Academy › Practical guides › Quality tools
Gage R&R: what the 10% / 30% thresholds really mean (and what they don't)
%GRR, ndc, %Tolerance, ANOVA vs X̄-R — and why a study that passed can still be useless.
Updated 13 November 2026 · 7 min read · Written for working labs and quality teams
Short answer
%GRR is the share of observed variation attributable to the measurement system. The AIAG MSA reference manual's conventional bands are under 10% acceptable, 10–30% potentially acceptable depending on the application, cost and customer, and over 30% needing improvement. These are industry conventions, not ISO clauses. Read them together with ndc (number of distinct categories, commonly ≥ 5), %Tolerance, and whether the parts in the study actually spanned the process variation.
On this page
The numbers you get from a study
The classic design is 10 parts × 3 operators × 3 trials, randomised, with parts selected to span the normal process range. ANOVA is preferred over the older X̄-R (range) method because it estimates the operator-by-part interaction instead of folding it into the residual.
| Metric | What it measures | Common convention |
|---|---|---|
| %GRR (vs total variation) | Measurement system share of observed spread | under 10% acceptable · 10–30% conditional · over 30% improve |
| %Tolerance | Measurement spread as a share of the spec width | Often judged against 10% / 30% too — but it is a different denominator |
| ndc / distinct categories | How many groups the system can reliably tell apart | ≥ 5 commonly expected; < 2 means the system cannot discriminate |
| Repeatability (EV) | Same operator, same part, same gage | Equipment / within-system variation |
| Reproducibility (AV) | Different operators measuring the same parts | Often operator-method or fixture issues, and operator × part interaction |
A result is not a verdict on its own
- A study can pass with the wrong parts. If the ten parts are all from one tight batch, process variation is understated and %GRR looks worse; if they are hand-picked across a far wider range than production, %GRR looks better than reality. Select parts to represent the process, not to make the number move.
- %GRR and %Tolerance answer different questions. A wide tolerance can give a comfortable %Tolerance while %GRR against process variation is poor — that matters when the chart is used to detect small shifts.
- Repeatability vs reproducibility tells you where to work. High EV points at the gage, fixturing or resolution; high AV points at operator technique, training or a poorly defined measurement method; a significant operator × part interaction usually means the method is being applied differently depending on the part.
- GR&R is one study, not the whole MSA. Bias, linearity, stability and (for attribute gages) agreement studies are separate. A destructive test, a chemical assay or a vision system needs a different design altogether.
Practical note. Resolution still matters: a gage that reads in 0.1 mm increments cannot convincingly support a study on a 0.05 mm spread. The old "10:1 rule of thumb" has been replaced by more nuanced guidance in current MSA references — check the resolution and the number of distinct categories actually obtained rather than quoting a ratio from memory.
What to do when the result is marginal
- Check the study design first: part selection, randomisation, blind measurement, and whether operators followed the same defined method.
- Look at the ANOVA table to see whether the problem is equipment or operator — the fix is completely different.
- If 10–30% is genuinely conditional, document the justification: characteristic criticality, customer agreement, cost of improvement, and what additional controls (e.g. averaging repeat readings) are used.
- After any change to the gage, fixture or method, re-run the study. A MSA result belongs to a specific system, not to the part number forever.
Where this goes wrong in practice
Treating 10% and 30% as pass/fail requirements in a standard. They are widely used AIAG MSA conventions; customer-specific requirements may set different thresholds.
Selecting ten near-identical parts, which inflates %GRR and triggers an unnecessary gage replacement project.
Reporting %Tolerance when the customer asked for %GRR against total variation (or the reverse) without saying which denominator was used.
Letting operators watch each other's readings, or measuring in part order instead of randomised order — both quietly destroy the study's independence.
Running one study and never repeating it after a fixture change, a new operator group or a gage repair.
Frequently asked
Are the 10% / 30% thresholds in ISO 9001 or IATF 16949?
No. They come from the AIAG MSA reference manual (and equivalent customer references) and are applied through customer-specific requirements, PPAP submissions and control plans. The standards require suitable measurement resources and valid results; the numeric bands are conventional.
What is ndc and why does it matter?
Number of distinct categories (also called number of distinct data classes) estimates how many separate groups the measurement system can reliably distinguish within the observed variation. A value below 2 means the system cannot tell products apart; ≥ 5 is the commonly expected minimum for a usable process-control gage.
Can I average multiple readings instead of improving the gage?
Sometimes — averaging repeated measurements reduces the repeatability component, but it must be defined in the measurement method, applied consistently, and reflected in the control plan. It does not fix a reproducibility problem or poor resolution.
How often should a Gage R&R study be repeated?
There is no universal interval. Re-run it when the gage, fixture, method, operator population or characteristic changes, when a customer requires it, and at the frequency your control plan or customer-specific requirements define.
Read the ANOVA, not just the headline
Run a Gage R&R study and see EV, AV, %GRR and ndc
Enter the standard 10 × 3 × 3 readings and get the ANOVA breakdown, repeatability vs reproducibility, %GRR, %Tolerance and the distinct-category count in your browser.
Open the free MSA / Gage R&R tool →Runs in your browser. Export the report free. No sign-up needed.
LabKitPro guides are written from published standards and accreditation-body documents — references, clause numbers and paraphrased intent only, never verbatim standard text. They are for preparation and study, not a certification or accreditation decision, and not legal advice. ISO standards are trademarks of ISO; IATF 16949 is published by the International Automotive Task Force and is not an ISO standard. The current laboratory standard is ISO/IEC 17025:2017; there is no 2025 edition. Global ACI replaced IAF and ILAC on 1 January 2026.
Academy by standard · All practical guides · Free browser tools