How to select samples for %GR&R Study

Let's interpret GR&R like this way.

Total Variation = MS Variation + Product/Process Variation
= Equipment Variation + Appraiser Variation + Part-to-Part Variation
GR&R= Root Square (Equipment Variation + Appraiser Variation)
%GR&R= Root Square (EV+AV) / Root Square (TV)

If EV+AV is kept as constant, to increase TV through increased Part-to-Part Variation will make %GR&R smaller which means the better performance. What does it mean to increase Part-to-Part variation???

Without doubt, we can not increase Part-to-Part variation as bigger as we wanted. If you want to pass %GR&R study, you can select samples with wide distribution. If you really want to assess your measurement system, you should select samples with narrow distribution.

In addition, 6Sigma brings another index for Measurement Analysis. It's %P/T which means R&R/Tolerance(Spec). It's another method to evaluate if your measurement system precision is enough for the spec.