Talk:Six Sigma/Archive 1

'''This is the first archive for Talk:Six Sigma, covering discussion from January 2004 to August 2005. Further discussion should take place at the current talk page.'''

This was from the Software engineering talk page.

Six Sigma: gotta vent
I'm not sure Six Sigma belongs here at all. Here's a typical example of six sigma silliness:

http://www.questdiagnostics.com/brand/b_home_six_sigma.html


 * Six Sigma stands for 3.66 defects per one million opportunities.


 * 3 Sigma 66,8 defects per one million opportunities
 * 4 Sigma 6,210 defects per one million opportunities
 * 5 Sigma 233 defects per one million opportunities
 * 6 Sigma 3.4 defects per one million opportunities
 * Good companies are 3-4 sigma, but good is not good enough in healthcare. We believe that the healthcare industry must strive for perfection.

What can one say of a methodology that does not recognize a difference between "3.4 defects per million" and "perfection?"

And what can one say of a metholody that doesn't seem to get the numbers right? Unless, of course, Excel is wrong... 2*normdist(-6,0,1,true) = 1.98e-9. In other words, the areas under the tails of a normal distribution 6&sigma; from the mean are 1.98 per BILLION, not 3.4 per million.

Apparently "six sigma" is just empty bragging about how great they are, with no true quantitative meaning.

Meanwhile, why, exactly, is 3.4 defects per million the right number to strive for? Why isn't it a nice round number like 1 per million? Or, say, 1.98 per billion (six sigma, in other words?)

It's things like this that give the word "methodology" a bad name. Dpbsmith 01:20, 5 Jan 2004 (UTC)


 * Meanwhile, why, exactly, is 3.4 defects per million the right number to strive for? Why isn't it a nice round number like 1 per million? Or, say, 1.98 per billion (six sigma, in other words?)


 * "Defect distributions" in probability approxmately follow Normal distribution. However, it has been observed that there are many scenarios where a random sample of defects do not fall under the normal distribution. In quality processes, we sometimes need to "force-fit" the normal distribution on to the sample graph, thereby introducing "statistical noise". Motorola apparently decided that one needs to set a "tolerance range" of 1.5 sigma(Standard deviation) for such graphs. So, the 3.4 per million you see, is actually "4.5 Sigma (standard deviation)" :-)


 * This step of "introducing tolerance" makes a lot of sense in manufacturing processes, where you need to assemble 1000 different parts, each with a preset accuracy. For want of a better example, assume we have a process A which yields 99.4% precision and a process B which also yields 99.4% precision. Each on their own, stand the test of 2 - Sigma. But when process B has to operate on the output of process A, we get an overall precision of 98.8% which is inferior to 2-Sigma. So, even when all processes taken individually seem to conform to the standards set, there is no asserting that the end product will stand the test. In terms of pure practical measurability, this is a distinct disadvantage - not the least because of the dynamic nature of the process itself. One way to eliminate such "marginal problems" is to offset the resultant accuracy expectation to a respectable distance away from "borderline cases" (in this case, offset it by 1.5 sigma). This does not eliminate the headache, but st least statistically tackles the majority case. This step is sometimes necessary, also in the sense that consistently detecting any less is may not be possible. Whether this step is "equally relevant" to all kinds of software processes is very debatable, indeed. chance 08:26, Jan 5, 2004 (UTC)


 * Perhaps this last entry should be moved to the Six Sigma talk page.


 * Dpbsmith seems to be taking out his or her Angst against 6s on the article itself.  Skepticism about the value of the methodology is legitimate and should be reflected in the article.   (I haven't read the whole thing to tell if it is.)   But I take, "I'm not sure Six Sigma belongs here at all," to imply that this article is illegitimate.   For good or ill, 6s is at very least a well-known methdology fad.   It may be more than that, but it's certainly that, and as such belongs in Wikipedia.   [EDIT: On further review it's not clear to me whether Dpbsmith's comment is referring to the Six Sigma article or to some part of Software Engineering that was moved here. ] - PhilipR 14:21, 3 Jun 2005 (UTC)

Most of the text of the article is devoted to an explanation of why it's called "six sigma" against all seeming reason and knowledge of the normal distribution. Note that the text makes a nod to the idea that six-sigma is nonsense relative to software engineering. Improvements to that text are invited. In fact, in the view of its author the present text is too pussy-footing and polite. Dandrake 02:52, Feb 24, 2004 (UTC)

the wikipedia article, along with all the top results in google, are nicely complicated. that's ok, it's a complicated subject with a long and changing history. but what is lacking, from wikipedia AND from the top results, is just the facts. can someone write or does anyone have a link to a page with just the factual numbers, with a six-sigma bell curve illustration? here's one, but it only shows to 3-sigma and doesn't have all the numbers: http://www.robertniles.com/stats/stdev.shtml

However market leaders ...
However market leaders have measurably reached six sigmas in numerous processes.

