Measure
The Measure domain quantifies the current process and proves the data are trustworthy. ICGB tests whether you validate the measurement system before calculating baseline capability.
Measurement system analysis first
- Gage R&R: % study variation below 10% is acceptable, 10–30% is conditional, above 30% is unacceptable.
- Attribute agreement: use for pass/fail or category judgments.
- Check resolution, bias, linearity, and stability before trusting data.
Data collection
Write operational definitions, classify data as continuous or discrete, and define the sampling plan. Stratify by shift, machine, operator, or time to expose hidden variation. Convenience samples produce confident but wrong baselines.
Baseline capability
- Cp = (USL − LSL) / 6σ; it ignores centering.
- Cpk = min(USL − mean, mean − LSL) / 3σ; it penalizes off-center processes.
- For discrete data, report DPMO, DPU, or rolled throughput yield.
Worked example
A process has mean 50, standard deviation 2, LSL 44, and USL 56. Cp = (56 − 44) / (6 × 2) = 1.00. Cpk = min(56 − 50, 50 − 44) / (3 × 2) = 6 / 6 = 1.00. If the mean drifts to 52, Cp remains 1.00 but Cpk = min(4, 8) / 6 = 0.67. Reasoning: capability fell even though spread did not change, because the process moved toward the upper specification. Report Cpk for centered performance.
Traps
- Calculating capability before MSA.
- Reporting Cp alone when the process is off-center.
- Mixing subgroup sigma with individual-data sigma.
- Sampling only good shifts or easy-to-reach units.
- Confusing defects with defective units.