The Mathematical Flaws of (most) Six Sigma Methods

Walter A. Shewhart was a very bright fellow who came up with some interesting ideas and techniques. These have proven immensely effective when properly applied.

So what the heck is up with Six Sigma? Well, the problem is, these people have taken a number that Shewhart pulled out of his entirely human ass and ignoring most everything he did that was worthwhile. Whether this was ignorance, stupidity, or a little of both is an open question. Now, 3 sigma unit limits to either side of the central line are not without value - but the main reason Shewhart kept them around is because they worked. But they're not based on the Central Limit Theorem. The normal distribution is conspicuously absent. Any figures giving parts per million defective are based upon the assumption of a normal distribution. This is generally all but unprovable, even when correct. It's also not required by the techniques that Six Sigma is derived from. Making a calculation to enough decimal places to give you, say, 3.4 parts per million based on a wild assumption... any good applied mathematician would observe an exponential rise in their own beating you with a big stick behavior.

And as far as five nines vs. six sigma... we're talking about two very, very different things. Five nines is a goal. Six sigma, or rather let's say control charts, are a technique to improve your process. Comparing the two shows a deep and terrible misunderstanding of what Shewhart was talking about. Six sigma should help you strive for that kind of reliability, but they are not the same thing.

Was Enron Really That Surprising?

At my school, only business majors are required to take Making Moral Decisions. You'd think they could at least figure out running a business, but from the sounds of thing, that math geek Shewhart had 'em beat - or at least the Six Sigma goons. He talked almost as much about using his techniques in a reasonable manner in the real world as he did about the math itself. He's very vehement about working towards improvement, not some arbitrary goal which is very likely not even achievable with the current process. Shewhart's limits are set based on the natural variation of the process itself - not the specifications handed down by someone who's probably never seen the factory floor, tech support farm, or whatever the process might be. Management is supposed to work with and for workers - I've read countless examples of this working perfectly.

Six Sigma and the McDojo Phenomenon

Young one, study the ancient and powerful teachings of the old masters. They will allow you to defeat many enemies with your great power! When you have learned all that there is, you will be given your black belt and you will be unstoppable!

McDojo - a term often applied to martial arts training centers that promise you the world, deliver little to nothing but the outside form, take your money, and run away cackling. I'm surprised the Six Sigma folk had the brazen nerve to adopt a colored belt system to their... teaching. It makes the comparison almost too easy. They attempt to tie their techniques to some ancient foundation, rather than allowing them to stand on their own merit. They have adopted some of the outer appearance - six sigma units - but not the core ideas that made the original techniques powerful. Heck, half of them may not even know how wrong they are.

A black belt, I should add, generally means you've mastered the basics of an art and are ready to move on to deeper understanding. And heck, even then, it doesn't mean anything except to you and your teacher. Six Sigma units are just a certain range Shewhart made up that seem to work effectively and economically. There. That's it. I grant you your Six Sigma Bling-do Black Belt. Go read a book on statistical process control and promote yourself up to however many stripes you want. And if you really want to improve your business, go buy a book like Wheeler and Chamber's Understanding Statistical Process Control. Our class worked from this book in the SPC course I took, where I learned most of what I've just talked about. Unless you start getting into the stuff in the back (that you don't need anyway), it's really pretty understandable and easy, even if you're not mathematically inclined.