I am familiar with the MSA method of using the average of multiple measurements to cut down the %R/R figure. But I recall, from a previous job, that the R/R number can be used to determine a GUARDBAND. Example, spec limit +/- 10 cm. R/R= 2 cm. Hence modify the limits to +/- 9 cm.

Is there a statistical justification for th is?

Is there a statistical justification for this?

Guardbands are a somewhat common practice.

I am not sure if this is statistical valid, but it is probably a good practice. By modifying your 'manufacturing' tolerance range versus your 'acceptable tolerance' range, you are probably assuring your customer receives a product that is suitable and acceptable.

Only one word of caution. Be certain that when a process capability is performed, it is clear against which tolerance the capability is calculated against. It may cause confusion if this is not done.

Regards,
Don

------------------
Just the ramblings of an Old Wizard Warrior.