performance / process variation was inflating over the long term
relative to the short term.” You won’t find anyone game enough to stick their
heads out and say exactly what “long term” and “short term” are supposed to be.
However, we can track it down. These terms derive from M Harry’s “Z shift”
equation (11). This equation is derived from the sums of squares equality where
for a groups of data points:
Overall variation = Variation between
groups + variation within groups
Or SSoverall = SSbetween + SSwithin
... where “SS” represents sums of
squares.
As we’ll see later, this typifies Six
Sigma, in that this equation dispenses with time. Who cares about the march of
time in business and production anyway? That is, the sequence of data sets is
irrelevant to the above equality. Despite this, Harry astoundingly renames the
SSwithin figure “short term” and the SSoverall figure “long term”!
To add further to the farce, both these
figures use exactly the same data! In other words, in this context there is no
“short term” nor “long term”. (12) He goes on “For the common case ng = 30 and
a type I decision error probability of .005, the equivalent mean shift will be
approximately 1.5S.st.“ Ah, there’s that magic 1.5 again! “g” is the subgroup
size, say the “common case” of 5, and “n” is the number of subgroups, which he
claims is supposed to be “commonly” 6 test points. This gives the ludicrous
situation of “long term” being 6 sample times!
This mangled mess relates directly to
the Shewhart Chart and shows a complete lack of understanding. A Shewhart Chart
compares local variation at each point in time, with variation overall. In this
way, local subgroups contain routine variation but are largely uncontaminated
by special cause variation over time. This approach provides a way to detect
special cause variation over time. There is no “short term”. There is no “long
term”. We simply have whatever time over which the data is collected. There are
no probabilities, no hypothesis tests, no Z shifts, no bell shaped
distributions and not even The Central Limit Theorem is needed. (20)
Power Curve Ball
Responders love Six Sigma’s shifts and
drifts “… it can be explained why a shift of 1 to 2 standard deviations may be
missed for a short period of time.” and “This concept of "drift" is
the amount of deviation a process can deviate BEFORE it is identified.“ There
is no mystery about process changes and the detection of them. It certainly
needs no “corrections” or “adjustments”. +/-1.5 sigma plays no role of any
kind, for any company, in any situation. Dr Wheeler provides a wealth of many
pages of what are called power function curves, that compare the effects of
different detection rules and their combinations, as well as subgroup sizes in
detecting process changes. (13)
Incredibly, as if stacks of disks, chi
square, dynamic mean offset, “worst case”, weren’t sufficient supportive
attempts (4,5,6,7,8), there is yet more 1.5 stupidity. Wait for it, yes, you
guessed it, this time it’s the power function curves (14). He forgets to
mention just which power function curve you take, but read off 50% on the
ordinate for a subgroup size of 4 and there it is, simple as that, 1.5! Who
wouldn’t fall for six sigma with an explanation like that! That pesky 1.5 just
keeps popping up like crop circles.
Monte Carlo Madness
Like stepping into another dimension, we
enter the twilight zone and see emerging from the murk, yet another dauntless
“proof” of the 1.5. (15) Harry makes this one sound especially mysterious and
it really needs twilight zone music to do it justice. It is his “Monte Carlo
Simulation”. Boy does that sound impressive: “Such a corrective device should
be methodically applied when attempting to establish a short-term qualification
standard deviation during the course of a DPQ.” Nope, I haven’t made a typo …
and there’s that “short term” again. He constructs sufficient obfuscation to
ensure you have no chance of reproducing his little Excel graph. However, his
party pooper ex-partner Reigle Stewart spills the beans here.(16) Once again,
stop and think about this for a minute. By typing a standard random number
generator into Excel, you are supposed to be able to “prove” what is the very
worst that can happen to your real world process, no matter what your
process happens to be. Isn’t Excel wonderful! Definitely worth wasting a few
minutes of your time on it, just for a good belly laugh.
