Gauge R&R Calculator — Measurement System Analysis

Every measurement you take is the true dimension plus some error contributed by the measuring system itself. Gauge R&R separates those two, and the reason it matters is that measurement variation is counted twice: it inflates your apparent process spread, so a capable process can look incapable, and it causes wrong accept-reject decisions at the limits, so good parts get scrapped and bad ones shipped. If a capability study comes back worse than the machine behaviour suggests it should, the gauge is the first thing to eliminate — not the last.

The result splits into two components. Repeatability, or equipment variation, is the same person measuring the same feature on the same part and getting different answers — that is the instrument, the fixturing and the resolution. Reproducibility, or appraiser variation, is different people getting different answers on the same part, which usually points to technique: where the micrometer is placed, how much force is applied, how a datum is set. The distinction is practical rather than academic, because the fixes are completely different. Poor repeatability means a better or better-maintained instrument. Poor reproducibility means training, a written method, or a fixture that removes the judgement.

The threshold most customers quote is 10 and 30 per cent: below 10 the measurement system is acceptable, 10 to 30 is conditional depending on the criticality and the cost of doing better, and above 30 it needs work before you rely on it. But the number those percentages are measured against is where studies most often mislead. The default compares GRR to the total variation in the study, which is only meaningful if the parts you sampled genuinely span the real process spread. Take ten parts from a tight, well-controlled run and the part-to-part variation is small, so the gauge looks terrible as a percentage of it — even though it resolves the drawing tolerance without difficulty.

That is exactly the job-shop situation, and it is why this tool reports both. Comparing GRR to the tolerance instead — the precision-to-tolerance ratio — answers the question an inspection gauge actually has to answer: can this instrument tell a good part from a bad one against this print? For an acceptance gauge that is the relevant question. For deciding whether a process is fit for SPC, the total-variation comparison is the right one. Quoting one when your customer meant the other is a common and avoidable argument, so state which basis you used.

One detail worth knowing precisely, because it is widely half-understood: AIAG changed the study-variation multiplier between editions, from 5.15 in the third to 6.0 in the fourth — a factor of about 1.165. It scales the absolute study-variation figures, and it scales the percentage measured against the tolerance, because the tolerance is a fixed number from the drawing that does not scale alongside it. That alone can move a gauge from a pass to a fail with nothing physical having changed. What it does not touch is the percentage against total variation, or the percentage contribution: in both of those the multiplier appears top and bottom and cancels exactly. So if a historic study disagrees with a new one, check the multiplier first — but only the tolerance-based figure can be explained that way.