GR&R Exceeds Limits, Don’t Rush to Replace the Measuring Tool — A Five-Step Method for Identifying and Addressing Measurement System Failures
1. Common Mistakes After Exceeding Limits
%GR&R measures 32%, and the first reaction at the quality meeting is often: "The measuring tool is no good, replace it!" So, they apply for an imported measuring tool, retrain the staff, and measure again—only to get 26%. Another, more subtle mistake is to repeatedly re-measure until a set of acceptable data is "measured," and then submit the report to meet the requirements. Both approaches treat GR&R (Gage Repeatability and Reproducibility study) as an "exam for the measuring tool," but GR&R evaluates the entire measurement system, which includes people, machines, materials, methods, and environment. Exceeding the limit is just a symptom, and the root cause could be in any of these areas. Replacing the measuring tool is often ineffective because the tool typically contributes only a portion of the variation. What GR&R truly answers is whether the measurement system can distinguish between conforming and nonconforming products in the process variation—tool precision is just one input.
Remember the criteria: %GR&R should not exceed 10% to be acceptable; 10% to 30% is conditionally acceptable, with the conditions clearly stated and continuous monitoring required; over 30% is unacceptable and must be improved. The correct sequence after exceeding the limit is to first identify the issue, then address it, rather than immediately replacing the measuring tool.
2. Step One: Check the Data, Then the System
When you receive the GR&R report, don't rush to analyze it. Go back to the original records and check three things: whether there are any out-of-control points on the range chart; whether the readings by the same evaluator for the same part are excessively different; and whether the data is "suspiciously neat"—for example, multiple readings are exactly the same, which often indicates data copying or fabrication. Under normal conditions, the range chart should mostly fall within the control limits, and the readings should show "consistent variation" rather than "complete uniformity."
Out-of-control points, uncalibrated measuring tools, and mixed-up records can all inflate the %GR&R. An out-of-control range chart indicates instability in the measurement process, making the calculated %GR&R meaningless. Address the stability first before discussing capability. If the data is problematic, redo the study; if the data is fine, proceed to the next step. Making decisions based on bad data will lead to incorrect actions at every step.
3. Step Two: Identify the Source of Variation by Repeatability and Reproducibility
GR&R breaks down measurement variation into two components, and identification starts from these two areas.
Poor repeatability, where the same person measures the same part repeatedly with divergent results, often indicates issues with the measuring tool itself, clamping, or the environment: tool wear, insufficient resolution, unstable clamping surfaces, or temperature fluctuations in the workshop.
Poor reproducibility, where different people measuring the same part show significant differences, often indicates issues with the method or the people: inconsistent measurement techniques (e.g., different force applied with a caliper, different reading postures), ambiguous judgment standards, or inadequate training.
Practical tips: Plot the data from three evaluators separately. If one evaluator's readings are systematically higher, go to the site to see "how they measure." A "zipper-like" cross pattern in the interaction chart between evaluators and parts suggests that different people have different feels for different parts, which is common in measuring soft materials or complex surfaces. Most of the time, the answer lies in the technique, not the tool.
4. Step Three: Check if the Sample Selection is Correct
The samples for GR&R must cover the range of process variation. If all "good parts" or parts from the same batch are selected, the variation between parts is too small, amplifying the measurement error and leading to an inflated %GR&R. NDC (Number of Distinct Categories) below 5 is often due to this reason—the measurement system cannot even distinguish between parts, let alone be considered qualified.
Be cautious of part internal variation mixing into measurement error: for soft materials, thin-walled parts, or parts with unstable clamping surfaces, measuring the same part in different positions yields different results. This is not the fault of the tool but rather the sampling and measurement points not being well defined. Too few sample parts (less than 8) can also distort the results. Remember three rules for sample selection: cover the tolerance band, cover the process variation, and ensure each part can be repeatably positioned.
5. Step Four: Address the Issues—Methods First, Then Tools, Finally Equipment
After identifying the issues, address them in the order of "methods first, then tools, and finally equipment."
Method issues are the most common: standardize measurement techniques, create clamping fixtures or dedicated inspection tools, and document the measurement steps in a work instruction. Tool issues are the next most common: calibrate, repair, or replace with a higher-resolution tool according to the 1/10 rule—the tool's resolution should not exceed one-tenth of the tolerance band. Personnel issues: provide training and qualification, and re-evaluate those who do not meet the standards.
The sequence is crucial: start with the least expensive method changes, then move to tool adjustments, and finally consider equipment replacement. Many cases where "replacing the tool doesn't help" are actually due to method issues being blamed on the tool. Document the actions: update the work instructions, archive training records, and prevent the same issues from recurring next month.
6. Step Five: Re-measure to Confirm and Incorporate into Routine Monitoring
After addressing the issues, redo the GR&R to confirm that it falls within the acceptable range. For results in the 10% to 30% conditionally acceptable range, specify the usage conditions, responsible person, and re-evaluation cycle.
Incorporate re-evaluation into the routine mechanism: after major repairs, calibration expiration, personnel changes, or product model changes, re-conduct MSA and link it with the tool inventory to avoid "one-time fixes" and forgetting about it.
A real example: a factory used an internal micrometer to measure hole diameters, with a %GR&R as high as 31%. Even after replacing the micrometer, the %GR&R was still 27%. Following the five steps: the data was correct, the sample coverage was reasonable, and the variation was entirely due to technique—three evaluators applied different forces, and the parts were not properly positioned. After installing a V-shaped positioning frame and standardizing the "light touch and stop" technique, the re-measured %GR&R dropped to 8.6%, and the NDC increased from 3 to 9. Since then, they have conducted re-evaluations quarterly and have not exceeded the limit again. A set of fixtures and a work instruction solved what a new tool could not. At the end, self-check with three questions: Is the data correct? Is the main source of variation from people or tools? Is the re-evaluation scheduled? If all three questions are answered positively, the issue is truly closed.
Exceeding the %GR&R limit is just a symptom. Identify the issue first, then address it: check the data, break down the variation, and then consider the tool.
Knowledge code: 6.2.1
Version: v20260829
Author: Quality Think Tank
Quality Think Tank is dedicated to providing systematic professional knowledge, methodologies, and practical tools to quality management practitioners, helping companies continuously improve their quality capabilities.