Minitab Crossed Gage R&R: A Task Guide for Gating Process Capability Analysis

The Problem: Capability Numbers Are Meaningless If the Gage Lies

Process capability indices such as Cp and Cpk compare process spread to specification limits. They assume the measurement system can reliably tell one part from another. When gage error is large, a capability study can label a stable process as incapable, or worse, certify scrap as conforming. The practical takeaway is to treat measurement system validation as a gate: run Minitab's Gage R&R Crossed study first, and only proceed to capability analysis if the measurement variation is small relative to part variation and tolerance.

Desired Outcome

The goal is a quantified assessment of measurement system variation relative to part-to-part variation and total variation. Minitab reports several acceptance metrics:

• %Contribution: the percentage of total variance attributable to the gage (repeatability plus reproducibility).
• %Study Variation: the standard deviation of the gage expressed as a percentage of total study standard deviation.
• %Tolerance: the gage standard deviation as a percentage of the tolerance width, available only when engineering specification limits are supplied.
• Number of distinct categories (ndc): the number of distinct part levels the measurement system can actually distinguish across the observed part variation.

The engineering decision rule is straightforward: halt process capability or tolerance analysis until the measurement system meets acceptance criteria, then document the study as objective evidence for your quality system.

Necessary Prerequisites

A crossed Gage R&R requires a crossed design: every operator measures every part the same number of times. If any operator misses a part, or if parts are nested within operators (for example, each operator measures a different set of parts), the crossed model is invalid and you must use a Nested or Mixed Effects design instead.

Prepare a clean Minitab worksheet with at least three columns:

Column	Contents
Part	Identifier for each distinct part (e.g., 1–10)
Operator	Identifier for each appraiser (e.g., A, B, C)
Measurement	Numeric measured value

There must be no missing cells. A common starting layout is 10 parts, 3 operators, and 2 trials, producing 60 rows. If you intend to report %Tolerance, you also need engineering specification limits—lower and upper—that come from design intent or customer requirements, not from the measured data itself.

The brief assumes use of the ANOVA method for balanced crossed data. ANOVA (Analysis of Variance) partitions total variation into repeatability, reproducibility, and part-to-part components. This method assumes approximately normal residuals; severe non-normality can bias the variance estimates.

Focused Procedure

Run the study from the Minitab worksheet containing your data. No special permissions are required beyond normal file access.

Step 1: Enter or import the crossed data

Arrange the data in the three-column layout above. For example, Part 1 measured by Operator A in Trial 1 occupies one row. Keep part and operator labels consistent; "A" and "a" are treated as different operators.

Step 2: Open the Gage R&R Crossed dialog

Choose Stat > Quality Tools > Gage Study > Gage R&R Study > Crossed.... Menu labels can vary slightly between Minitab releases, so confirm the path in your version.

Step 3: Assign variables

• Part numbers: select the Part column.
• Operators: select the Operator column.
• Measurement data: select the Measurement column.

Step 4: Choose method and options

• Method: select ANOVA.
• Options: if you have engineering limits, enter the lower and upper spec limits to enable %Tolerance reporting. Alternatively, select %Study Variation if no tolerance is defined.

Step 5: Run and review

Click OK. Minitab produces a session-window summary, an ANOVA table, variance component estimates, and graphical output including interaction and measurement-by-part plots. Do not rely on a single number; review the components.

Expected Checks and Acceptance Criteria

After the study runs, apply these checks before allowing capability analysis to proceed:

1. %Contribution under 10% and %Study Variation under 30% are widely used acceptance thresholds for a measurement system to be considered adequate. Systems between these limits and stricter cutoffs may be conditionally acceptable depending on the cost of error.
2. ndc of 5 or higher. An ndc below 5 means the system cannot resolve enough distinct part levels to support meaningful capability indices or control charting.
3. Repeatability as the dominant gage component when the system is acceptable. If reproducibility is large, the problem is usually operator technique or training, not the gage itself.
4. Residual normality. Check the residual plots from the ANOVA output. Strong skew or outliers can invalidate the variance component estimates.
5. Non-negative, stable variance components. Negative variance components or wildly unstable estimates across replicates indicate a design or data problem.

Recovery Options When Criteria Fail

If the study fails, do not simply re-run the same data and hope for a better result. Address the root cause:

• Operator retraining or standardization when reproducibility dominates. Standardize fixturing, reading technique, and environmental conditions.
• Gage maintenance, calibration, or replacement when repeatability dominates. Check for wear, dirt, or inadequate resolution.
• Increase measurement resolution if the gage cannot discriminate between parts. This may mean a more precise instrument or a finer scale.
• Improve the fixture or measurement method to reduce setup variation.
• Expand the part sample if part variation is too small. A narrow part range inflates %Gage and deflates ndc, producing a false failure. Select parts that span the full expected process range before re-running.

Limitations and Practical Verification

Several limitations affect interpretation. First, menu names and default options for %Study Variation versus %Tolerance can differ between Minitab releases; confirm the selected method in the dialog before running. Second, unbalanced or missing cells force a switch to Nested or Mixed methods, which change the interpretation rules. Third, specification limits used for %Tolerance must be engineering limits; using data-derived limits will misstate the result. Fourth, a small part sample or narrow part range can inflate %Gage and deflate ndc, giving a false failure of the measurement system.

To verify the result, re-run the Gage R&R with the same data and confirm that %Study Variation, %Contribution, and ndc are reproducible and stable across runs. On a small subset, compare Minitab's Repeatability and Reproducibility variance components to manual calculations to validate your interpretation. Finally, review the operator-by-part interaction plot for unusual patterns, and confirm that the selected parts span the expected process range and that operators represent normal production usage.