What the Gage R&R numbers mean
A Gage R&R study (repeatability and reproducibility) separates the variation caused by the measurement system from the real variation between parts. If gauge noise is large compared with the part-to-part spread or the tolerance, the measured data cannot be trusted for process control or capability studies.
EV: equipment variation (repeatability)
The variation seen when one operator measures the same part several times with the same gauge. It is estimated from the average range of the repeated readings.
AV: appraiser variation (reproducibility)
The variation between operators measuring the same parts. It is estimated from the differences between the operator averages, corrected for the share of repeatability they contain.
GRR, PV and TV
GRR combines both gauge sources. PV, the part variation, is the spread of the part averages. Total variation TV combines gauge and part variation, and %GRR is the share of total variation used up by the measurement system.
GRR = √(EV² + AV²)
TV = √(GRR² + PV²)
%GRR = 100 · GRR / TV
ndc: number of distinct categories
The number of distinct categories says how many groups of parts the gauge can reliably tell apart. It is rounded down to a whole number.
ndc = 1.41 · PV / GRR
Acceptance criteria (AIAG MSA)
The MSA manual of AIAG gives these guidelines for %GRR and ndc:
- %GRR below 10 %: the measurement system is acceptable.
- %GRR from 10 % to 30 %: conditionally acceptable, depending on the importance of the application, the cost of the gauge and the cost of repairs.
- %GRR above 30 %: not acceptable; the measurement system needs improvement.
- ndc of 5 or more: the gauge can distinguish enough categories within the process spread.
%GRR can be related to total variation, as in this calculator, or to the tolerance. For a gauge used to sort parts close to the limits, the tolerance-based figure is often the more demanding and more meaningful one.
How to run a study
A typical crossed study follows these steps:
- Choose 10 parts that cover the whole range of the process, including parts near the limits, not 10 consecutive parts from one batch.
- Select 3 operators who normally use the gauge, and use a calibrated gauge in its normal condition.
- Number the parts so that operators cannot see the identity or earlier results, and let each operator measure every part 3 times.
- Measure in random order and record each reading separately for part, operator and trial.
- Enter the readings in the grid, evaluate %GRR and ndc, and if the result is not acceptable look at resolution, fixture, method and training.
Fewer parts or trials give a wider uncertainty and are meant for a first look only. With a single operator the appraiser variation cannot be estimated.
Worked example
The example above uses 10 parts, 3 operators and 3 trials. Equipment variation is 6.70 % and appraiser variation 2.71 % of total variation, which gives %GRR = 7.23 % and ndc = 19. The part-to-part share is 99.74 %. %GRR is below 10 % and ndc is well above 5, so the gauge is acceptable. Repeatability makes up most of the gauge variation, so any improvement should start with the gauge and method, such as fixture or resolution, rather than with retraining the operators.
Frequently asked questions
What is an acceptable %GRR?
Below 10 % is acceptable, 10 % to 30 % is conditionally acceptable depending on the application, and above 30 % is not acceptable. Many customers also require ndc of at least 5. Check the requirement in your customer's quality agreement.
Why do %EV, %AV and %PV not add up to 100 %?
They are ratios of standard deviations to the total, so they combine as squares: %GRR² + %PV² = 100². The numbers shown in a column therefore do not sum to 100 %.
How many parts, operators and trials do I need?
The usual design is 10 parts, 3 operators and 3 trials. The parts must represent the whole process range. Fewer parts give a less reliable estimate of part variation.
What if there is only one operator?
With one operator the appraiser variation cannot be estimated, so the study shows repeatability only. The result applies to that person and should not be generalized.
Which method does this calculator use?
The average and range method of the AIAG MSA manual, 4th edition, with a crossed design. The ANOVA method additionally estimates the interaction between parts and operators and is preferred by many customers when software is available.
Save the result as a PDF report and work with studies
Keep your measurements, run complete studies and export the result as an audit-ready PDF report.
Start free trial