Measure

MSA Type 2

Measurement System Analysis Type 2 — Gage R&R (Repeatability & Reproducibility)

Overview

The MSA Type 2 study (Gage R&R, Repeatability & Reproducibility) decomposes measurement-system variation into three components: repeatability (same part, same operator), reproducibility (same part, different operators), and part-to-part variation. It is the standard check before process capability or control charts become trustworthy.

Repeatability (EV, Equipment Variation): Variation when one operator measures the same part multiple times. It comes purely from the instrument and the measurement procedure — not from humans or parts.

Reproducibility (AV, Appraiser Variation): Variation when different operators measure the same part. Comes from handling, reading, fixturing. A high AV share indicates training needs or an unclear procedure.

Part-to-part (PV, Part Variation): Variation between the selected parts themselves. Should dominate — if not, the system is mostly measuring itself, not the process.

GR&R in % (of variation or tolerance): Ratio of measurement-system variation to total variation or tolerance. < 10% is accepted, 10–30% conditionally accepted, > 30% not accepted.

Number of Distinct Categories (ndc): How many distinct levels the system can tell apart within the observed part range. Rule of thumb: ndc ≥ 5. Below 5 the process is seen too coarsely.

Gage R&R is a statistical experiment, not a number game. Part selection, order, operator blinding — all influence the result. A craftsmanlike setup matters more than the software.

Approach

Typical setup: 10 parts × 3 operators × 2–3 repeats = 60–90 measurements.
Pick parts spanning the process range — not all near the mean.
Use operators who run the process, not specially trained testers.
Randomize order, blind operators to which part is which.
Same procedure, same instrument, same fixture.
Analyze with the ANOVA method (not the range method for modern studies).
Assess GR&R percent, ndc, and component shares.
On failure, identify the cause (instrument, operator, fixture, procedure) and fix it.

Pitfalls

Parts too similar: If all 10 parts sit close together, the study fails from too little part variation — GR&R looks big because PV is small. Parts must span the relevant process range.

Operators know which part they have: Without blinding, they remember prior readings and "correct" unconsciously. Repeatability looks too good, real variation hides.

Order not randomized: Systematic order (e.g. operator 1 first, then operator 2) confounds drift with operator effect. Randomization is mandatory.

Only best operators tested: Pre-training the selected operators measures potential — not everyday reality. Results look great until real shifts run.

ndc ignored: GR&R 20% can come with ndc 3 — the system sees only three levels of the process. Read both indices together.

Range method instead of ANOVA: The classic range method underestimates interactions. Modern studies use ANOVA — it reports repeatability, reproducibility, and the operator×part interaction separately.

Interpreting "10% is good": Whether 10% GR&R is enough depends on the use. For safety-critical tolerances even 10% may be too much; for rough manufacturing 20% is fine. Context beats table.

Examples

This module ships with the following example datasets — load any of them in the app with a single click.

MSA Type 2: acceptable Gage R&R — Project template

Crossed ANOVA study: 10 parts × 3 operators × 3 replicates (90 measurements). Small repeatability and reproducibility variation, dominant part variance → %GRR ≈ 6 % of study variation. Textbook acceptable measurement system.

MSA Type 2: problematic Gage R&R — Project template

Crossed ANOVA study with large repeatability and reproducibility variation versus small part variance. 10 parts × 3 operators × 3 replicates → %GRR in the 40–50 % range. Textbook case where the measurement system masks part-to-part variation.

Available in the following cycles

DMAIC: Measure
DMADV: Measure
8D: D2 — Describe the Problem