A machining cell produces 1.2 percent oversize bores. Everyone has a theory: the casting, the tool, the night shift. Three months later the tool supplier has changed twice, the night shift has been retrained, and the rejects are still there.
DMAIC exists to stop this pattern. It is a fixed sequence of questions, and each phase ends with a decision.
Five phases, five questions
- Define: is this the right problem? A charter with a measurable gap, a target, a benefit and a team.
- Measure: how bad is it really? A checked gage, a data plan, a baseline over time, capability and cost.
- Analyze: why does it happen? Causes collected, narrowed down with data, tested and verified at the process.
- Improve: does the solution work? Solutions linked to verified causes, a pilot, the same measurement as the baseline.
- Control: will it hold? A control plan, a control chart at the process, updated standards, a handover.
The order matters. A solution chosen in Define is a guess. A cause accepted without a checked gage may be a measurement artifact.
The tollgate is a decision, not a presentation
At the end of each phase the champion, usually the manager who owns the problem, reviews the evidence and decides one of three things: continue, continue with changes, or stop.
The evidence is defined in advance:
| Tollgate | Evidence |
|---|---|
| Define | Signed charter, SIPOC, critical characteristic with limits, cost estimate |
| Measure | Gage study, baseline chart, capability, cost confirmed by the controller |
| Analyze | Tested hypotheses, causes verified at the process |
| Improve | Pilot results, new capability, confirmed benefit |
| Control | Control plan, control chart in use, handover signed |
Stopping a project is a valid result. If Measure shows that the problem costs EUR 8,000 a year instead of the assumed EUR 80,000, the best decision is to stop and free the team for a better project.
One page per tollgate
A tollgate does not need thirty slides. One page is enough: the problem statement and the limits at the top, one chart and one number for the phase in the middle, the decision needed at the bottom.
If the problem statement on that page has changed since Define, the team should say so openly. Silent changes of scope are the most common reason projects drift.
Rhythm
In a project of four to six months, tollgates fall every four to five weeks. Put all five dates into the calendar of the champion at the start. A tollgate that is postponed twice is a signal that the project has lost its sponsor.
The phases are not strictly linear. Analyze may send the team back to Measure for more data. That is fine, as long as the decision to go back is made at a tollgate and not by drifting.
DMAIC is not complicated. Its value lies in the discipline of answering each question with evidence before moving to the next one.