Measure Phase
Purpose and deliverables
Measure establishes a valid baseline: what the process does now, with trustworthy data. Deliverables: operational definitions, data collection plan, measurement system analysis (MSA), baseline capability, and a refined problem statement.
Core concepts
- Operational definition: makes a measure unambiguous (what counts as a defect, exactly).
- Data types: continuous vs. discrete (attribute); drives tool choice.
- MSA: Gage R&R for continuous data (repeatability = equipment, reproducibility = appraiser); attribute agreement analysis for discrete. Rule of thumb: %GRR < 10% acceptable, 10–30% conditional, >30% unacceptable.
- Sampling: random, stratified, systematic; sample size affects confidence.
- Baseline capability: Cp/Cpk (continuous, normal), DPMO, sigma level, yield (RTY = product of step yields).
- Control charts to see whether the process is stable before judging capability.
Traps in questions
- Computing capability on an unstable process - first check stability with a control chart.
- Confusing repeatability (same appraiser, same equipment) with reproducibility (different appraisers).
- Using Cp when the process is not centered - Cpk accounts for centering.
- Vague operational definitions that create measurement disagreement.
- Treating attribute data with continuous-only tools.
Rule of thumb
Stability → capability. Never report Cpk from a process with out-of-control points; fix the special cause first.
Worked example
A process has mean 50, spec 40–60 (USL−LSL = 20), σ = 2. Cp = 20 / (6×2) = 1.67. If the mean shifts to 53, Cpk = min((60−53), (53−40)) / (3×2) = 7/6 ≈ 1.17 - same spread, worse capability because it is off-center.