What is a Gage R&R study?
A Gage R&R study splits measurement variation into repeatability (the gauge) and reproducibility (the people), and compares them with the variation between parts and with the tolerance. It is a core part of measurement system analysis, and is required by IATF 16949 and the AIAG PPAP process. Run it before capability studies and SPC, because a poor gauge hides the real process.
Reading the results
- %Study Var (Gage R&R): can the gauge monitor the process? Under 10% good, 10–30% marginal, over 30% unacceptable.
- %Tolerance: can the gauge judge conformance to specification? Same thresholds.
- ndc: 5 or more distinct categories are required.
- Repeatability vs reproducibility: the larger one tells you where to act.
- Interaction plot: crossing operator lines indicate a part-by-operator interaction.
Gage R&R example
Three operators measured the bore diameter of 10 parts three times each with a digital caliper (90 measurements, tolerance 4 mm). The interaction is not significant (p = 0.27), so it is pooled into repeatability. Gage R&R is 19.2% of the study variation and 19.2% of the tolerance: conditionally acceptable. Reproducibility (15.4%) is larger than repeatability (11.5%), and the interaction plot shows why: operator Ben measures every part about 0.1 mm higher. Aligning the measuring technique is the fastest improvement. The system still distinguishes 7 categories of parts.
Improving a measurement system
- Repeatability dominates: maintain or replace the gauge, improve fixturing, check resolution, control temperature.
- Reproducibility dominates: write a clear measuring instruction, train operators, use fixtures that remove technique.
- Significant interaction: look for parts that are hard to measure (burrs, form errors) and specific operator habits.
How to use this Gage R&R tool
- Plan the study. Select about 10 parts across the process range and 2 or 3 operators who normally use the gauge.
- Measure blind and in random order. Each operator measures every part 2 or 3 times without seeing earlier results.
- Paste the data. Paste one row per part and operator, with a column for each trial.
- Enter the tolerance. Add the tolerance (USL − LSL) to see %Tolerance as well as %Study Var.
- Read the verdict. Check %Study Var, %Tolerance, ndc and whether repeatability or reproducibility dominates.
- Improve and export. Fix the dominant source, repeat the study, and export the report to PDF or Excel.
Frequently asked questions
What is Gage R&R?
Gage repeatability and reproducibility (Gage R&R) is a study that measures how much of the observed variation comes from the measurement system. Repeatability is the variation when the same operator measures the same part repeatedly (equipment variation). Reproducibility is the variation between operators (appraiser variation).
What are the acceptance criteria?
The AIAG MSA manual gives: Gage R&R under 10% of study variation (or tolerance) is acceptable; 10% to 30% may be acceptable depending on the application, the cost of the gauge and the cost of repair; over 30% is unacceptable. The number of distinct categories (ndc) should be 5 or more.
How many parts, operators and trials do I need?
The classic study uses 10 parts, 3 operators and 2 or 3 trials (60 to 90 measurements). Choose parts that cover the real process range; if the parts are too similar, %Study Var looks worse than it is.
What is the difference between %Study Var, %Tolerance and %Contribution?
%Contribution compares variances (they add up to 100%). %Study Var compares standard deviations with the total observed variation, so it tells you whether the gauge can monitor the process. %Tolerance compares 6 standard deviations of the measurement system with the tolerance, so it tells you whether the gauge can judge parts as good or bad.
Why use the ANOVA method instead of average and range?
The ANOVA method uses all the information in the data, gives better estimates, and can detect a part-by-operator interaction (some operators measure some parts differently). The average and range method cannot. This tool tests the interaction and pools it into repeatability when it is not significant (p above 0.05), as most statistical software does.
What is the number of distinct categories?
ndc = 1.41 × (part-to-part standard deviation / Gage R&R standard deviation), rounded down. It estimates how many groups of parts the system can reliably tell apart. Below 5, the gauge is not adequate for process control.