Gage R&R (Repeatability and Reproducibility) is a statistical method to assess the quality of a measurement system. It quantifies how much variation in measurements comes from the measurement system itself versus actual process variation. Gage R&R is part of Measurement System Analysis (MSA) and validates both accuracy (measuring the true value) and precision (getting consistent results) before collecting data for Six Sigma projects.

What are the Gage R&R acceptance criteria?

For continuous Gage R&R: (1) %R&R (Study Variation) 30% not acceptable. (2) %Tolerance 30% not acceptable. (3) Number of Distinct Categories (NDC) ≥ 5 is good, 3-4 acceptable, 30% or NDC <3), you must fix the measurement system before continuing your project.

What is the difference between repeatability and reproducibility?

Repeatability (Equipment Variation) measures variation when the same operator measures the same part multiple times. It reflects measurement equipment variation and test method consistency. Reproducibility (Appraiser Variation) measures variation when different operators measure the same part. It reflects operator-to-operator differences in measurement technique, interpretation, or skill. Together, R&R = Repeatability + Reproducibility quantifies total measurement system variation.

How do you calculate Gage R&R?

For continuous Gage R&R using AIAG method: (1) Collect data from 10 parts, 3 operators, 3 repetitions (90 measurements). (2) Perform ANOVA to estimate variance components: σ²observed, σ²R&R, σ²part-to-part. (3) Calculate three metrics: %R&R = (6×σR&R)/(6×σobserved)×100%, %Tolerance = (6×σR&R)/(USL-LSL)×100%, NDC = √2×(σpart-to-part/σR&R). (4) Compare results to acceptance criteria. If any metric fails (>30% or NDC <3), the measurement system is not acceptable.

What is %R&R in Gage R&R?

%R&R (also called % Study Variation or %GRR) is the proportion of observed variation caused by the measurement system. Formula: %R&R = (6×σmeasurement system)/(6×σobserved)×100%. It indicates the relative usefulness of the gage for control charting and process capability analysis. Low %R&R means most variation you observe is real process variation, not measurement error. %R&R 30% not acceptable.

What is NDC in Gage R&R?

NDC (Number of Distinct Categories) represents the number of non-overlapping confidence intervals that span product variation range, or how many distinct groups your measurement system can discern. Formula: NDC = √2×(σpart-to-part/σR&R). NDC ≥ 5 is excellent (can detect meaningful process changes), NDC = 3-4 is marginal (limited discrimination), NDC < 3 is poor (essentially a go/no-go gage). Higher NDC indicates better measurement system discrimination power.

What is attribute Gage R&R?

Attribute Gage R&R (Attribute Agreement Analysis) evaluates measurement systems for pass/fail, good/bad, or OK/KO decisions. Protocol: 3 operators evaluate 50 parts (balanced 50% good/50% bad) 3 times each. Metrics include: % repeatability (operator consistency with self), % reproducibility (operator-to-operator agreement), % accuracy vs reference, % concordant appraisals, false reject rate, and false accept rate. ALL agreement percentages must be ≥90% for an acceptable system. Below 80% requires operator retraining or process redesign.

How do you improve a failing Gage R&R?

For high repeatability (equipment variation): calibrate/maintain equipment, increase measurement resolution (1/10th rule), improve fixturing, control environment, simplify test method. For high reproducibility (operator variation): train operators, clarify procedures with detailed work instructions, reduce subjectivity, use digital vs analog displays, implement error-proofing fixtures. For attribute systems: create crystal-clear operational definitions with visual examples, provide limit samples showing borderline cases, optimize inspection environment.

What is the AIAG Gage R&R method?

The AIAG (Automotive Industry Action Group) Gage R&R method is the industry standard for continuous measurement systems. Protocol: select 10 parts representing full process variation, have 3 operators each measure every part 3 times in random order (total 90 measurements). Analyze using ANOVA to calculate variance components. Evaluate using three metrics: %R&R (study variation), %Tolerance (acceptance risk), and NDC (discrimination power). Sample size requirements: ≥90 measurements is good, 40-89 acceptable, <40 not acceptable.

