Understanding Gage R&R Results for Measurement System Acceptability

1. The Key Acceptance Criteria

After running a Gage R&R study, statistical software provides several key metrics to evaluate your measurement system. The Automotive Industry Action Group (AIAG) has established standard guidelines for these metrics, which are widely used in manufacturing.

There are four primary criteria we use to judge the capability of a measurement system:

1. % Study Variation (%StudyVar): The percentage of the total variation in the study that is consumed by measurement error.
2. % Tolerance (P/T Ratio): The percentage of the customer's specification width (tolerance) that is consumed by measurement error.
3. % Contribution: The percentage of total study variance (the square of standard deviation) attributable to the measurement system.
4. Number of Distinct Categories (NDC): An index that represents how many separate groups of parts the measurement system can distinguish within the tested sample.

The first step in interpretation is deciding which metric is most relevant to your objective.

2. Choosing Your Primary Metric: Process Control vs. Product Control

The two most common metrics are % Study Variation and % Tolerance. The one you choose depends on the primary purpose of your measurement.

• Use % Tolerance (P/T Ratio) for Product Control: If your main goal is to sort parts—to determine if they conform to specifications and can be shipped to the customer—then % Tolerance is your metric. It directly compares the measurement error to the specification limits.

• Use % Study Variation for Process Control: If your main goal is to understand and reduce the variation in your process, perhaps using control charts, then % Study Variation is your metric. It compares the measurement error to the natural variation of the process itself.

The following video explains this choice and introduces the fundamental concepts of interpreting the results.

3. The AIAG Acceptance Guidelines

Once you've chosen your primary metric, you compare your results against the industry-standard AIAG guidelines. These are typically presented in a "traffic light" format (Green, Yellow, Red).

Let's break down these rules:

Note: % Contribution uses different thresholds (1% and 9%) because it's based on variance, not standard deviation. It's more sensitive to measurement error.

A measurement system must be acceptable across all relevant criteria. The most critical metrics for a go/no-go decision are % Study Variation (or % Tolerance) and the Number of Distinct Categories.

This video provides a practical walkthrough of finding these values in a Minitab output and applying the rules.

4. What to Do in the "Marginal" Zone (The Yellow Light)

An "Acceptable" result is great, and an "Unacceptable" result clearly requires action. But what do you do with a "Marginal" result (e.g., %StudyVar is 22%)?

A marginal gage may be acceptable depending on the context. You must consider the risks:

• Producer's Risk (False Rejects): You produce a perfectly good part, but the measurement error makes it look bad, so you scrap it. This costs money and reduces efficiency.
• Consumer's Risk (False Accepts): You produce a defective part, but the measurement error makes it look good, so you ship it to the customer. This creates quality problems and damages your reputation.

When your gage is marginal, you must weigh these risks against the cost and difficulty of improving the measurement system. If the measurement is for a non-critical feature, or if the process is highly capable (e.g., Cpk > 1.67), you might decide to accept the marginal system. If the feature is critical to safety or function, you must improve the measurement system.

5. Troubleshooting an Unacceptable System: Graphical Analysis

If your Gage R&R is unacceptable or marginal, the next step is to diagnose the problem. Is the issue with the equipment (Repeatability) or the operators (Reproducibility)? The graphical output from your study provides the clues.

To understand how to read these charts, please study the following article. It provides a chart-by-chart guide to interpreting Minitab's Gage R&R six-pack report.

Here is a summary of what to look for:

• Components of Variation Chart: Your primary diagnostic tool. It visually shows if the largest component of gage error is Repeatability or Reproducibility. If Reproducibility is high, you have an operator problem. If Repeatability is high, you have an equipment or method problem.
• R Chart by Operator: Assesses repeatability. All points should be in control. An out-of-control point indicates that an operator was inconsistent when measuring a specific part.
• Xbar Chart by Operator: Assesses the system's ability to distinguish parts. This chart should look out of control, with most points outside the control limits. This shows that the variation between parts is much greater than the measurement error.
• Measurements by Operator (or Box Plot): Compares operators. The average lines for each operator should be roughly equal. If one operator is consistently higher or lower, you have a reproducibility problem.
• Part-Operator Interaction Plot: Looks for inconsistencies where some operators measure certain parts differently than others. The lines should be parallel, indicating no interaction.

6. A Practical Caution: The Part Selection Trap

Finally, even with perfect execution and interpretation, a Gage R&R study can be misleading if the initial part selection was flawed. As we discussed in the last lesson, the parts for the study must represent the full range of normal process variation.

The following case study illustrates this pitfall perfectly and introduces a useful "sanity check" metric called Category Resolution.

The key takeaway is that you can't just trust the %Gage R&R value alone. By calculating the Category Resolution (Part Range / Number of Distinct Categories), you get a practical feel for your gage's power. For example, a resolution of 2.06 lbs means your scale can only reliably distinguish between weights that are more than 2.06 lbs apart. If your tolerance is +/- 1 lb, this gage is clearly not up to the task, no matter what the %Gage R&R says.

Conclusion

You now have the tools to move beyond simply generating Gage R&R results to confidently interpreting them and making sound business decisions. This is a core competency for a Six Sigma Green Belt.

Key Takeaways:

• A measurement system is judged "Acceptable," "Marginal," or "Unacceptable" based on AIAG guidelines for four key metrics: %StudyVar, %Tolerance, %Contribution, and Number of Distinct Categories (NDC).
• The go/no-go criteria are generally <10% for %StudyVar/%Tolerance and an NDC of 5 or more.
• Marginal systems (10-30% %StudyVar) require a risk assessment: weigh the cost of improvement against the risks of making bad parts (consumer's risk) or scrapping good ones (producer's risk).
• If a system is unacceptable, use the graphical outputs to diagnose whether the root cause is with the equipment (Repeatability) or the operators (Reproducibility).
• Always perform a sanity check on your results, considering the range of parts used in the study and what the Category Resolution implies about the gage's practical discrimination power.

Preview of the Next Lesson:
We have focused on continuous data, which involves measurements on a scale. But what about inspections that result in a simple pass/fail or a defect classification? In our next lesson, we will tackle this by learning how to perform an attribute agreement analysis to validate a visual inspection system.