A practical guide to Measurement System Analysis (MSA) — Gage R&R (repeatability and reproducibility), percent study variation, number of distinct categories, and bias, linearity, and stability.
What Is Measurement System Analysis?
Every measurement is the sum of the true value and measurement error. If the measurement system adds too much variation, the data it produces can mislead every decision built on it — process capability, control charts, acceptance, and improvement. Measurement System Analysis (MSA) is the discipline of quantifying that measurement error and confirming the system is good enough for its purpose before trusting the numbers. It is a required step in the Measure phase of Six Sigma.
The total observed variation decomposes as:
σ²total = σ²part-to-part + σ²measurement system
MSA aims to make the measurement-system term small relative to the part-to-part term so that observed variation truly reflects the process.
The Properties MSA Evaluates
| Property | Question it answers |
|---|
| Bias | Does the gage read consistently high or low versus a reference (accuracy)? |
| Linearity | Does the bias change across the measurement range? |
| Stability | Does the gage drift over time? |
| Repeatability | How much variation when one operator re-measures the same part? |
| Reproducibility | How much variation between different operators? |
The first three concern location/accuracy; the last two concern spread/precision and are the focus of a Gage R&R study.
Bias, Linearity, and Stability
- Bias is the difference between the average of repeated measurements and a known reference (master) value. It is corrected by calibration.
- Linearity describes how bias varies across the operating range — a gage may be accurate at low readings but biased at high readings. Linearity is assessed by measuring several reference parts spanning the range.
- Stability is the change in bias over time, monitored by periodically measuring a master part and plotting the results on a control chart. Drift signals the need for recalibration or maintenance.
Gage R&R: Repeatability and Reproducibility
A Gage R&R study estimates the precision of the measurement system by separating two sources:
- Repeatability (Equipment Variation, EV) — variation observed when the same operator measures the same part multiple times with the same gage. It reflects the instrument itself.
- Reproducibility (Appraiser Variation, AV) — variation observed when different operators measure the same parts. It reflects differences in technique, training, or procedure.
The combined gage variation is:
σR&R = √(σ²repeatability + σ²reproducibility)
A typical crossed study uses around 10 parts, 3 operators, and 2–3 trials each, with parts measured in random order and operators blind to prior readings. Results are analyzed by ANOVA (preferred) or the average-and-range method.
Percent Study Variation and Acceptance
The headline result is usually expressed as a percentage. %Study Variation compares gage variation to total study variation:
%SV = (σR&R / σtotal) × 100
When the goal is acceptance against specifications, %Tolerance is used instead, comparing gage variation to the tolerance width. Common AIAG acceptance criteria:
| %Study Variation (or %Tolerance) | Verdict |
|---|
| Under 10% | Acceptable measurement system |
| 10% – 30% | May be acceptable depending on application, cost, and importance |
| Over 30% | Unacceptable — improve the measurement system |
Number of Distinct Categories (ndc)
The number of distinct categories estimates how many distinct groups of parts the gage can reliably tell apart across the observed range:
ndc = 1.41 × (σpart / σR&R), truncated to an integer
The widely used guideline is ndc ≥ 5. An ndc of 1 means the gage can only tell "is it different or not"; an ndc of 5 or more means it can resolve the process variation into enough buckets to support SPC and capability work. Note that ndc rewards a large part-to-part spread relative to gage error — the same gage can have a poor ndc on a tightly grouped set of parts and a good ndc on a more varied set.
Why MSA Comes First
Control charts and capability studies assume the data is trustworthy. If the measurement system contributes excessive variation, a control chart may show false alarms (or hide real signals) and a capability index will be understated because gage error inflates the observed spread. That is why MSA precedes building a control chart or running a capability study — you must validate the measuring instrument before trusting what it tells you about the process.
Improving a Poor Measurement System
- High repeatability (EV) — investigate the gage: wear, resolution, fixturing, and clamping. Better fixtures and higher-resolution instruments help.
- High reproducibility (AV) — investigate operators: standardize the procedure, train appraisers, and use clearer work instructions or automated measurement.
- Bias or linearity problems — recalibrate and verify against masters across the full range.
- Stability problems — increase calibration frequency and monitor with a gage control chart.