What "market leaders"? In what market? This sentence is really bloody annoying. Isn't this an example of Weasel words?

--221.249.13.34 05:58, 10 Dec 2004 (UTC)

Using a normal distributon table, 6 sigma actually translates to about 2 defects per billion opportunities only.

But how we get 3.4 defects per million opportunities, which we normally define as 6 sigma? Motorola has determined, through years of process and data collection, that processes vary and drift over time - what they call the Long-Term Dynamic Mean Variation. This variation typically falls between 1.4 and 1.6.

After a process has been improved using the Six Sigma DMAIC methodology, we calculate the process standard deviation and sigma value. These are considered to be short-term values because the data only contains common cause variation -- DMAIC projects and the associated collection of process data occur over a period of months, rather than years.

Long-term data, on the other hand, contains common cause variation and special (or assignable) cause variation. Because short-term data does not contain this special cause variation, it will typically be of a higher process capability than the long-term data. This difference is the 1.5 sigma shift. Given adequate process data, you can determine the factor most appropriate for your process.

"By offsetting normal distribution by a 1.5 standard deviation on either side, the adjustment takes into account what happens to every process over many cycles of manufacturing… Simply put, accommodating shift and drift is our 'fudge factor,' or a way to allow for unexpected errors or movement over time. Using 1.5 sigma as a standard deviation gives us a strong advantage in improving quality not only in industrial process and designs, but in commercial processes as well. It allows us to design products and services that are relatively impervious, or 'robust,' to natural, unavoidable sources of variation in processes, components, and materials."

More Comments
There are numerous examples of companies that have achieved at or near six sigma levels of quality. The hard drive on your computer has a bit error rate that is much better than six sigma... usually 9 or 10 sigma. Airline safety exceeds six sigma. Many automotive part suppliers are at close to six sigma quality levels.

Yes, the whole business of calling 3.4 DPPM "six sigma" is silly in the extreme. It was probably invented for marketing purposes. However, six sigma as a program, properly done, is anything but silly. It can profoundly improve a company's competitive position.

Why Six?
The debate towards the end of the Why Six section should be resolved I think (doesn't that kind of thing belong in Talk anyway?). I lack the expertise to do a proper job, but if noone else jumps in I'll have a bash in the next day or so.

--Stephenh 12:28, 12 Jan 2005 (UTC)

hmmm
This link is a partial explanation: http://www.qualitydigest.com/dec97/html/motsix.html

From what I can piece together (note that I am not a six-sigma-black-belt, but I have a brown-belt in seeing through marketing hype):

Part to part variation: 6 sigma of samples within acceptable tolerances (ie 2 per billion are outside spec)

Long term variation: This is quoted at 1.5 sigma (defined by Motorola as an aim, I guess). I believe what is being said is that the long term variation degrades the original six-sigma-of-samples-within-tolerances to 4.5-sigma-of-samples-within-tolerances. This gives a combination of 4.5 and 9 sigma error, of which the 4.5 sigma variation dominates to give 3.4 per million.

So where does this leave us? Motorola's process appears to relate to their customers (ie final delivered item) and not to (say) their transistors (1 million transistors needs more than 4.5 sigma per transistor!).

This is a long way from any mathematical definition, as far as I can tell.

capitalization / punctuation
I did a little work on the first section; as always feel free to do whatever's necessary to improve upon my changes. My inference is that Six Sigma (both cap) is the name of the methodology, whereas six-sigma is an adjective meaning "corresponding to the sixth standard deviation". Therefore, in one instance I corrected Nine Sigma to nine-sigma, b/c AFAIK no one has published a metholodolgy called Nine Sigma. Does the community agree with me? - PhilipR 14:24, 3 Jun 2005 (UTC)

Why six sigma, again
Has anyone noticed that this discussion is a mess? Yes, some have, but why not do something?

