What Is Gage R&R? Definition and Uses

Gage R&R — gage repeatability and reproducibility — is a designed study that measures how much of the variation in your data comes from the measurement system rather than from the parts being measured. Repeatability asks whether the same person, measuring the same item with the same instrument, gets the same value twice. Reproducibility asks whether different people agree with one another. If the answer to either is no, the process may be innocent — and your data guilty.

How it works

The classic crossed study takes around ten parts spanning the real range of process variation, two or three appraisers, and two or three measurement trials each, run in randomized and blinded order. Analysis — usually ANOVA — splits total observed variation into part-to-part variation and measurement-system variation, then splits the latter into repeatability and reproducibility. The headline result is percent GRR: the share of observed variation consumed by measurement itself. A parallel study exists for pass-or-fail judgments — attribute agreement analysis — but the continuous-data version is what most people mean by the term.

Widely used guidelines call a system under 10 percent acceptable, 10 to 30 percent marginal depending on the application, and over 30 percent unacceptable. A failing result is not a dead end; it points at its own cause. High repeatability error indicts the instrument or the method. High reproducibility error indicts training and technique — the appraisers are measuring differently.

An illustration: the machine shop

A machine shop checks turned shafts with hand calipers, and the control chart has been jumpy for weeks. A Gage R&R study — ten shafts, three inspectors, three trials — shows the measurement system consuming close to a third of observed variation, most of it reproducibility: each inspector seats the caliper differently on the shaft. The shop writes a one-page measurement method, adds a simple locating fixture, and re-runs the study. Percent GRR drops into the acceptable band, and half the “process instability” disappears without anyone touching the process. The lesson generalizes: when a control chart misbehaves, the first suspect worth interviewing is the measurement system.

• Choose study parts that span the process’s full range — ten identical parts prove nothing
• Randomize and blind the trials; inspectors who recognize a part remember its number
• Calibration is not R&R — an accurate gage can still be inconsistent
• Fix the method and the training before shopping for a new instrument
• Re-run the study after any change to instrument, method, or people

Before you measure the process, qualify the measurement.

Gage R&R lives in the Measure phase of DMAIC and is taught hands-on in our Green Belt program — $299, roughly 35 hours, with a 100-question proctored exam and lifetime access to the material. Black Belt training goes deeper into the ANOVA behind the percentages.