Gage R&R: Complete Guide to Measurement System Analysis (MSA), Repeatability & Reproducibility

Q: What is the difference between repeatability and reproducibility?

Repeatability (Equipment Variation) measures variation when the same operator measures the same part multiple times. It reflects measurement equipment variation and test method consistency. Reproducibility (Appraiser Variation) measures variation when different operators measure the same part. It reflects operator-to-operator differences in measurement technique, interpretation, or skill. Together, R&R = Repeatability + Reproducibility quantifies total measurement system variation.

Q: How do you calculate Gage R&R?

For continuous Gage R&R using AIAG method: (1) Collect data from 10 parts, 3 operators, 3 repetitions (90 measurements). (2) Perform ANOVA to estimate variance components: σ²observed, σ²R&R, σ²part-to-part. (3) Calculate three metrics: %R&R = (6×σR&R)/(6×σobserved)×100%, %Tolerance = (6×σR&R)/(USL-LSL)×100%, NDC = √2×(σpart-to-part/σR&R). (4) Compare results to acceptance criteria. If any metric fails (>30% or NDC <3), the measurement system is not acceptable.

Q: What is %R&R in Gage R&R?

%R&R (also called % Study Variation or %GRR) is the proportion of observed variation caused by the measurement system. Formula: %R&R = (6×σmeasurement system)/(6×σobserved)×100%. It indicates the relative usefulness of the gage for control charting and process capability analysis. Low %R&R means most variation you observe is real process variation, not measurement error. %R&R 30% not acceptable.

Q: What is NDC in Gage R&R?

NDC (Number of Distinct Categories) represents the number of non-overlapping confidence intervals that span product variation range, or how many distinct groups your measurement system can discern. Formula: NDC = √2×(σpart-to-part/σR&R). NDC ≥ 5 is excellent (can detect meaningful process changes), NDC = 3-4 is marginal (limited discrimination), NDC < 3 is poor (essentially a go/no-go gage). Higher NDC indicates better measurement system discrimination power.

Q: What is attribute Gage R&R?

Attribute Gage R&R (Attribute Agreement Analysis) evaluates measurement systems for pass/fail, good/bad, or OK/KO decisions. Protocol: 3 operators evaluate 50 parts (balanced 50% good/50% bad) 3 times each. Metrics include: % repeatability (operator consistency with self), % reproducibility (operator-to-operator agreement), % accuracy vs reference, % concordant appraisals, false reject rate, and false accept rate. ALL agreement percentages must be ≥90% for an acceptable system. Below 80% requires operator retraining or process redesign.

Q: How do you improve a failing Gage R&R?

For high repeatability (equipment variation): calibrate/maintain equipment, increase measurement resolution (1/10th rule), improve fixturing, control environment, simplify test method. For high reproducibility (operator variation): train operators, clarify procedures with detailed work instructions, reduce subjectivity, use digital vs analog displays, implement error-proofing fixtures. For attribute systems: create crystal-clear operational definitions with visual examples, provide limit samples showing borderline cases, optimize inspection environment.

Q: What is the AIAG Gage R&R method?

The AIAG (Automotive Industry Action Group) Gage R&R method is the industry standard for continuous measurement systems. Protocol: select 10 parts representing full process variation, have 3 operators each measure every part 3 times in random order (total 90 measurements). Analyze using ANOVA to calculate variance components. Evaluate using three metrics: %R&R (study variation), %Tolerance (acceptance risk), and NDC (discrimination power). Sample size requirements: ≥90 measurements is good, 40-89 acceptable, <40 not acceptable.

Master Measurement System Analysis (MSA) with Gage R&R studies. Learn repeatability, reproducibility, acceptance criteria (%R&R, %Tolerance, NDC), continuous and attribute Gage R&R methods with step-by-step examples.

What is Gage R&R?

Gage R&R (Repeatability and Reproducibility) is a statistical method to assess the quality of a measurement system. It quantifies how much variation in measurements comes from the measurement system itself versus the actual process variation.

Gage R&R answers critical questions: "Can I trust my data? Do different operators get the same results? Does repeating a measurement give consistent values?"

