Is there free software that does Gage R&R with the ANOVA method?

Yes. SVEND's Gage R&R calculator runs a crossed study by ANOVA variance decomposition and reports %GRR, number of distinct categories, the ANOVA table and the variance components. It runs in a browser, requires no account and nothing is uploaded. Gage linearity and bias is free on the same bench. The range method is quicker to hand-calculate but ANOVA is the method the AIAG MSA manual prefers, because it separates the operator-by-part interaction that the range method folds into reproducibility.

Can I run a measurement system analysis without installing software?

Yes. SVEND is browser-based with nothing to install, which matters in a calibration lab or on a shop floor where getting software approved and installed is slower than doing the study. The free calculators need no account at all; the full bench, including SPC, capability and reliability alongside MSA, is $19 per month billed monthly.

What counts as an acceptable %GRR?

The AIAG MSA manual's conventional bands are under 10% acceptable, 10 to 30% conditionally acceptable depending on the application and the cost of the measurement, and above 30% unacceptable. Number of distinct categories is the companion figure and is often the more honest one: ndc of 5 or more is the usual minimum for a measurement system used to control a process, because it describes how many groups the gauge can actually tell apart. A study can pass on %GRR and still be too coarse to control with.

How do I know the calculations are right?

The reference values are not ours. SPC and capability are cross-checked against R's qcc package, 28 of 28 checks, and hypothesis tests and regression against base R, 40 of 40; statistics are additionally calibrated against 14 NIST Statistical Reference Datasets certified to 15 significant digits. The calibration certificate is published, including any package that fails: every package currently passes.

I — Datum

Measurement System Analysis

Every capability number you have ever quoted was measured with something. If the gauge cannot tell two parts apart, the Cpk is a statement about the gauge and not about the process — which is why MSA comes before capability rather than after it.

Ref: AIAG MSA Manual, 4th Ed. · VDA 5 · Montgomery · Wheeler & Lyday

Free Gage R&R Calculator → Digital gage integration → Start Free → Pricing →

II — What Runs Today

This table is generated from the analysis catalog when the page is requested, so it cannot drift from the product. Standby rows are listed rather than hidden — finding out after you have paid is worse than reading it here.

Method	What it answers	Live
Gage R&R (crossed)	ANOVA variance decomposition — repeatability, reproducibility, operator-by-part interaction, %GRR, number of distinct categories. Run it in the free calculator, or issue a randomized run sheet and record the study in the authenticated Gage R&R bench.	●
Test Uncertainty Ratio	The Z540.3 ratio of tolerance span to calibration uncertainty	●
Intraclass Correlation (ICC)	Reliability across raters — one column per rater	●
Bland-Altman Agreement	Bias and limits of agreement between two measurement methods	●
Attribute Agreement (Fleiss' κ)	Inter-rater agreement on categorical ratings — one column per rater	●
Krippendorff's Alpha	Inter-rater reliability for any measurement level — one column per rater	●

5 live, 0 standby. The Gage R&R calculator is public and needs no account, and nothing you paste into it is uploaded.

III — Where This Is Unusual

Most quality packages give you crossed Gage R&R and stop. Two things here are not the norm.

Linearity and bias is free. A gauge that reads true at the middle of its range and drifts at the ends passes a single-point bias check and fails in production. Almost nobody gives that study away.

Agreement statistics sit next to shop-floor MSA. ICC, Bland-Altman and Krippendorff's α come from laboratory and clinical method comparison, and they answer questions AIAG's crossed study does not: do these two instruments agree, does the disagreement grow with the reading, how do I handle raters who did not measure every part. If you are qualifying a new gauge against an incumbent, that is the analysis you actually want.

And the arithmetic is checked rather than asserted: SPC and capability match R's qcc package 28 of 28, hypothesis tests and regression match base R 40 of 40, and the calibration certificate is published with its own failures on it.

IV — Where MSA Studies Actually Go Wrong

Not in the arithmetic. Any package computes %GRR correctly. The study fails earlier, in three ways that no calculator sees.

The plan. Ten parts spanning the process range, three operators, three trials, randomized so nobody measures part 1 first every time — and parts that actually represent production variation rather than ten good ones off the top of the bin. A study built on unrepresentative parts reports a small %GRR because the part variation it is dividing by is small.

The transcription. Readings go from the gauge to a clipboard, from the clipboard to a spreadsheet, and from the spreadsheet into the analysis. Three copies by hand, ninety numbers, and a transposed digit changes a verdict.

The interval. A gauge qualified in March is not qualified in September. Stability is a control chart on repeat measurements of a master part over time, and it is the study most often skipped entirely.

V — Boundaries

What this does not do yet. We would rather you find it here.

Crossed and nested are both built. Crossed assumes every operator can measure the same part. Destructive testing cannot, so a nested study deals each appraiser its own parts and the bench allocates them that way.

The study bench issues the sheet and judges it. It preserves a randomized run sheet, accepts keyboard-wedge readings, records corrections as later events, and judges the completed study. identified serial-device capture remain separate work. The full study reference →

Gage stability is calibration work, and sits with the register. It has its own bench — one master of known value, measured again and again for as long as the instrument is in service — but it moved to the Team tier on 2026-08-10 rather than staying beside the study bench, on the view that checking a master against its known value over months is calibration rather than an analysis run on a dataset you already hold.

VI — Where It Sits

Dedicated metrology software is priced for a calibration department and licensed per seat. General statistics packages do crossed Gage R&R and little else in MSA. The person stuck in between — a quality engineer who needs a defensible gauge study this week, on a laptop, without raising a purchase order — is who this is for.

The Gage R&R calculator is free and needs no account. The full bench, including SPC, capability, reliability and value stream mapping alongside MSA, is $19 per month billed monthly, with no implementation engagement attached.

Run a Gage R&R free → Start Free →