Gage R&R Study Example
If My Parts Are So Good, Why Is My Gage R&R So Bad?
Almost every week, we get this question from someone trying to conduct a Gage R&R study because their customers are requiring it. They think if their product is good, their Gage R&R should be too.
Hint: Gage R&R doesn't evaluate your product,
it evaluates your ability to measure your product accurately and consistently.
Gage R&R Study Basics
Gage repeatability and reproducibility calculates the total variation (TV) from: 1. appraisers (AV), 2. equipment (EV, i.e., gages) and 3. part variation (PV). It then uses total variation to determine how much variation is attributable to appraisers and equipment (%R&R).
If your parts are identical, there is NO Part Variation and all of the variation falls to the appraisers and equipment.
Gage R&R Example
I have dozens of examples, but here's a recent one. The supplier produces parts:
- Target: 43.11
- Tolerance: +/-0.13 (0.26 total)
They measured 10 parts with three appraisers. As you can see from the data table below, all parts are only off from the target by a few thousands. NO PART VARIATION.
If we run a frequency histogram on this data, you'll see that the capability indices (Cp, Cpk, Pp, Ppk) are excellent.
Even though the parts are good, they are too good for the measurement system. When we put this into the QI Macros Gage R&R template, the results are pretty bad.
%R&R = 97.8% (should be less than 10%). Look at part variation: PV = 0.0001.
NDC: Now look at the Number of Distinct Categories (0). If your NDC is less than 5, there isn't enough part variation to use the Average and Range Method or the ANOVA method to calculate Gage R&R:
The only method left is:
- Specification Tolerance Method
If you set the specification tolerance (AIAG MSA V3 pgs. 116-117), you can use the specification tolerance (0.26) instead of TV to calculate %R&R (1.1%):
You can even use the specification tolerance with the ANOVA method (%R&R = 1.3%):
But, it can still be hard to explain this to your customers.
A better choice is to go back and make sure you have 10 parts that span the tolerance: (42.98-43.24):
42.98, 43.01, 43.04, 43.07, 43.10, 43.13, 43.16, 43.19, 43.22, 43.24
When you do, there's much more part variation and the Gage R&R will change accordingly. Now, NDC is higher than 5 and %R&R is 0.6%:
Hint: You need BAD parts to do a GOOD Gage R&R.
Here's the Point
Gage R&R doesn't measure the quality of your parts, only the quality of your measurement system. If you want customers to trust your part quality, you have to learn how to properly evaluate measurement variation.
Gage R&R Video Series
- Gage R&R Template for Excel
- Gage R&R Type 1 Study
- Gage R&R Bias and Linearity
- Gage R&R Destructive Testing
- MSA Gage R&R Study Quick Reference Card (PDF, 204 KB)