Measurement System Analysis (MSA)

Gage R&R is part of Measurement System Analysis (MSA), which validates the accuracy (measuring the true value) and precision (getting consistent results) of your measurement system before collecting data for your Six Sigma project.

Why Validate Your Measurement System?

Too many business problems are analyzed with non-reliable data. If data quality is poor, you must stop and fix it during the Measure phase before proceeding with analysis.

Avoid wrong decisions: Bad measurements lead to incorrect conclusions about process performance
Prevent Type I errors (Producer risk): Scrapping good parts thinking they're defective
Prevent Type II errors (Customer risk): Shipping bad parts thinking they're good
Enable reliable capability analysis: Process capability calculations require valid measurement systems
Support control charts: SPC requires confidence that variation is real, not measurement error

Type I Error (Producer Risk)

Problem: Measurement system rejects good parts as defective.

Impact: Scrapping good products, increased costs, lower yield than actual.

Example: A diameter measuring 10.00mm (within spec 9.95-10.05mm) is measured as 10.08mm and rejected.

Type II Error (Customer Risk)

Problem: Measurement system accepts bad parts as good.

Impact: Shipping defective products to customers, quality escapes, customer complaints.

Example: A diameter measuring 10.07mm (out of spec) is measured as 10.02mm and shipped to customer.

Accuracy vs Precision

A complete Measurement System Analysis evaluates both Accuracy (measuring the true value) and Precision (consistency of measurements).

Accuracy

Definition: How close measurements are to the true value (reference standard).

Accuracy Components:

Bias: Systematic offset from true value
Linearity: Constant bias across measurement range
Stability: Consistent measurements over time
Resolution: Ability to detect small changes (1/10th rule)

Test method: Measure reference standards and compare to known values.

Precision (Gage R&R)

Definition: How consistent measurements are when repeated.

Precision Components:

Repeatability (Equipment Variation): Same operator, same part, repeated measurements
Reproducibility (Appraiser Variation): Different operators, same part
Operator×Part Interaction: Certain operators struggle with certain parts

Test method: Gage R&R study with multiple operators and parts.

Variance Decomposition

σ²_observed = σ²_{true process} + σ²_{measurement system}

σ²_{measurement system} = σ²_{repeatability} + σ²_{reproducibility}

σ²_{reproducibility} = σ²_operators + σ²_{operator×part interaction}

Continuous Gage R&R Method

For continuous data (measurements like diameter, weight, temperature), use the AIAG (Automotive Industry Action Group) standard Gage R&R method.

Study Design - AIAG Recommended Protocol

n = 10 parts (minimum) selected to represent full process variation
p = 3 repetitions — each operator measures each part 3 times
q = 3 operators — at least 3 different operators
Total measurements: n × p × q = 10 × 3 × 3 = 90 measurements

Sample Size Requirements

Good: n×p×q ≥ 90 — Reliable results. Recommended for important characteristics.

Acceptable: 40 ≤ n×p×q < 90 — Adequate but less confident. Use when constraints exist.

Not Acceptable: n×p×q < 40 — Insufficient data. Increase parts, repetitions, or operators.

Gage R&R Acceptance Criteria

Metric	Good (<10%)	Acceptable (10-30%)	Not Acceptable (>30%)
% Study Variation (%R&R)	< 10%	10% - 30%	> 30%
% Tolerance (%P/T)	< 10%	10% - 30%	> 30%
Number of Distinct Categories (NDC)	≥ 5	3 - 4	< 3

Critical: If ANY metric is in the red zone (>30% or NDC < 3), your measurement system is NOT acceptable. You MUST fix the gage issue before continuing your Six Sigma project! Proceeding with bad data will lead to wrong conclusions.

Understanding the Three Gage R&R Metrics

1. % Study Variation (%R&R or %GRR)

Formula: %R&R = (6 × σ_{measurement system}) / (6 × σ_observed) × 100%

What it measures: Proportion of observed variation caused by the measurement system.

Why it matters: Indicates relative usefulness of the gage for control charting and process capability analysis. Low %R&R means most variation you observe is real process variation, not measurement error.

