Scenario
You spent two weeks adjusting the process, improving Cpk from 0.98 to 1.35. Just as you were about to submit the improvement report, QA called you in: "Have you performed a Measurement System Analysis?" You paused — Measurement system? I'm using the same caliper!
The Gage R&R result came back: 38%. Half of your "improvement" might be an illusion caused by measurement instrument error.
Plain English Explanation
Gage R&R asks a fundamental question: Is your measurement tool reliable?
It breaks down into two parts:
- Repeatability (重複性): Does the same person get consistent results when measuring the same part multiple times? (An issue with the instrument itself)
- Reproducibility (再現性): Do different people get consistent results when measuring the same part? (A human factor issue)
%GRR means: What percentage of the total variation is accounted for by measurement error?
If %GRR = 38%, it means that 38% of the product variation you observe is noise introduced by the measurement tool, not actual product differences. In this situation, your Cpk figure does not accurately reflect the true process capability.
Practical Judgment
| %GRR | Assessment | Recommendation |
|---|---|---|
| < 10% | Excellent | Measurement system is reliable; Cpk can be used with confidence. |
| 10–30% | Marginal | Acceptable under certain circumstances; not recommended for high-precision processes. |
| > 30% | Unacceptable | Fix the measurement system first, then discuss process improvement. |
Most common reasons for exceeding GRR limits:
- Using contact-type measuring tools on soft materials (rubber, foam), with inconsistent application of force.
- Insufficient resolution of the measuring tool (e.g., using a 0.1mm measuring tool for a ±0.05mm specification).
- Lack of SOP for fixed measurement position and direction.
How InsightFab Works
Running Gage R&R in Minitab requires designing experimental tables, manually entering each measurement value, and identifying the primary source of variation yourself. In InsightFab, after inputting multiple measurement data points, it directly tells you: "The main problem is operator variation (Reproducibility), and it is recommended to standardize the measurement SOP."
Golden Quote
"Fix the measurement system before fixing Cpk. No matter how fast you walk in the wrong direction, you'll only arrive at the wrong destination sooner."