The comments arguing against 6-sigma look not unreasonable, but it's the sort of treatment that (as I think someone has already said) belong in the Talk page, not the text. To begin with, it jumps into Cp and Cpk without a hint of definition. Foo. Should I venture in and start the exdit wars, or submit it to peer review by people who handle statistics better than I? Trying to figure that out.

And then there's the amazing little gem I reproduce here:


 * When many parts have to fit together, tolerances actually work in the favor of the manufacturer. It is quite possible to make six sigma assemblies out of three sigma parts, since it is highly unlikely that all parts will simultaneously be at one extreme of the tolerance range. Intelligently allocating variation is called "Statistical Tolerancing", and is a useful part of Design for Six Sigma.

What on Earth is this supposed to mean? What it seems to mean is that when you put two parts together, each of which has variations with some standard deviation, the combination has a smaller standard deviation. Talk about goofy!

Concretely, let an assembly have some measurement with a tolerance of 5. It is made from two pieces, and their deviations from spec are phsyically additive. Suppose part A is off by 3.0. Part B is off by 2.1 in the same direction: result, off by 5.1, out of spec. Part B is off in the opposite direction: result, off by 0.9, in spec. The measurement might, for instance, be the diameter of a hole and of the pin that fits in it. This is the sort of thing we're addressing, right?

Now, if A and B each have a sigma of 1, the sigma for the combination isn't less than 1 because errors will cancel out so often; it's more, because they won't cancel out so often. In fact, it's about 1.4 (square root of 2). Does anybody who deals with statistics not know that? For those who need a refresher,.

I'm about to replace this with something that isn't flagrantly wrong. If I'm missing the point completely, someone tell me, please.

Dandrake 19:13, Jun 3, 2005 (UTC)

Things To Keep in mind
1. The article is not a talk page. Hypotheticals, questions, arguments, etc. should be minimized.

2. NPOV - It doesn't matter if the topic is daft, just marketing, etc. Describe what it *is*, not how you feel about it.

Suggestion:

Describe and explain from adherent's point of view. Raise points from detractors. Give treatment to rebuttals.

Wikipedia is not in the business of assigning validity to any idea/practice - it merely describes things. -- 23:13, 4 Jun 2005 (UTC)


 * Good advice. In fact, it's precisely my agenda for the "Why six?" section. Now if anyone responds to my request for peer review with some useful and technically savvy review of the section, things will be improving. --Dandrake 00:48, Jun 5, 2005 (UTC)


 * Forgive the "Amen" comment, but I just wanted to say that I'm glad Dandrake is taking steps to move it in an NPOV/more accurate direction. - PhilipR 13:16, 6 Jun 2005 (UTC)

RfC
Wikipedia policy is clear enough on the matter of criteria for excluding links. Spangineer is making a good-faith attempt to offer reasonable and consistent criteria to establish what is link-spamming and what is not — he should be treated with according civility. If anyone feels that there are shortcomings in reasonably prominent external links, readers should be given some indication of the issues in the entry via NPOV presentation of external authorities (i.e. do not try to make a point of your own or present original research). If a site is so far outside of the mainstream that controversies it generates are not regarded as notable, the site probably ought not be linked. Adding links to an external site from someone who is not a recognised authority allows subversion of these editorial policies: if you want to contribute on the subject for wikipedia, please contribute here and under wikipedia policies and guidlines. Anything else is subject to charges of self-promotion. Buffyg 15:55, 2 August 2005 (UTC)


 * This article is NOT as accurate as it could be and now shows signs of gross biasness as well as some immaterial information and links to marketing companies. The information provided does not reflect on anything other than a 'self admitted ignorance' of the topic by the people involved in the writing it. There is NOT enough information for people to know what six sigma is neither about nor of the original concepts as conceived by Bill Smith and Motorola. It is an insult to the late Bill Smith and all the pioneers of six sigma. As already proven, any attempt to change it to reflect more real world information will be undone by these highly biased people. Posted by 69.220.197.99 --Spangineer (háblame)  14:13, August 12, 2005 (UTC)


 * Actually, I mentioned above that it would be great if we could do some work on this article to improve it. By all means, feel free to change/add information and cite your sources.  If you do and you follow relevant wikipedia policy (NPOV, etc.), I won't complain. --Spangineer  (háblame)  14:13, August 12, 2005 (UTC)