MSA Study Explained: Gage R&R, Attribute Agreement, and What Passes

Every capability index, control chart, and dimensional result in your quality system rests on an assumption so basic it usually goes unstated: the measurement is telling the truth. A Cpk of 1.4 computed from a gage that cannot repeat its own reading is not a capability result. It is noise with a decimal point.

Measurement System Analysis exists to test that assumption. An MSA study quantifies how much of the variation in your measurement data comes from the measurement system itself, the gage, the operator, the method, rather than from the parts being measured. That share is measurement error, and it rides along in every reading the measurement process produces. If the measurement system consumes too much of the tolerance or masks the differences between parts, every downstream decision made on that data inherits the problem: false accepts, false rejects, and process improvements chasing variation that was never in the manufacturing process.

The requirement is not optional in regulated supply chains. MSA is Element 8 of the Automotive Industry Action Group (AIAG) PPAP package, an expectation within AS9145 for aerospace APQP, a requirement under IATF 16949 for automotive quality systems, and the subject of AS13003, the aerospace standard for measurement systems analysis in aero engine supply chains. When a customer asks for your MSA studies, this is the body of practice they mean. This guide covers the two study types that carry most of those requirements, the AIAG gates that decide them, and the audit failures that break otherwise valid studies.

The Two Studies That Cover Most Requirements

The AIAG MSA reference manual defines several study types, including bias, linearity, and stability studies that examine a gage's accuracy over its range and over time. MSA is not a substitute for calibration: calibration establishes a gage's accuracy and measurement uncertainty against a reference standard, while an MSA study tests how the whole measurement system, gage, operators, and method together, performs on your parts. In day-to-day supplier quality work, two study types carry most of the load: variable Gage R&R for measured values, and attribute agreement analysis for pass/fail judgments.

Variable Gage R&R (crossed study). Gage repeatability and reproducibility, often written gauge R&R, is the classic design: multiple operators each measure the same set of parts multiple times with the same gage, every operator on every part, with measurement order randomized. Ten parts, three operators, three trials is the common textbook design, 90 measurements. The analysis decomposes total variation into:

• Repeatability (EV, equipment variation): the gage's ability to agree with itself, the same operator, part, and gage, measured again.
• Reproducibility (AV, appraiser variation): the variation between operators measuring the same parts with the same gage.
• Gage R&R (GRR): the combined measurement system variation, reported as a percentage of study variation and against the tolerance (P/T, percent of tolerance).
• ndc (number of distinct categories): how many distinct groups of parts the measurement system can reliably distinguish across the study range.

Part selection matters as much as the math. %GRR compares measurement system variation to total variation, which also includes part-to-part, or process, variation. Parts chosen to span the expected process variation give an honest result; a set of near-identical parts shrinks that denominator and makes a sound measurement system look inadequate.

Two calculation methods exist: the Average and Range method, the manual worksheet approach, and ANOVA, which additionally separates the operator-by-part interaction. ANOVA is the more complete analysis; running both and comparing is common practice, and modern tools compute both automatically.

Attribute agreement analysis. When the inspection is a judgment call, go/no-go gaging, visual inspection, thread checks, the study design changes: appraisers make repeated accept/reject calls on parts with known reference answers. The analysis reports effectiveness (percent of correct decisions), miss rate (bad parts accepted, the dangerous error), false alarm rate (good parts rejected, the expensive one), and kappa statistics measuring agreement beyond chance, within appraisers, between appraisers, and against the standard.

What Passes: The AIAG Gates

For variable Gage R&R, the AIAG acceptance criteria are the reference point nearly every customer flow-down uses:

%GRR	Verdict
Under 10%	Acceptable without conditions
10% to 30%	Marginal; requires evaluation and documentation, with approval based on risk, cost, and process capability
Over 30%	Unacceptable; the measurement system must be improved before use in production or in a PPAP submission

ndc of 5 or more is the accompanying pass threshold, and it stands on its own: an ndc below 5 marks the measurement system inadequate whatever the %GRR, because it cannot distinguish enough part-to-part difference to be useful for process analysis. P/T is read against the same 10% and 30% gates.

For attribute studies, the AIAG gates apply to each appraiser: effectiveness of 90% or better is acceptable (80% to 90% marginal), miss rate at or under 2%, and false alarm rate at or under 5%. For kappa, AIAG treats 0.75 or higher as good agreement, and 0.90 or higher is preferred. An attribute study is judged per appraiser, and the system's verdict follows its weakest appraiser, which is exactly the point: one inspector who accepts marginal parts is a system problem, not a personnel footnote.

A marginal result is not an automatic failure, but it is never a shrug. It is a documented engineering decision, and auditors read it as one.

Why MSA Studies Fail Audits (It Is Rarely the Math)

The statistics are the settled part of MSA. Where studies actually break down in audits and customer reviews is provenance:

• The study is disconnected from the gage. A spreadsheet says "caliper"; the calibration system knows twelve calipers. Which one, and was it in calibration on the study date?
• The study is disconnected from the feature. A GRR run on a convenient dimension says little about the critical characteristic the control plan actually gates.
• The study is stale. The gage was replaced, the measurement method changed, or the tolerance tightened, and the study on file describes a measurement system that no longer exists.
• The blind is broken. Operators who can see each other's readings converge, and the study flatters the system.
• The math is unauditable. A single %GRR cell in a spreadsheet, with the formulas long since pasted-over, invites exactly the question it cannot answer: show me how this number was calculated.

Excel and standalone statistics tools produce correct numbers and weak evidence. The number is one output; the traceable chain from feature to gage to calibration status to raw trials to result is what a reviewer accepts.

Running MSA Studies Inside the Quality Platform

That provenance problem is a data-model problem, which is why it is better solved where the gages, features, and measurements already live. Net-Inspect now includes MSA Studies: variable and attribute Gage R&R studies planned, collected, calculated, and approved against the part features and calibrated tools already on file.

A study starts from a real part number, feature, and tool ID, with the tool's calibration status and next due date visible at study creation. Collection runs as a blind study, with operators unable to see other readings during data entry. Results report both ANOVA and Average and Range methods side by side, with every figure traceable: selecting a result opens the formula with the study's actual values substituted, so "show me how this number was calculated" has a literal answer. AIAG gates are applied automatically, approval locks the study read-only, and a full audit trail records who entered, calculated, and approved what. Because a study links to an APQP/PPAP checklist item, PPAP Element 8 is satisfied by a live, reviewable study rather than an uploaded spreadsheet, and a customer can open it read-only through the linked package.

The measurement data those studies qualify then feeds the same platform's SPC and measurement collection workflows, which is the quiet payoff: capability results built on measurement systems you have evidence for.

The second payoff is the gage record. MSA Studies works from the same tool records your Tool Calibration program already maintains, so the study and the calibration history point at the same tool ID. When an auditor asks which caliper was used and whether it was in calibration on the study date, the answer comes from one system instead of two spreadsheets.

The Standard to Hold Your Measurement System Analysis To

An MSA study makes a specific, checkable claim: this measurement system, on this feature, with this gage, can be trusted to this degree. Run the study designs the AIAG manual defines, hold the results to the published gates, and keep the chain from feature to gage to raw trials intact, because the chain is what gets audited.

To see how MSA studies, tool calibration, and PPAP evidence work in one system, explore the APQP/PPAP module or see how Tool Calibration keeps every gage in a study traceable to its calibration record.

Explore APQP and PPAP in Net-Inspect