MSA & Gage R&R: can you trust your measurement system?

Before you trust the data, you have to trust the way it was measured. Measurement Systems Analysis asks a blunt question: how much of the variation you see is the part — and how much is your measurement system?

What MSA is for

Every measurement contains two kinds of variation: real differences between parts, and noise from the measurement system itself — the gauge, the method, the environment and the people. Measurement Systems Analysis (MSA) separates the two so you know whether your data reflects the product or just the way you measured it.

Repeatability and reproducibility

Gage R&R, the workhorse of MSA, splits measurement-system variation into two parts:

• Repeatability — the same person measuring the same part with the same gauge gets slightly different results. This is the equipment’s own scatter.
• Reproducibility — different people (or setups) measuring the same part get different results. This is the appraiser/method variation.

Together they are the R&R: the total noise your measurement process adds.

Reading the numbers: %GRR and ndc

Two outputs decide whether the system is fit for use:

• %GRR — the measurement-system variation as a percentage of the total (or of the tolerance). A common rule of thumb: under 10% is good, 10–30% is marginal (acceptable depending on application and cost), over 30% is unacceptable.
• ndc (number of distinct categories) — how many distinct groups the system can reliably tell apart. Five or more is generally wanted.

A capable process measured by an incapable system produces data you cannot trust — and decisions you cannot defend.

Average-range vs ANOVA

The older average-and-range method is quick by hand. The ANOVA method is preferred because it also captures the part-by-appraiser interaction and gives a more complete, statistically sound breakdown. If your software offers ANOVA, use it.

Beyond R&R: bias, linearity, stability

R&R covers precision, not accuracy. A full MSA also checks bias (a consistent offset from the true value), linearity (does bias change across the range?) and stability (does the system drift over time?). Each can pass R&R yet still mislead if ignored.

Where it fits

MSA is most associated with production and IATF 16949, but the logic applies to any lab: prove the measurement system before you act on the measurements. Built into the same tool as your calibrations and SPC, it becomes a routine check rather than a special project.

More from the Cali blog: browse all calibration & metrology guides →