What needs to be finished first
A concrete problem statement with impact (cost, quality, throughput) and initial sizing data is available.
The affected process is documented or mapped during Define, so measurement points can be identified.
What needs to be ready before start
Whiteboard or shared tool with five phases; data access (databases, logs, measurement systems); statistical tools (for example Minitab, R, Python); templates for project charter, SIPOC, process map, FMEA; steering cadence.
A Lean Six Sigma coach (Green or Black Belt); a project owner with decision authority; 3-6 project team members with process context; champion or sponsor from management; data analyst.
Project charter with problem, goal, scope, team, timeline; available measurement data; stakeholder list; current metrics (defect rate, lead time, Cost of Poor Quality); agreed phase gates.
Project typically 3-6 months, phase workshops 2-8 h per phase
Use project charter as anchor. Phase board with gate criteria: Define is complete when problem, target, and scope are clear; Measure when baseline and measurement system are validated; Analyze when root causes are evidenced; Improve when solutions are piloted; Control when sustainable embedding is in place. Fix steering dates in calendar.
The one question this method answers
Which data-supported root causes explain the problem, which solutions reduce them measurably, and how is improvement made durable in the process?
Marker: Phase
| Step | Duration | Action | Hint |
|---|---|---|---|
1Phase 1: Define | 2-3 weeks | Define project charter: problem, impact, target, scope, team, timeline. Create SIPOC. Gather Voice of Customer. Validate phase gate with champion. | If the problem remains at "we are slow," it is too vague. Required format: "Defect rate Y in process X is at A%, target is B% by date, impact is Z EUR/year." Gate is mandatory, not optional. |
2Phase 2: Measure | 3-4 weeks | Validate measurement system (repeatability, reproducibility). Collect baseline data (minimum 30 points or 4 weeks). Calculate process capability. Create stratified data by layer, machine, person. | A common mistake is to use existing data without validation. If measurement system is not validated, all further analysis is invalid. Gage R&R or equivalent is mandatory. |
3Phase 3: Analyze | 4-6 weeks | Run root-cause analysis with data: Fishbone for hypotheses, statistical tests (hypothesis testing, regression, ANOVA), process capability. Distinguish verified root causes from assumptions. | Statistics alone is insufficient. Explain a mechanism for each statistical correlation. If mechanism is implausible, the correlation is randomness or a third variable. |
4Phase 4: Improve | 4-8 weeks | Develop solutions (brainstorming, FMEA for risks). Pilot in a narrow scope and measure effect. Roll out after pilot. Document target process. | A solution without pilot is a gamble. Compare effects against baseline during pilot. Redesign solution if pilot fails; do not force rollout. |
5Phase 5: Control | 4-6 weeks, then ongoing | Set up Control Charts, document standard work, conduct training, define escalation triggers. Close project after 90-day stability and hand over to owner. | A common recurrence is improvement falling back after 6 months. Link Control Charts to action points. Owner handover must be explicit. |
What comes out at the end
Project storyboard with all five phases, charter, SIPOC, measurement system validation, baseline data, statistical analyses, solution pilots, control plan, and evidence of sustained improvement. One gate document per phase with champion signature.
One storyboard area per project with date and champion. Gate documents are dated and signed. If control re-opens in Control phase, add new iteration as an annex; preserve original storyboard.
Compact working template for DMAIC with context, input, output artifacts, and next step.
Data-driven improvement method with the phases Define, Measure, Analyze, Improve, and Control.
When and for what do we use this method?
Which data, observations, decisions, or materials are available?
Short notes along the runsheet.
Owner, date, and success signal.
Concrete filled scenario, fictional example
Concrete filled scenario, fictional example
Champion: Operations lead. Team: 4 people plus data analyst.
Problem: return rate for hardware bundle "Pro Kit" is 14.2% in Q1 2026, industry benchmark 6%, additional cost 280k EUR/year. Target: below 8% by end of Q3.
Measurement system (return categorization) showed reproducibility issue: two categories were coded inconsistently. Cleaned. Baseline n=842, mean return rate 14.2%, sigma 1.8.
Fishbone produced 9 hypotheses. Statistical tests:
Solutions: pre-fulfillment configuration checklist (pilot weeks 28-30, return rate 7.4%); instructions redesign with video walkthrough (pilot weeks 31-33, return rate 6.8% with checklist). Rollout to 100% from 18.08.
Control Chart with UCL 9%, action point at 7 consecutive values above mean. Standard Work integrated into shipping SOP. Train all 14 shipping employees. Handoff to shipping lead by 30.09., 90-day review on 30.12.
Recognize symptoms and steer against them
Define ends with "improve quality" without a concrete target or EUR/defect-rate impact.
Approve charter only when target has number, date, and impact in EUR or percentage. Otherwise return to Define.
Existing reports are treated as truth and validation is skipped.
Gage R&R or equivalent is mandatory in Measure. If measurement system is insufficient, improve it before Analyze starts.
Significant correlation is accepted as root cause without explaining causal mechanism.
Formulate a mechanism for each statistical result: why should X cause Y. If no plausible mechanism exists, probe further or check a third variable.
Improve phase moves directly to full rollout and assumes effect instead of measuring.
Pilot with comparison baseline is mandatory. Roll out only after evidence and effect are proven. Pilot each solution separately.
After project close, return rate moves back to baseline after 6 months and nobody responds.
Link Control Charts to action points and named responders. Conduct 90-day and 6-month reviews. On drift, return to Analyze phase, not repeat DMAIC mechanically.
Team jumps from Define to Improve because solution seems obvious, skipping Measure and Analyze.
Champion enforces gate discipline. If solution is clearly obvious, DMAIC is the wrong method. Use Just-Do-It actions outside DMAIC.
Done signals checkable in under a minute
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.