Data collection planning
A data collection plan specifies what data to collect, how, when, where, by whom, and in what form. Without one, the team collects the wrong data or data that cannot answer the hypothesis. The plan must be built before any data is collected in Measure.
Elements of a data collection plan
| Element | Decision to make |
|---|---|
| Measure (Y) | What CTQ characteristic are you measuring? Is it continuous or attribute? |
| Operational definition | Exact, unambiguous description of what counts as a defect or how the measurement is made — so any operator produces the same result |
| Data type | Continuous (weight, time, dimension) or attribute (pass/fail, count of defects) |
| Sampling strategy | Random, systematic, stratified, or rational subgroup? Size n per subgroup, number of subgroups |
| Stratification factors | Variables to track alongside Y — shift, operator, machine, material lot — that may explain variation later in Analyze |
| Data collection form | Check sheet, electronic system, or automated sensor — who fills it in and when |
| Measurement system | Which gauge or method? MSA completed (or in parallel)? |
Operational definitions — why they matter
An operational definition removes ambiguity. Compare:
- Vague: "Count defective tablets" — every operator has a different threshold
- Operational: "A defective tablet is any tablet with a visible chip ≥0.5 mm, coating color deviation ΔE > 2.0, or weight outside 490–510 mg as measured on the calibrated bench scale (asset ZZ-SC-004) at the end of the coating step"
Continuous vs attribute data
Continuous data gives more information per measurement and requires smaller sample sizes for the same statistical power. Convert attribute problems to continuous wherever possible (e.g., measure the actual dimension rather than go/no-go).
Common exam traps
- Stratification factors must be recorded at collection time. Shift, lot or machine information that was not recorded with the measurement often cannot be reliably reconstructed later.
- Sample size for a baseline study: the AIAG Statistical Process Control manual recommends 25 or more subgroups containing 100 or more individual readings (for example, 25 subgroups of 4 or 5) before calculating control limits and capability. Smaller samples give wide confidence intervals on Cpk/Ppk.
- Rational subgroups: Parts in the same subgroup should have been produced under conditions as similar as possible; variation between subgroups should capture the process variation you want to study.
Try 5 free CSSBB practice questions on this topic. No account needed.
Take the free CSSBB diagnosticMore: DMAIC overview · Project charter · VOC & CTQ · SIPOC · MSA in DMAIC · Process capability · Data collection planning · Graphical analysis · Root cause analysis · Multi-vari analysis · Hypothesis testing · Regression · DOE · FMEA · Define tollgate · Change management · Financial benefits · Team dynamics · Improve tollgate · Piloting & scaling · Poka-yoke (BB) · Solution selection · Control plan · SPC in DMAIC · Sustaining gains
CSSBB Quest by Bizarre Act. Original study material; not affiliated with ASQ. CSSBB is a certification of ASQ. Terms · Privacy