QE Capability Enhancement (4) | How to Set MSA Plans: Which Gauges, How Deep, and How Often to Re-evaluate

A manufacturing company was undergoing a customer process audit. The quality manager presented a stack of measurement system analysis reports, a total of eighteen, which seemed to cover "all that needed to be done." The auditor asked only two questions: How many gauges do you have in your gauge inventory? For the over-tolerance complaint last quarter, which gauges were used to measure the dimensions, and have they been analyzed?

The answers were embarrassing. The inventory listed over 1,200 gauges, and the eighteen reports covered only ordinary dimensions that did not require special attention. The two critical dimensions involved in the complaint were measured using a dedicated inspection tool that had not been evaluated for three years and a homemade go/no-go gauge from the production line. The auditor issued a nonconformity, stating bluntly: The measurement system analysis was not planned based on risk, and the coverage was insufficient to support the conclusions on product characteristics.

This is not an issue of execution but of planning. MSA is never about "submitting a report," but rather about "using limited resources to control measurement risk within an acceptable range." The plan must address three questions: which gauges, how deep, and how often to re-evaluate.

1. Key Principles: Variability Decomposition and Risk Grading

The entire logic of measurement system analysis is based on an additive relationship: the variability observed in the data equals the variability of the part itself plus the variability introduced by the measurement system.

σ² (observed) = σ² (part) + σ² (measurement system)

The variability of the measurement system can be broken down into repeatability (the same person using the same gauge repeatedly), reproducibility (changing the operator or the setup), bias and linearity (inaccurate measurements), and stability (drift over time). GR&R manages the first two, bias and linearity manage the third, and stability manages the fourth. The cost of analyzing these four aspects increases sequentially, so it is impossible to perform a full analysis on all gauges. Risk grading must be applied. This is the fundamental reason for the existence of an MSA plan.

Grading typically involves two dimensions: the importance of the characteristic being measured (safety/legal characteristics > key characteristics > general characteristics) and the performance limitations of the gauge itself (dedicated inspection tools, homemade fixtures, attribute gauges, and gauges that rely on human eyes for readings have significantly higher risks than mature digital universal gauges). The intersection of these two dimensions determines the priority for more or less detailed analysis.

2. Practical Steps: Five Steps to Turn the Plan into a Table

Step One: Create a "Characteristic—Gauge" Mapping Matrix to Ensure Coverage. Match each characteristic listed in the control plan to the specific gauge number in the gauge inventory, filling in each cell. This matrix is the foundation of the MSA plan and is the easiest step to skip, yet it is the one that most clearly exposes problems. Criteria: The analysis coverage for gauges used to measure safety and legal characteristics must be 100%, key characteristics ≥95%, and general characteristics ≥80%. Any blank rows in the matrix are mandatory items for the annual plan.

Step Two: Determine "Depth" by Level, Avoiding Both Over- and Under-Analysis. It is recommended to use three levels. Level A (safety/legal characteristics, customer-specified gauges): Full analysis including GR&R, bias, linearity, and stability, plus measurement uncertainty evaluation. Level B (key characteristics): GR&R and bias, with stability checked through daily standard part inspections. Level C (general characteristics): Only GR&R, and a simplified two-person, two-time method is allowed. Criteria: No gauge’s %GRR should be exempted from analysis due to "tolerance not being strict." If a characteristic has no impact on assembly, it should be removed from the control plan rather than skipping the analysis.

Step Three: Set a Unified "Pass Line" for Evaluation. When %GRR is calculated with tolerance as the denominator: <10% is acceptable; 10% to 30% is conditionally acceptable, requiring three additional measures (relaxed judgment, increased monitoring, and time-limited rectification) and written customer approval; >30% is unacceptable and cannot be used for release judgment. Additionally, resolution must be verified: the gauge’s minimum readable value should not exceed 1/10 of the tolerance band, and the number of distinguishable part categories (ndc) should be ≥5. Sample size and personnel: For a conventional GR&R, use 10 parts (covering the range of process variation, not from the same batch) × 3 operators × 3 repetitions, totaling 90 measurements. For destructive measurements, use a nested design and explain the batch assumption. For attribute gauges, use 30 to 50 samples, with at least 20% being nonconforming or borderline, and each sample should be re-evaluated 2 to 3 times. At least one unskilled operator must be included to avoid systematic underestimation of reproducibility.

