Gage R&R
Gage repeatability and reproducibility (Gage R&R) is a statistical study that measures the variation introduced by a measurement system rather than the parts being inspected. It evaluates repeatability, the equipment variation when one appraiser measures the same part repeatedly, and reproducibility, the variation between different appraisers. By comparing combined measurement error against product tolerances or total process spread, Gage R&R determines whether inspection data is reliable for quality control, capability analysis, and process improvement.
- Five shafts on one study sheet
- This sheet records a gage R&R study on one shaft micrometer. Five shafts fill the rows, with reference diameters from 9.96 to 10.07 mm. Three appraisers, Ana, Ben, and Cho, measure each shaft twice. Every dot is one reading on a 9.90 to 10.10 mm scale. The dark tick marks each shaft's reference diameter. The shafts differ by up to 0.11 mm. The gage must resolve differences that small.
- Repeatability: one appraiser, two trials
- Ana measures shaft 2 twice with the same gage. Her two indigo dots on one lane are those two readings. The gap between them is repeatability, also called equipment variation. On the good gage the 15 pairs average 0.004 mm apart. On the poor gage they average 0.008 mm apart. Dividing the mean range by 1.128 gives repeatability as one sigma.
- Reproducibility: three appraisers, one shaft
- Cho's amber dots sit right of Ana's and Ben's on every row. Cho reads about 0.02 mm high on every shaft, both trials. That offset is reproducibility, also called appraiser variation. It comes from the spread between appraiser averages, divided by 1.693. On the poor gage it computes to 0.012 mm, larger than repeatability. The instrument is the same; the method differs between people.
- The %GRR readout
- Gage R&R combines repeatability and reproducibility: the square root of the sum of squares. Multiplying by 5.15 covers 99% of measurement error. Dividing by the 0.20 mm tolerance gives %GRR. Under 10% is acceptable. From 10% to 30% is marginal. Over 30% is unacceptable. The good gage scores 9%. The poor gage scores 36%, so its readings cannot support a pass or fail decision.
- The tolerance it is judged against
- Every row's scale runs from 9.90 to 10.10 mm. Those are the specification limits, 10.00 ± 0.10 mm. The distance between them is the 0.20 mm tolerance. %GRR reports how much of that tolerance measurement error consumes. On the poor gage, error takes 36% of the tolerance. A shaft near either limit could then pass one day and fail the next.
Key facts
- Core components
- Repeatability (equipment) and reproducibility (appraiser)
- Acceptable threshold
- Under 10% of tolerance or process spread
- Marginal threshold
- 10% to 30% of tolerance or process spread
- Unacceptable threshold
- Over 30% of tolerance or process spread
- Standard multiplier
- 5.15 (spans 99% of normal distribution)
By Matthew Savas — Founder of Kaizumi. Reviewed 1 September 2026.
Gage repeatability and reproducibility (Gage R&R) is a statistical study that evaluates the amount of variation in a measurement process that comes from the measurement system itself rather than from the items being measured. In quality engineering and manufacturing operations, observed process data reflects both true product variation and the error introduced by tools, fixtures, procedures, and human operators. A Gage R&R study isolates these sources of error by having multiple appraisers measure the same set of parts across multiple independent trials. The study quantifies repeatability, which is the variation observed when one person measures the same part multiple times with the same device, and reproducibility, which is the variation observed between different people measuring the identical part. These components combine to yield the total measurement system variation, which is evaluated against the product tolerance or the total process spread.
Core components of measurement variation
Measurement systems analysis categorizes total observed variation into distinct components. Total process variation consists of true part-to-part variation combined with measurement system error. Gage R&R focuses strictly on assessing the measurement system error, which breaks down into equipment variation and appraiser variation:
- Repeatability: Also designated as equipment variation, repeatability represents the inherent precision of the measurement device. It is the variation recorded when a single appraiser measures the exact same characteristic on the exact same part multiple times under identical conditions. High repeatability error indicates mechanical instability, loose fixturing, electronic noise, or insufficient device resolution.
