Blog

Notes, articles, and updates.

More Articles

Apr 17, 2003 · 1 min read

Six Sigma and Software Engineering & Reliability

I recently finished reading the book "What is six-sigma?" by Peter Pande, and Larry Holpp. In terms of Software Engineering, Six Sigma is much more than a specific analysis of software reliability. It is a quality improvement framework and mindset focused on the measurement of process variation as the culprit for lack of quality. I want to point out that the term "six sigma" when used in conjunction with software reliability, has little or nothing to do with statistics, with distributions, with their moments, etc. It is a buzzword and will remain a buzzword until such a time as it is defined in statistically correct ways. ** The real Sense for Six Sigma** Six Sigma as the name implies stands for six standard deviations from the mean. Sigma is a statistical measure of variability around the average. The concept of Six Sigma comes from reliability engineering prediction of system or component failure probabilities. For example: the wearout time of a component may be normally distributed - that is mean - standard deviation. So, we want a component having a very small of failure before its design life. If we set this at one sigma from the mean we get ~80% reliability, 2 sigmas gives us ~95%, 3 sigmas ~99%, and so on. Six Sigma gives us ~99.9997% relaibility - near perfect. Or in other ways 3.4 defects per million.