Tuesday, January 12, 2010

On 6 Sigma

My father told me an exceedingly long and dry story yesterday about data analysis but which actually had an interesting point.

The background is that in statistics any typical test results can be graphed along a cosine, the bell curve. 85% of results fall in the middle, then the outlying 15% will fall 7.5% to either extreme. Apparently "1 Sigma" refers to 31% or a 69% defect rate among products. "2 Sigma" drops to 31% defect, then levels 3-6 jump to the 93.3 percentile, 99.38%, 99.977% and 99.99966%. There exists a 7 Sigma at 99.9999981% but that's a pretty moot divergence from 100%.

Anyway it seems a guy my father used to work with was ordering parts from a distributor at one Sigma level and he was looking to move up a notch. He asked the distributor, "Hey, we're getting parts at like 95% defect-free. Do you think you could get us 97%?"

The distributor told the guy, "Oh, sure! You know, I actually have great quality control. I mean I could get you 100% defect-free if you wanted. You just always asked for 95% free so every time you ordered 100 parts I just threw in five defective ones."

Apparently, the moral of the story, kiddies, is always ask your cosine for what you really want. Don't be afraid and don't short change yourself.

Because cosines are dicks.

No comments :

Post a Comment

Note: Only a member of this blog may post a comment.