- Reproducibility: Also designated as appraiser variation, reproducibility reflects the difference between the average readings obtained by different appraisers measuring the same parts using the same measurement tool. High reproducibility error points to differences in operator technique, variations in part positioning, inconsistent visual alignment, or ambiguous measurement instructions.
When repeatability and reproducibility are mathematically combined, they define the total gage variation. This metric indicates whether a measurement tool and its associated operating protocol are capable of accurately inspecting parts, monitoring processes, or generating data for statistical analysis.
Mathematical computation and the average-and-range method
The Automotive Industry Action Group (AIAG) standardizes several methods for computing Gage R&R, with the average-and-range method being the most widely deployed. This method relies on range calculations across trials and appraisers, converted to standard deviations using standard statistical d2 constants based on sample subgroup sizes.
To conduct the calculation:
- The range between trials is calculated for every appraiser and part combination. The mean of these ranges is computed to establish the average trial range.
- Repeatability is calculated by dividing the average trial range by the d2 constant corresponding to the number of trials, then multiplying by the standard distribution multiplier.
- The average measurement value for each appraiser across all parts is calculated. The appraiser range is the difference between the highest appraiser average and the lowest appraiser average.
- Reproducibility is calculated by dividing the appraiser range by the d2 constant corresponding to the number of appraisers, multiplying by the standard distribution multiplier, and subtracting a correction factor. The correction term is calculated as repeatability squared divided by the product of the number of parts and the number of trials. This term removes the portion of repeatability error that naturally enters appraiser averages. If this correction term exceeds the initial appraiser variation term, reproducibility is set to zero.
- Total Gage R&R is calculated as the square root of the sum of repeatability squared and reproducibility squared.
The standard calculation uses a multiplier factor of 5.15, which spans 99 percent of a normal distribution. Multiplying 5.15 by the standard deviation of the measurement system defines the full width of the interval within which a single measurement reading is expected to fall due to measurement noise alone.
Worked example: machined shaft measurement
Consider a machine shop turning steel shafts to an engineering specification of 10.00 millimeters plus or minus 0.10 millimeters. This specification defines a total allowable tolerance of 0.20 millimeters. To verify the measurement system, the team executes a standard study: 5 parts, measured by 3 appraisers, across 2 trials, producing 30 total readings on a single study sheet.
Five shafts are selected to represent the operational range of the turning process, with reference diameters of 9.96, 9.99, 10.02, 10.04, and 10.07 millimeters.
The good gage scenario
In the first scenario, a precision digital micrometer mounted on a stable stand is tested:
- Across the 15 pairs of trials, the readings sit close together, averaging a difference of 0.004 millimeters between repeat measurements on the same shaft.
- The average readings of the three appraisers agree with one another within 0.001 millimeters.
- Repeatability computes to 0.004 millimeters (0.0035 millimeters before rounding).
- In the reproducibility calculation, the correction term exceeds the raw appraiser range term, setting reproducibility to 0.000 millimeters.
- Total Gage R&R computes to 0.004 millimeters.
Dividing the Gage R&R value of 0.004 millimeters by the total tolerance of 0.20 millimeters and multiplying by 100 yields a percent Gage R&R of 9 percent.
The poor gage scenario
In the second scenario, a worn mechanical caliper is evaluated across the same five shafts by the same three appraisers: Ana, Ben, and Cho:
- The trial pairs spread apart, averaging a difference of 0.008 millimeters on repeated measurements.
- Cho consistently applies heavier clamping force than Ana and Ben, reading approximately 0.02 millimeters higher on every shaft across both trials.
- Repeatability computes to 0.007 millimeters.
- Reproducibility computes to 0.012 millimeters due to the persistent offset in Cho's technique.
- Total Gage R&R computes to the square root of 0.007 squared plus 0.012 squared, which equals 0.014 millimeters.