Temporal Time
These days Harry hangs his hat on his
old Chi Square “proof” for the “1.5 sigma shift”, in his summary (15). It’s
easy to show how silly it is. (4) You need to wade through a swamp of drivel in
his “proof” such as “The basic nature of temporal effect is time dependent.”,
and gems such as “Although the typical shift factor will tend to 1.5 over many
heterogeneous CTQs each CTQ will retain its own unique magnitude of dynamic
variance expansion (expressed in the form of a dynamic mean offset)”. Now
that’s enough, stop laughing, this is serious.
Back to the responders. “First are alpha
and beta risk.” No there is not. People fail to understand that while based on
probability, control charts are not probability charts and do not depend on any
probability model. (17)
Motorola and TQM
A couple of responders mention Motorola.
They forgot to mention that Motorola’s growth and winning the Baldridge Award
was the result of 8 years work using quality circles, TQM and BPI. Six Sigma
was tacked on afterwards. (18) In the next decade Motorola fell off the cliff.
A Baldridge examiner said to me recently
that people were turning to Lean, not because of the six sigma nonsense, but
because the maths is too hard. How sad is that! The maths isn’t too hard. It is
simply that unscrupulous consultants make more money out of long training
courses teaching volumes of stuff that has no relevance to processes. As Dr
Deming stated: “Analysis of variance, t-test, confidence intervals, and other
statistical techniques taught in the books, however interesting, are
inappropriate because they provide no basis for prediction and because they
bury the information contained in the order of production. Most, if not all
computer packages for analysis of data, as they are called, provide flagrant
examples of inefficiency.” (19)
It is so trivially easy to tear “six
sigma” to shreds yet the believers keep lining up to be milked. I should say
that whenever I think such thoughts, Halon’s Razor comes to mind. "Never
attribute to malice that which is adequately explained by stupidity."
Normal Nonsense
Take control charts for example. The
first PC wasn’t invented till 40 years after the invention of the control
chart. There was no normalization and no need for it. 76 years after the
invention of the control chart, by examining 1143 different distributions, Dr
Wheeler proved that simple Shewhart Charts work for essentially all processes.
(20) However, normalizations by computer are endemic today and are not only not
needed, they lose the meaning of the data. The meaning and information
contained in the once simple histogram is most often these days buried by a
meaningless and fictitious distribution curve drawn over the top of it. Such
curves are useless and serve no purpose. It takes 3200 data points to know a
distribution out to 2.95 sigma and by the time you collect the data the process
will have changed (20). Even for those folk wanting, and permitted, to fiddle
and experiment with their processes, there is no need to be befuddled by the
complexities of ANOVA. Dr Wheeler shows a far simpler and more effective
approach. (21) There is clearly a profound need to throw away the nonsense, the
irrelevant and unnecessary complexities, and get back to basics.
Dozens of DMAICs
Some responders claim that Six Sigma is
more than six sigma and it is DMAIC that is key. People forget that 25 years
ago there were dozens of different versions. Every company had their own
5,6,7,8 step or whatever, methodology. All have their origins in The Scientific
Method. One responder agrees “the core of DMAIC is the scientific method.” The
Scientific Method combines inductive and deductive reasoning. Learning the
basis of The Scientific Method is far more important than any derivative.
Ad Hominem
An ad hominem attack was inevitable. It
it a sign of desperation. No doubt some will cling to their six sigma tables to
the death. For those for whom the messenger is more important than the message,
my PhD was in Chemical Engineering; my BE was also in Chemical Engineering,
with 1st class honours; I was Dux of my high school. However, when it comes to
process statistics, I do not consider myself close to being in the same league
as Dr Wheeler. I have great respect for Dr Wheeler but have no financial
connection with him nor SPC Press. One stated “Dr. Burns should back up his
assertions.“ This article has sufficient references to satisfy anyone who has
any doubts about six sigma stupidity. Sadly though, I have no doubt that no one
will check them. People don’t think and investigate, despite professing to
adhere to methodologies. If people did follow The Scientific Method or any
other methodology, Six Sigma would never have got off the ground.
