Summary
In short
- Percentage of tolerance and percentage of study variation are different questions with different answers. Reporting one without saying which produces confident numbers that mean different things to different readers.
- Percentage of tolerance asks whether the measurement can judge conformance. Percentage of total variation asks whether it can distinguish parts for process improvement.
- Part selection determines the answer when using study variation as the denominator. Parts spanning the process range produce a favourable result; parts from one shift produce a poor one, with the same gauge.
- The number of distinct categories, sometimes called ndc, should be at least five. Below that the system cannot reliably separate parts into enough groups to control a process.
- Reproducibility dominating repeatability points at method, training or fixturing rather than at the instrument. Replacing the gauge will not fix it.
- The study must reflect real conditions: the actual appraisers, the actual method, the actual location, in random order and ideally blind.
What it is
What it is
What is a gage R&R study?
A measurement system analysis quantifying repeatability, the variation when one appraiser measures the same part repeatedly, and reproducibility, the variation between appraisers measuring the same parts. Together they express how much observed variation comes from the measurement system rather than from the parts.
What do the acceptance thresholds mean?
Under AIAG MSA guidance, a gage R&R below 10 percent is generally acceptable, 10 to 30 percent may be acceptable depending on the importance of the application and the cost of improvement, and above 30 percent is considered unacceptable. The number of distinct categories should be five or more.
When to use it
When to use it, and when not to
This study assesses a measurement system for a specific characteristic. It is not calibration.
Use it for
- Qualifying a measurement system before it is used for control or conformance decisions
- New gauge, new fixture, new method or new characteristic
- After a change in appraisers, location, fixturing or measurement procedure
- Where capability studies produce unexpected results and the measurement is a candidate cause
- Periodic requalification where a customer or scheme requires it
Not for
- Calibration, which compares the instrument to a traceable standard and says nothing about reproducibility
- Bias, linearity and stability studies, which are separate elements of measurement system analysis
- Process capability studies, which assume the measurement system has already been qualified
- Attribute agreement analysis, which is the equivalent method for pass or fail judgements
- Instrument selection, which follows from the resolution and capability the application requires
Standards
What it is built against
Measurement system analysis is mandated in automotive quality management and is good practice wherever measurement drives decisions.
| Clause | Requirement | Where it lands |
|---|---|---|
| IATF 16949 cl.7.1.5.1.1 | Measurement systems analysis on each type of inspection, measuring and test equipment system in the control plan | Header |
| AIAG MSA | Reference manual defining gage R&R methods, acceptance criteria and the number of distinct categories | Study design |
| ISO 9001 cl.7.1.5 | Monitoring and measuring resources suitable and fit for purpose | Conclusion |
| ISO 9001 cl.9.1.1 | Methods for monitoring and measurement to ensure valid results | Conclusion |
| ISO/IEC 17025 cl.7.2 | Selection, verification and validation of methods, including measurement uncertainty | Results |
| IATF 16949 cl.9.1.1.1 | Statistical process control including analysis of measurement variation | Results |
| ISO 22514 series | Statistical methods in process management, capability and performance | Results |
| ISO 9001 cl.8.5.1 | Controlled production including monitoring and measurement at appropriate stages | Conclusion |
What it does not cover
- Calibration, which establishes traceability and says nothing about reproducibility between appraisers.
- Bias, linearity and stability studies, which are separate MSA elements addressing different properties.
- Process capability studies, which assume the measurement system is already qualified.
- Attribute agreement analysis, the equivalent method where the judgement is pass or fail.
- Measurement uncertainty budgets, required under ISO 17025 and computed differently.
Filling it in
Filling it in well
Design the study to reflect reality, choose the denominator deliberately, and read what the components are telling you.
Where percentage of study variation is the denominator, part selection determines the result. Parts spanning the range the process actually produces give a meaningful answer; parts taken from a single shift understate part variation and make the measurement system look worse than it is. Where percentage of tolerance is used, part selection matters less and the tolerance must be stated.
The people who actually take the measurement, in the location they take it, with the fixturing and method they use. A study conducted in a laboratory by a metrologist qualifies a measurement system nobody operates. Randomise the order and, where practicable, keep appraisers blind to previous results.
Repeatability dominating points at the instrument, its resolution, or the fixturing. Reproducibility dominating points at the method, the training or how parts are located, and no amount of buying a better gauge will address it. The split is the most actionable output of the study and is frequently ignored in favour of the headline percentage.
At least five, per AIAG guidance. This asks whether the system can separate parts into enough distinct groups to support process control. A system can produce an acceptable percentage against a wide tolerance while being unable to distinguish parts at all, and ndc is what reveals it.
