Measure: Baselines, Measurement Systems and Capability
Measure establishes a trustworthy baseline. The candidate must understand data types (continuous vs discrete/attribute) because they determine which tools are valid later. A operational definition makes each metric unambiguous so two people measure the same thing the same way. Data collection plans specify what, who, when, where and how, and guard against convenience sampling.
Central to this phase is the measurement system analysis (MSA). For continuous data, Gage R&R separates repeatability (same operator, same part) from reproducibility (different operators) and expresses variation as a percentage of study or tolerance variation; a common rule of thumb is that under 10% is acceptable, 10–30% may be acceptable depending on risk and application, and above 30% needs improvement. For attribute data, attribute agreement analysis checks agreement within and between appraisers against a standard. Skipping MSA is a classic exam trap: a capability study on an untrusted gauge is meaningless.
Also required: baseline performance (DPMO, yield, cycle time), process capability (Cp, Cpk, Pp, Ppk), and sigma level conversion. Cpk accounts for centring; Cp ignores it, so a process can have high Cp and poor Cpk. Ppk uses overall variation, Cpk uses within-subgroup variation.
Worked example / rule of thumb: with a lower spec of 10, mean 12 and standard deviation 0.5, Cpk = (12 − 10) / (3 × 0.5) = 1.33. Before quoting it, ask whether the gauge's R&R is acceptable - if the measurement variation is 40% of the total, the number is not decision-worthy.