Six Sigma is about reducing costly variation in manufacturing and business processes. From a statistical perspective, Six Sigma uses the normal distribution model, which predicts 3.4 defects-per-million (DPM) for processes that have six standard deviations between the process average and the nearest specification limit (see image below). Of course, no process follows the normal distribution model perfectly, so the 3.4 DPM prediction is only theoretical.  The important point is that Six Sigma processes have extra “cushion” between the process results (data) and the specification limits, making them less likely to produce defects over the long run.

The Six Sigma philosophy also assumes that the process mean (average) will drift over time, moving the outer edges (also called “tails”) of the process results closer to the specification limits.  This process drift is assumed to be a maximum of 1.5 “sigmas” (sigma = standard deviation = a measure of variability that is calculated from our process data), and the 3.4 DPM level takes this theoretical mean shift into account.  It is unlikely that any given process drifts by exactly 1.5 standard deviations over the long run – some processes may not drift at all, and some may drift more than 1.5 standard deviations – but the assumption of a 1.5 sigma mean-shift is better than ignoring mean shift as a factor in long term variation.

While books have been written explaining the theory behind Six Sigma’s probability model, the real power of Six Sigma lies in its DMAIC approach to process improvement.  The DMAIC model is a collection of quality improvement tools that are combined into a 5-step approach:  Define -> Measure -> Analyze -> Improve -> Control.

Six Sigma Process