Audit findings
Common audit findings
Findings here concern study design more often than measurement performance.
| Finding | Clause | What fixes it |
|---|---|---|
| Result reported without stating the denominator used. | AIAG MSA | State whether against tolerance or study variation; they answer different questions. |
| Parts selected from a narrow range where study variation is the denominator. | AIAG MSA | Span the process range; narrow selection understates part variation. |
| Study conducted by metrologists rather than the actual appraisers. | IATF 16949 cl.7.1.5.1.1 | Use the people who take the measurement in production. |
| Number of distinct categories below five. | AIAG MSA | The system cannot separate parts adequately for control regardless of the percentage. |
| Repeatability and reproducibility not reported separately. | AIAG MSA | The split tells you whether to address the instrument or the method. |
| Measurement order not randomised. | AIAG MSA | Randomise and blind where practicable; sequence effects inflate apparent agreement. |
| Capability studies performed on an unqualified measurement system. | ISO 9001 cl.9.1.1 | Qualify measurement first; capability inherits measurement error. |
| Attribute characteristics assessed with a variable gage R&R method. | AIAG MSA | Use attribute agreement analysis for pass or fail judgements. |
| Study not repeated after a change of fixture, method or location. | IATF 16949 cl.7.1.5.1.1 | These affect reproducibility directly and are common triggers. |
| Marginal result accepted without assessing application importance. | AIAG MSA | Between 10 and 30 percent is a judgement about application and cost, recorded rather than assumed. |
Worked case
Case in point: the gauge that was not the problem
A supplier failed a customer's measurement system requirement on a critical dimension, returning a gage R&R of 34 percent. The response was to purchase a higher-resolution instrument, which cost several thousand pounds and produced a result of 31 percent.
The study report, when examined, showed repeatability at around eight percent and reproducibility at over thirty. The three appraisers were consistent with themselves and disagreed with each other, which points at how the part was being located and held rather than at the instrument reading it.
The parts were being positioned by hand against a datum face that could be interpreted two ways. A simple locating fixture, made in-house, brought reproducibility down and the combined figure below ten percent.
Definitions
Definitions and key terms
- Repeatability
- Variation when one appraiser measures the same part repeatedly with the same instrument. Equipment variation.
- Reproducibility
- Variation between appraisers measuring the same parts. Appraiser variation, pointing at method or fixturing.
- Percentage of tolerance
- Gage R&R expressed against the specification width, appropriate where measurement judges conformance.
- Percentage of study variation
- Gage R&R expressed against total observed variation, appropriate for process understanding and improvement.
- Number of distinct categories
- How many distinct groups the system can separate parts into. AIAG guidance requires at least five.
- Discrimination
- The resolution of the instrument, which should be no coarser than a tenth of the process variation or tolerance.
- Attribute agreement analysis
- The equivalent study for pass or fail judgements, assessing agreement within and between appraisers and against a standard.
- Bias
- Systematic difference between the measured average and a reference value, a separate MSA element from R&R.
FAQ
Frequently asked questions
Which denominator should we use?+
Percentage of tolerance where the measurement is used to judge conformance to specification, and percentage of total study variation where it is used to understand or improve the process. Where the measurement does both, report both. A result quoted without stating which is being used is ambiguous, and the same system can pass on one and fail on the other.
What are the acceptance criteria?+
Under AIAG MSA guidance, below 10 percent is generally acceptable, 10 to 30 percent may be acceptable depending on the importance of the application, the cost of the gauge and the cost of repair, and above 30 percent is considered unacceptable. The number of distinct categories should be five or more. A marginal result requires a recorded judgement rather than an assumption.
Does part selection affect the result?+
Substantially, where study variation is the denominator. Parts spanning the range the process actually produces give a meaningful answer. Parts taken from one shift or one machine understate part variation, which inflates the measurement system's apparent share and makes an adequate gauge look unacceptable.
What does it mean if reproducibility dominates?+
That appraisers disagree with each other more than they disagree with themselves, which points at method, training, part location or fixturing rather than at the instrument. Buying a better gauge addresses repeatability and will not move the larger component, which is a common and expensive mistake.
How does this relate to calibration?+
They answer different questions and neither substitutes for the other. Calibration establishes that the instrument reads correctly against a traceable standard. Gage R&R establishes whether the measurement system, including the people, the method and the fixturing, can distinguish what you need it to distinguish. A perfectly calibrated instrument can produce an unusable measurement system.
The agents
What the agents do with it
The study qualifies a measurement. What fails is the undeclared denominator and the component split nobody read.
Holds studies against the characteristic and the control plan, records the denominator used, and flags where capability studies rest on unqualified measurement.
Surfaces the repeatability and reproducibility split so remedial action addresses the dominant component rather than the headline figure.
Links gauges to calibration status and fixture condition, both of which affect study results and change over time.
Connects appraiser training and method standardisation where reproducibility dominates, since that is a competence and method finding.
This template lives in KnowQuality — quality and food safety. HACCP, nonconformance, traceability, laboratory and customer complaints.
Sources
Sources
- AIAG Measurement Systems Analysis reference manual
- IATF 16949:2016 clauses 7.1.5.1.1 and 9.1.1.1
- ISO 9001:2015 clauses 7.1.5 and 9.1.1
- ISO 22514 series, statistical methods in process management
- ISO/IEC 17025:2017 clause 7.2, selection, verification and validation of methods