Analyse: Finding and Verifying Root Causes
Analyse turns data into verified causes. Start with process mapping and value stream mapping to see where flow breaks, then use stratification to split data by shift, machine, supplier or operator - patterns often hide in aggregate figures.
Graphical tools are examined heavily: histograms (shape, spread, centring), run charts and control charts (distinguishing common from special cause), scatter plots (association, not proof), box plots (compare distributions), Pareto charts (prioritise the vital few) and fishbone/Ishikawa or 5 Whys for cause generation. Note the trap: fishbone and 5 Whys generate hypotheses; they do not verify them.
Verification uses hypothesis tests. Know when to use: 1-sample t (mean vs target), 2-sample t (two means), paired t (before/after on same units), ANOVA (three or more means), chi-square (association between categorical variables), F-test (equal variances), and regression/correlation (relationship and prediction). Interpret p-values correctly: p below alpha means the observed difference is unlikely under the null - it does not measure effect size or importance, so report confidence intervals and practical significance too. Watch for the multiple-comparison trap and for confusing statistical with practical significance.
Worked example / rule of thumb: defect rates by shift are 4.1%, 3.9% and 6.2%. A chi-square test gives p = 0.01, so shift matters - then stratify further by machine before concluding shift is the cause. Correlation is not causation: only designed experiments confirm cause.