Example: If %R&R = 8%, then 8% of what you observe comes from the gage, and 92% is true process variation. Excellent!

2. % Tolerance (%P/T)

Formula: %P/T = (6 × σ_{measurement system}) / (USL - LSL) × 100%

What it measures: Measurement system variation relative to specification tolerance.

Why it matters: Indicates relative usefulness for determining part acceptance/rejection. If measurement variation is large compared to tolerance, you'll make many wrong accept/reject decisions.

Example: Tolerance = 0.10mm, 6σ_measurement = 0.015mm. %P/T = 15%. Acceptable but should improve.

3. Number of Distinct Categories (NDC)

Formula: NDC = √2 × (σ_part-to-part / σ_R&R)

What it measures: Number of non-overlapping confidence intervals that span the product variation range. Represents how many distinct groups your measurement system can discern.

Why it matters: Indicates measurement system's ability to detect differences in the measured characteristic. Higher NDC = more discrimination power.

NDC ≥ 5: Excellent discrimination. Can detect meaningful process changes.
NDC = 3-4: Marginal. Limited ability to distinguish parts.
NDC < 3: Poor. Essentially a go/no-go gage.

Continuous Gage R&R Calculation Example

Scenario: Measuring shaft diameter (mm)

Specification: 10.00mm ± 0.05mm (LSL=9.95mm, USL=10.05mm, Tolerance=0.10mm)
Study design: 10 parts, 3 operators, 3 repetitions = 90 measurements
Results from ANOVA:
- σ_observed = 0.0187mm
- σ_R&R (measurement system) = 0.0041mm
- σ_part-to-part = 0.0182mm

Step 1: Calculate % Study Variation (%R&R)

%R&R = (6 × σ_R&R) / (6 × σ_observed) × 100%

%R&R = (6 × 0.0041) / (6 × 0.0187) × 100%

%R&R = 0.0246 / 0.1122 × 100%

%R&R = 21.9%

Acceptable (10-30%) but should be improved

Step 2: Calculate % Tolerance (%P/T)

%P/T = (6 × σ_R&R) / (USL - LSL) × 100%

%P/T = (6 × 0.0041) / (10.05 - 9.95) × 100%

%P/T = 0.0246 / 0.10 × 100%

%P/T = 24.6%

Acceptable (10-30%) but improvement recommended

Step 3: Calculate Number of Distinct Categories (NDC)

NDC = √2 × (σ_part-to-part / σ_R&R)

NDC = 1.414 × (0.0182 / 0.0041)

NDC = 1.414 × 4.44

NDC = 6.3 → rounded to 6

Good (≥ 5) — excellent discrimination ability

Interpretation

Overall Assessment: The measurement system is marginally acceptable.

%R&R = 21.9% (Yellow zone) — measurement system contributes 22% of observed variation. Should be improved for better process capability analysis.
%P/T = 24.6% (Yellow zone) — measurement variation consumes about 25% of tolerance. May cause some accept/reject errors.
NDC = 6 (Green zone) — good discrimination power. Can distinguish 6 different quality levels.

Recommendation: Use the gage but prioritize improvements. Investigate main sources of variation (repeatability vs reproducibility) to target improvements.

Attribute Gage R&R Method

For attribute data (pass/fail, good/bad, OK/KO decisions), use the Attribute Agreement Analysis method.

Study Design - Recommended Protocol

3 operators evaluate the same 50 parts
3 repetitions — each operator evaluates each part 3 times (blind — don't see previous evaluations)
Reference standard: Expert panel determines true value (good/bad) for each part
Balanced sample: Ideally 50% good and 50% bad parts in the sample
Total evaluations: 50 parts × 3 operators × 3 repetitions = 450 evaluations

Attribute Gage R&R Metrics

Without Considering Reference (Precision):

% Repeatability (Operator 1, 2, 3): How consistent is each operator with themselves?
% Reproducibility (All Operators): Do all operators agree on each part?

