QFD: Quality Function Deployment. Table customer requirements on columns against design/solution components on rows, mark each requirement with a importance score of 1 to 5, and mark each impact/contribution of components to each requirement with a score 1 to 5. Then multiply the two scores and prioritize components based on the produced score.
FMEA: Failure Model Effect Analysis: same QFD table concept between failures and solutions.
Cause & Effect Fishbone diagram: to identify KPIVs: key process input variables & limit scope.
Process Flow Diagram: another method to find the KPIV.
Correlation: In Define-Analyze-Improve phases.
Simple correlation test:
1. Find KPIVs: through SIPOC, Fishbone, and/or Process Flow diagrams.
2. plot & line-chart KPI and KPIVs against time.
3. find which KPIV plot harmonize with KPI plot. BUT correlation doesn’t define causality, so to detect causality:
4. Run controlled cause and effect tests: freeze all other KPIV values and just change the suspect KPIV and check/plot the KPI against it.
Intermediate correlation test:
5. First run regression, build the curve then:
6. Measure the fitness of datapoints into this curve. Then you conclude the correlation level.
Advanced correlation test to add: If your data does NOT look conclusive:
7. run statistical significance test for correlation. For ordinal data: T-test for nomral and Man-Wittney (Wilcoxon) test for non-normal distributions. Wilcoxon is more robust than t-test.
H0: There is no difference between the KPI values before and after applying the change (increase or decrease) on the KPIV. I.e. the two KPI value sets of before and after the KPIV change belong to the same population.