Multiplying that 0.014 millimeters by the 5.15 study factor gives a measurement spread of 0.072 millimeters. Dividing 0.072 millimeters by the 0.20 millimeter tolerance yields a percent Gage R&R of 36 percent.
Practical consequence of measurement error
A 36 percent Gage R&R means that measurement error alone consumes more than one-third of the total allowable part tolerance. Under this condition, measurement noise directly causes two types of operational errors:
- False rejection: A shaft with an actual physical diameter of 10.08 millimeters is within the specification upper limit of 10.10 millimeters. Measurement error can cause it to read as 10.11 millimeters, leading the inspector to scrap or rework an acceptable part.
- False acceptance: A defective shaft with an actual physical diameter of 10.11 millimeters can read as 10.08 millimeters, causing nonconforming material to ship directly to the customer.
Conducting a 30-reading study requires approximately one hour of labor. In contrast, operating with an unvalidated gage yielding 36 percent error can lead an engineering team to spend months attempting to solve manufacturing process instability when the observed variation exists entirely within the measurement process. In this poor gage scenario, Cho's 0.02 millimeter discrepancy is an operational method issue. Establishing a written standard operating procedure and executing a brief calibration session corrects the reproducibility error without requiring capital expenditure for new tooling.
Acceptance criteria and interpretation
The AIAG outlines standard benchmark thresholds for evaluating the percentage of tolerance or process variation consumed by Gage R&R:
- Under 10 percent: The measurement system is acceptable. It possesses sufficient resolution and consistency to monitor process shifts, verify quality tolerances, and support statistical process control.
- 10 percent to 30 percent: The measurement system is marginal. It may be acceptable based upon the criticality of the application, the cost of the measurement device, or the economic consequences of part failure.
- Over 30 percent: The measurement system is unacceptable. The measurement error is too large relative to the specification limits or the process spread. The system must be improved through hardware upgrades, fixture redesign, calibration, or operator retraining before drawing conclusions from inspection data.
Integration with quality and continuous improvement frameworks
Gage R&R serves as an essential preliminary step within structured problem-solving and process control methodologies:
- Six Sigma and DMAIC: During the Measure phase of a DMAIC project, teams must validate data integrity before analyzing root causes. A Gage R&R study confirms that the data collected in the Measure phase reflects true process behavior rather than measurement noise. If measurement variation is unaddressed, subsequent steps in the Analyze and Improve phases risk optimizing false signals.
- Capability analysis: Process capability indices such as Cpk compare actual process spread to engineering tolerances. Because total measured variance includes measurement variance, an elevated Gage R&R artificially inflates the standard deviation calculation, deflating the reported Cpk value and falsely indicating that a capable process is failing.
- Statistical process control: A control chart relies on subgroup averages and ranges to detect special causes of variation. Excessive measurement variation expands control limits, obscuring genuine process shifts and reducing the sensitivity of the control chart.
- Lean workflow balancing: Accurate cycle time measurement is critical when balancing assembly lines. Using tools such as a Free Yamazumi builder to balance manual operations requires repeatable observation data; large timing measurement discrepancies between observers distort line-balancing calculations.
Common misconceptions and pitfalls
Several operational assumptions frequently compromise the integrity of measurement systems:
- Calibrated equipment guarantees acceptable measurement capability: Calibration checks only whether a tool reads accurately against a certified standard under controlled conditions at zero or set points. A calibrated micrometer with an active calibration sticker can easily produce an unacceptable 36 percent Gage R&R if operators apply inconsistent pressure, position parts at incorrect angles, or read analog vernier scales inconsistently.
- Gage R&R is an intrinsic property of the tool: A measurement error of 0.014 millimeters is independent of the part drawing. Its 0.072 millimeter study spread is 36 percent of a tight 0.20 millimeter tolerance, making the system unacceptable. The same device against a loose 1.00 millimeter tolerance consumes only 7 percent of the specification width, rendering it fully acceptable.