Step Four: Set Frequencies and Document Trigger Conditions. The cycle is just a minimum, and trigger conditions are the main focus. It is suggested that Level A gauges be re-evaluated annually, Level B gauges every two years, and Level C gauges every three years or through sampling of similar gauges (for batch gauges of the same model and specification, sample 10% and no less than 3). The following seven events must unconditionally trigger re-evaluation: gauge replacement or major repair, gauge relocation or movement, changes in measurement methods or procedures, operator changes or additions, calibration results exceeding or nearing the upper limit of tolerance, %GRR in the 10% to 30% risk range, and customer requests. Criteria: Evaluation must be completed within 15 working days after a trigger event, and the release of characteristics measured by the gauge must be tightened (100% inspection or double-check by two people) until the evaluation is complete.

Step Five: Create an Auditable Plan Table and Close the Loop. The table fields should include at least: gauge number, name, characteristic measured, characteristic level, analysis items, sample size, judgment criteria, responsible person, cycle, last completion date, next due date, and conclusions with rectification status. Criteria: The proportion of overdue items should be ≤5%, and gauges with "unacceptable" conclusions must have a disposal decision (replacement or repair, change in measurement method, or change in judgment criteria) within 30 days. Simply writing "operators have been trained" is considered an incomplete loop.

3. Common Pitfalls

Pitfall One: Evenly Distributing Resources by Gauge Count, Spending on Unimportant Dimensions. Some companies, to make their inventory look good, perform GR&R on ordinary calipers and tape measures that do not affect judgment, while overlooking homemade inspection tools from the production line. The correct order is to first assess the importance of the characteristic, then the risk of the gauge, and finally the quantity.

Pitfall Two: Using a Single %GRR for All Similar Dimensions. The denominator of %GRR is the tolerance. The same gauge measuring two dimensions with different tolerances can yield completely different conclusions: a dimension with a tolerance of ±0.5 may be acceptable, but a dimension with a tolerance of ±0.05 will inevitably exceed the limit. Criteria must be evaluated for each characteristic individually, not just by gauge number.

Pitfall Three: Equating Calibration Compliance with Measurement System Compliance. Calibration only answers whether the measurement is accurate, not whether it is stable or consistent across operators. A calibration certificate with all green results can still have a %GRR exceeding 30%, a common misjudgment in the field.

Pitfall Four: Sampling from the Same Batch of Continuous Parts. Part variation is compressed to near zero, making the denominator smaller and artificially inflating %GRR, leading to the conclusion that "the gauge is not working." Conversely, deliberately selecting parts with significant differences can mask real issues. Samples must cover the actual range of process variation and document the selection rationale.

Pitfall Five: Uniformly Setting the Cycle to "Once a Year," Only Doing Regular Re-evaluations. The most common loophole on the shop floor is: if the gauge is not broken, it is not re-evaluated, even when operators change, methods are altered, or the gauge is moved. The cycle is only a baseline, and trigger conditions prevent omissions.

Pitfall Six: Selecting Only Skilled Operators for GR&R. The consistency between skilled and new operators is precisely the variability that needs to be exposed. Excluding new operators from the evaluation hides systematic risks in the reports.

4. Self-Check List

• I have a "Characteristic—Gauge" mapping matrix, with 100% coverage for safety/legal characteristics, and all blank rows have been included in the annual plan.
• The %GRR criteria for each characteristic are clearly defined, and conditional acceptance (10% to 30%) includes additional measures and customer confirmation.
• The gauge resolution does not exceed 1/10 of the tolerance band, ndc ≥ 5, and the samples for attribute gauges include at least 20% nonconforming or borderline parts.
• The seven trigger conditions for re-evaluation are documented, and evaluations are completed within 15 working days after a trigger, with tightened release criteria during this period.
• Gauges with "unacceptable" conclusions are 100% given a disposal decision (replacement, method change, or judgment criteria change) within 30 days.

The value of an MSA plan lies not in the thickness of the reports but in the ability to immediately retrieve the analysis records for a gauge when a customer points to a dimension and asks, "How dare you release it?"

Planning based on risk for coverage, depth, and frequency is the key to a robust plan.

Knowledge code: 6.2.1

Version: v20260914

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously improve their quality capabilities.