We first isolate when and where the problem change took place. We do this by doing a time plot or a position plot of every measurement of the process or product we have that is indicative of the change. From these plots we can often define the time and/or the position of the change to a very narrow range. If the change indicated by the plot is large compared with other data changes before and after the incidence and the timing corresponds to the observed problem recognition, it is generally worthwhile to check for correlations.

The next thing to do is to look for correlations with input variables, often using graphs of historical data. If we don't know the KPIVs, we must do a fishbone diagram or a process flow ...

Get Statistics for Six Sigma Made Easy now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.