← Insights

Measurement

Gauge R&R explained, minus the mystery

July 2026 · 8 min read

Here is an uncomfortable question for any factory: when your inspector measures a part at 10.02 mm, how much of that number is the part - and how much is the gauge, the method and the person holding it?

Every measurement system has variation of its own. If that variation is small relative to what you are trying to measure, your data means something. If it is large, you are sorting parts with a pair of dice: good parts get rejected, bad parts get shipped, and your control charts chase ghosts that live in the gauge rather than the process. A Gauge R&R study is how you find out which world you are in, and it is the foundation of Measurement Systems Analysis (MSA).

The two Rs

Repeatability - equipment variation (EV). The same operator measures the same part with the same gauge several times and does not get the same answer. This is the gauge's own noise: play in the mechanism, resolution limits, fixturing.

Reproducibility - appraiser variation (AV). Different operators measuring the same parts get systematically different answers. This is the human layer: technique, feel, how the part is seated, how the scale is read.

Combined, they give GRR - the total noise of the measurement system. The study's job is to compare that noise with the variation you actually care about, and it also yields PV (part variation - the real differences between parts) and TV (total variation).

The standard study

The classic AIAG average-and-range study: 10 parts, 3 operators, 3 trials - 90 measurements. The discipline matters more than the arithmetic:

• Choose parts spanning the real process range. This is the most common study mistake: ten nearly identical parts make PV artificially small, which makes %GRR artificially huge, and a decent gauge fails the study. Pick parts across the tolerance - including near both limits.
• Randomise and blind. Operators must not know which part they are measuring or what they scored last time - number the parts where the operator cannot see.
• Normal conditions. The gauge as used at the line, not a laboratory ritual. You are qualifying reality.

From the averages and ranges, published constants (K1, K2, K3, and D4 for the stability check) turn the raw spreads into EV, AV, GRR, PV and TV. The constants for a 10 x 3 x 3 study are fixed and published - K1 = 0.5908, K2 = 0.5231, K3 = 0.3146, D4 = 2.575. This, incidentally, is where cheap MSA templates quietly fail: wrong or hidden constants that no auditor can check.

Reading the verdict

Two numbers decide the outcome:

%GRR - measurement system variation as a percentage of total variation. The AIAG thresholds: under 10% acceptable; 10-30% conditionally acceptable, depending on the importance of the application and cost of improvement; over 30% unacceptable - fix the measurement system before trusting any data that comes through it.

ndc - the number of distinct categories the system can genuinely tell apart across the part range, which should be 5 or more. An ndc of 2 means your elegant variable gauge is functioning as a go/no-go gauge with extra steps.

The verdict tells you what to fix, not just whether you passed. If %EV dominates, the problem is the equipment: maintain, fixture or replace the gauge. If %AV dominates, the problem is the method: clarify the technique, train, add a fixture that takes feel out of the measurement.

One check before believing any of it: the range chart. If any operator's ranges exceed UCLr, that operator's measuring process was not stable during the study - resolve that first, because the rest of the numbers are built on it.

Where it fits

Every measured characteristic on your control plan implies a gauge, and every gauge doing that work deserves a study behind it - that is the MSA element of PPAP, and it is the first thing a capability claim rests on. Measure first, then control, then claim capability: gauge, chart, Cp/Cpk, in that order.