By Considering Reference (Accuracy):

% Accuracy vs Reference (Operator 1, 2, 3): How often does each operator match the reference?
% Accuracy All Operators vs Reference: Overall accuracy across all operators
% Concordant Appraisals: (# measures equal to reference / total measures) × 100%
% OK rated KO: False reject rate (Producer risk)
% KO rated OK: False accept rate (Customer risk)

Acceptance Criteria

% Agreement Metric	Good (≥90%)	Acceptable (80-89%)	Not Acceptable (<80%)
Repeatability (each operator)	≥ 90%	80% - 89%	< 80%
Reproducibility (all operators)	≥ 90%	80% - 89%	< 80%
Accuracy vs Reference (each operator)	≥ 90%	80% - 89%	< 80%
Concordant Appraisals	≥ 90%	80% - 89%	< 80%
% OK rated KO (False Reject)	≤ 1%	1% - 5%	> 5%
% KO rated OK (False Accept)	≤ 1%	1% - 5%	> 5%

Critical: For an acceptable attribute measurement system, ALL agreement percentages must be ≥ 90% (ideally). If ANY metric is below 80%, the measurement system is NOT acceptable. Retrain operators, clarify inspection criteria, or redesign the inspection process.

Attribute Gage R&R Example

Scenario: Visual inspection of welds (Good/Bad)

Study design: 30 parts (15 Good, 15 Bad per reference), 3 operators, 2 repetitions
Total evaluations: 30 × 3 × 2 = 180 evaluations
Reference: Expert welding inspector determined true quality

Results

Repeatability (Operator consistency with self):

Operator A: 27/30 parts consistent = 90.0% (Good)
Operator B: 24/30 parts consistent = 80.0% (Acceptable)
Operator C: 28/30 parts consistent = 93.3% (Good)

Accuracy vs Reference:

Operator A: 54/60 evaluations match reference = 90.0% (Good)
Operator B: 46/60 evaluations match reference = 76.7% (Not Acceptable)
Operator C: 55/60 evaluations match reference = 91.7% (Good)

Error Rates:

% OK rated KO (False Reject) = 1.1% (Good)
% KO rated OK (False Accept) = 6.7% (Not Acceptable)

Interpretation

MEASUREMENT SYSTEM NOT ACCEPTABLE

Critical Issues:

Operator B's accuracy (76.7%) is below 80% threshold — needs retraining
False accept rate (6.7%) is too high — bad welds are being shipped to customers!
Operator B's repeatability (80%) is marginally acceptable but concerning

Action Required: STOP production inspection. Retrain Operator B on weld defect identification. Clarify inspection criteria with visual aids. Rerun Gage R&R after training.

How to Improve a Failing Gage R&R

For High Repeatability (Equipment Variation)

Improve equipment: Calibrate, maintain, or replace worn gages
Increase resolution: Use more precise measurement device (1/10th rule)
Improve fixturing: Better clamping, positioning, or alignment
Control environment: Temperature, humidity, vibration
Redesign test method: Simplify measurement procedure

For High Reproducibility (Operator Variation)

Train operators: Standardized training on measurement procedure
Clarify procedures: Detailed work instructions with photos
Reduce subjectivity: Use objective criteria, limit judgment calls
Automate reading: Digital displays vs analog scales
Poka-yoke (error-proofing): Fixtures that force correct measurement position

For Attribute Measurement Systems

Operational definitions: Crystal-clear defect definitions with visual examples
Limit samples: Reference samples showing borderline cases
Reduce categories: If using 3+ categories (Good/Marginal/Bad), consider reducing to 2
Better lighting/environment: Optimize inspection conditions
Regular recertification: Periodic operator requalification

Automated Gage R&R with AI

DMAIC Suite™ automates Gage R&R calculations, ANOVA analysis, and AI-powered interpretation for both continuous and attribute MSA.

Automated Calculations

One-click %R&R, %Tolerance, NDC calculation
Automatic ANOVA with variance components
Attribute agreement analysis (repeatability, reproducibility, accuracy)
Visual charts: X-bar, Range, by Operator

AI Interpretation

AI explains if measurement system is acceptable
Identifies main contributors (equipment vs operators)
Suggests specific improvement actions
Flags operators needing retraining