dmaic-improve

Start with confirmed root causes from Analyze, generate multiple solution alternatives per cause, classify them on an Impact × Effort quadrant, run an FMEA on the selected solution, then pilot it and validate the improvement statistically against the Measure baseline.

FMEA (Failure Mode and Effects Analysis) evaluates what could go wrong before implementing a solution. RPN equals Severity times Occurrence times Detection, each scored 1-10; values above 200 require mandatory preventive action before piloting.

Use a pilot when the solution is new or unproven, when any FMEA mode has RPN above 100, when error costs are high, or when root cause confidence is moderate. Full rollout is acceptable only for validated solutions in low-risk, non-critical processes.

5S addresses disorganization and search waste, Kanban limits work-in-progress, SMED reduces setup times, Jidoka catches defects automatically, Heijunka levels variable demand, and MUDA elimination removes non-value-added activities.

Compare post-improvement data against the Measure baseline using a two-sample t-test for means, chi-square for proportions, or F-test for variances, requiring p < 0.05. Also compute the new Sigma Level and confirm control CTQs show no regression.

Skip it when root causes are not yet statistically confirmed, when the solution was decided without analysis, or when the solution scope exceeds the project charter. Implementing without validated causes is guessing, not DMAIC.