- Confusing display resolution with accuracy and repeatability: A digital caliper displaying four decimal places does not guarantee repeatability to four decimal places. Electronic resolution often exceeds the mechanical precision and human reproducibility of the measurement process.
Applications across industries
Measurement systems analysis applies to any operational environment where human judgment or physical tools generate data used for decision-making.
Manufacturing
Prior to launching a Six Sigma initiative to reduce dimensional variability on a precision machining line, a project team conducted a Gage R&R study involving 3 operators measuring 10 production parts 3 times each. The initial study revealed that 35 percent of the total observed process variation stemmed directly from the measurement system. Reproducibility accounted for the majority of the error, as operators mounted the components into the test fixture using different clamping sequences. Standardizing the clamping procedure and adding alignment guide pins reduced measurement error to 6 percent, allowing the team to accurately measure true tooling wear.
Healthcare
In a clinical study assessing patient blood pressure monitoring procedures, researchers evaluated measurement variability among clinical staff. Multiple nurses recorded blood pressure readings on a set of standardized test subjects. The initial analysis revealed that 25 percent of the total variation in recorded blood pressure resulted from measurement system inconsistency, driven primarily by reproducibility differences. Variation was traced to inconsistent cuff placement, improper cuff sizing, and varying patient rest periods prior to recording the measurement. Implementing a standardized protocol for cuff positioning and enforcing a five-minute rest period reduced the measurement contribution to 8 percent.
Transactional and service operations
Measurement systems analysis also applies to subjective visual inspections and administrative scoring systems. In a customer contact center, 4 quality assurance supervisors evaluated and scored 15 recorded customer service calls. An initial attribute Gage R&R study demonstrated that 45 percent of the variation in call quality scores resulted from differences in how individual supervisors interpreted evaluation criteria. Conducting calibration sessions to align scoring definitions and publishing a clarified grading rubric with concrete examples reduced measurement variation to 12 percent, establishing consistent performance evaluations across customer service teams.
Frequently asked questions
- What constitutes an acceptable Gage R&R percentage?
- According to Automotive Industry Action Group benchmarks, a Gage R&R result under 10 percent indicates an acceptable measurement system. Results between 10 percent and 30 percent are marginal and may be accepted depending on application criticality, part failure risk, or equipment costs. Any measurement system with variation exceeding 30 percent is unacceptable and requires corrective action before inspection data can be trusted.
- Does a calibrated tool guarantee an acceptable Gage R&R result?
- No, calibration checks only whether a tool reads accurately against a certified standard under controlled conditions. An instrument with an active calibration sticker can easily fail a Gage R&R study if operators use inconsistent clamping force, position parts incorrectly, or read scales subjectively. Gage R&R evaluates the entire operational process—including fixturing, human technique, and environmental noise—rather than just the device's static baseline.
- Is Gage R&R an intrinsic property of the measurement device?
- A Gage R&R percentage is not an intrinsic property of a tool because it is calculated relative to a specific tolerance or process spread. A measurement variation of 0.014 millimeters consumes 36 percent of a tight 0.20 millimeter tolerance, rendering the tool unacceptable for that application. That exact same variation consumes only 7 percent of a loose 1.00 millimeter tolerance, making the tool fully acceptable.
- How does excessive measurement error affect process capability studies like Cpk?
- Because observed process variation combines actual part variation with measurement noise, high Gage R&R artificially inflates the calculated standard deviation. This inflated standard deviation directly deflates the reported Cpk index. Consequently, an unvalidated measurement system can make a capable manufacturing process appear nonconforming and out of statistical control.
- What operational risks occur when a measurement system has high Gage R&R?
- Excessive measurement variation leads directly to false rejection and false acceptance. False rejection happens when measurement noise makes an acceptable, in-spec component measure as nonconforming, creating unnecessary scrap and rework expenses. False acceptance occurs when measurement error causes a genuinely defective part to read within allowable limits, resulting in defective material being shipped to customers.