Gage R&R (MSA): theory, step-by-step guide and case study

Theory: where measurement variation comes from

Every measured value is the true value plus a measurement error. If the error is large compared with the differences between parts, you cannot tell good parts from bad ones, and capability results and SPC charts are distorted. A Gage Repeatability and Reproducibility study (Gage R&R) estimates how much of the observed variation comes from the measurement system.

• Repeatability (EV, equipment variation): the variation when the same operator measures the same part several times. It reflects the gauge itself: resolution, clamping, fixture, measuring force.
• Reproducibility (AV, appraiser variation): the difference between operators measuring the same parts. It reflects the method: how the part is held, where it is measured, how the reading is taken.
• Part-to-part variation (PV): the real differences between the parts, which is what you want to see.

The ANOVA method

A two-way analysis of variance splits the total sum of squares into part, operator, operator × part interaction and repeatability. The interaction is kept only when it is significant (p < 0.25 by AIAG convention); otherwise it is pooled with repeatability. From the mean squares the variance components are estimated.

The key figures

The number of distinct categories (ndc) should be at least 5.

Step by step: running a Gage R&R

1. Choose 10 parts that represent the whole process range, ideally including parts near both specification limits. Number them invisibly to the operators.
2. Choose 3 operators who use this gauge in production.
3. Measure in random order. Each operator measures all parts, then repeats the round 2 or 3 times. Operators must not see their earlier results.
4. Enter the data. In the tool, set the number of parts, operators and trials and click Create table, then fill in the values, or paste the study from Excel.
5. Enter the tolerance (USL − LSL) to get % of tolerance.
6. Read the verdict on %GRR and ndc, and look at the variance components: is repeatability or reproducibility larger?
7. Check the charts. The range chart by operator shows inconsistent operators; the operator-by-part lines show interaction (lines crossing).

Case study: the AIAG reference study

The AIAG MSA manual uses a study with 10 parts, 3 operators (A, B, C) and 3 trials, 90 measurements in total. It is the data set behind Load AIAG example in the tool, so you can check the calculation against the manual.

Results (ANOVA method)

• %GRR = 27.9 % of the study variation: conditionally acceptable.
• ndc = 4: below the required 5.
• The operator × part interaction is not significant (p ≥ 0.25) and was pooled with repeatability.
• Reproducibility is larger than repeatability: the operators measure differently, the gauge itself is consistent.

What the team does

Because reproducibility dominates, a new gauge would not help. The team writes a clear measuring instruction (measuring point, number of readings, how to hold the part), trains the three operators with a reference part, and repeats the study with the same 10 parts. Operator C, whose average is visibly lower in the chart, gets an additional check of the method.

Common mistakes

• Parts that are too similar, for example 10 consecutive parts from the same setting. ndc drops and %GRR rises.
• Operators know which part they measure and remember the previous value.
• Judging only on % study variation when the gauge is used for go/no-go decisions against the tolerance.
• Skipping the Type 1 study. A gauge with bias or poor resolution should be fixed before the full study.
• Doing the study once and never again. Repeat it after gauge repairs, new operators or method changes.

For pass / fail inspection without a measured value, use the attribute agreement analysis instead.

Frequently asked questions