In summary, ask yourself, why does Dr
Wheeler attempt to clarify, simplify and facilitate, when Six Six consultants
attempt to obfuscate and add cost and complexity? Why does Dr Wheeler, the
world's leading process statistician, call Six Sigma "goofy"? Ask yourself why have there
been so many attempts to prop up “six sigma” and “+/-1.5 sigma” … is it to
maintain the consultants’ money flow or just stupidity? McKenna would say that
many quality managers are unable to separate shite from shinola. They hurl
themselves off the cliff like Walt Disney lemmings, into the Six Sigma abyss.
It is time to THINK. It is time to actually start using The Scientific Method,
firstly by checking the references to what you have just been reading. It is
time to trash the nonsense and get back to basics. The basics are quality as it
should be, taught by Dr Shewhart, Dr Taguchi, Dr Deming, and Dr Wheeler, with
Lean adding to Deming’s comments on waste.
References
1. Terence Mckenna denounces Relativism
https://www.youtube.com/watch?v=YK3BahMxH4M
2. Six Sigma Psychology – Dr Burns http://www.qualitydigest.com/inside/quality-insider-column/six-sigma-psychology.html
3. ABC News http://abcnews.go.com/Technology/ufos-exist-americans-national-geographic-survey/story?id=16661311
4. Six Sigma Lessons from Deming, Part 1
- Dr Burns http://www.qualitydigest.com/inside/six-sigma-article/six-sigma-lessons-deming-part-1#
5. Six Sigma Lessons from Deming, Part 2
- Dr Burns
http://www.qualitydigest.com/inside/six-sigma-article/six-sigma-lessons-deming-part-2#
6. Sick Sigma - Dr Burns http://www.qualitydigest.com/node/5900
7. The Tail Wagging It’s Dog - Dr Burns http://www.qualitydigest.com/inside/six-sigma-article/sick-sigma-part-2
8. The Six Sigma Zone - Dr Wheelerhttp://www.spcpress.com/pdf/DJW177.pdf
9. Good Limits from Bad Data – Dr
Wheeler http://www.qualitydigest.com/inside/quality-insider-column/good-limits-bad-data.html
10. Kilogram - Stability of the
international prototype kilogram https://en.wikipedia.org/wiki/Kilogram
10b. Avogadro Project https://www.nist.gov/physical-measurement-laboratory/silicon-spheres-and-international-avogadro-project
11. Z Shift https://www.isixsigma.com/ask-dr-mikel-harry/why-15-subracted-short-term-estimate-capability-zst-get-long-term-estimate-capability-zlt/
12. Short term - long term https://www.isixsigma.com/ask-dr-mikel-harry/why-15-subracted-short-term-estimate-capability-zst-get-long-term-estimate-capability-zlt/
13 Power Functions for Control Charts.
Ch 9. “Advanced Topics in SPC” - Dr D Wheeler
14. Statistical Reason for the 1.5 Shift
- Davis R. Bothe
15. “Resolving the Mysteries of Six
Sigma” - Mikel Harry.
16. Monte Carlo in Excel https://www.isixsigma.com/topic/1-5-shift-simulation/
17. Advanced Topics in SPC – Dr Wheeler
1995 SPC Press
18. The Mists of Six Sigma - Alan Ramias
http://www.performancedesignlab.com/the-myths-of-six-sigma
19. Out of the Crisis. - Dr Deming. 1986
20. Normality and the Process Behaviour
Chart – Dr Wheeler SPC Press 2000.
21. The Analysis of Experimental Data -
Dr Wheelerhttp://www.qualitydigest.com/inside/quality-insider-column/analysis-experimental-data.html
22 How to Build Ultra-High Reliability
Work Processes - Sondalini http://www.lifetime-reliability.com/free-articles/work-quality-assurance/Ultra_High_Reliability_Work_Processes.pdf
* Six Sigma is a Motorola US Service and
Trade Mark, although Mikel Harry claims he created it.
