DMAIC step by step: deliverables and tools per phase
The five DMAIC phases of Six Sigma with tollgate deliverables, the right tools for each phase, how to write the problem statement and how to prove a root cause.
DMAIC is the five-phase roadmap of Six Sigma for improving a process that already exists. Its strength is discipline: no solution before the problem is measured, no action before the cause is proven, and no project closed before the gain is secured. Each phase ends with a short review (a tollgate) with the sponsor.
The five phases, deliverables and tools
| Phase | Deliverables at the tollgate | Free tools |
|---|---|---|
| Define | Problem statement, goal with baseline and target, scope, team, savings estimate, signed charter. | SIPOC, COPQ, Pareto |
| Measure | Operational definitions, validated measurement system, baseline capability or yield, data collection plan. | Gage R&R, Attribute agreement, Capability, FPY |
| Analyze | Possible causes collected, narrowed with data, root causes verified statistically or by switching them on and off. | Fishbone, IS / IS NOT, Multi-vari, Hypothesis tests, Regression |
| Improve | Solutions or optimal settings found, risks assessed, pilot run with data before and after. | DOE, FMEA, Sample size and power |
| Control | Control plan, control charts, standardised work, training, handover to the process owner, savings confirmed. | Control plan, Control charts, LPA, Work instruction |
Define: a charter that can be measured
A good problem statement says what, where, since when and how big, in numbers, without naming a cause or a solution: “Since March, 4.2 % of housings from line 3 fail the leak test, costing €18,000 a month.” The goal uses the same unit with a baseline, a target and a date.
Measure: trust the data first
Before the baseline, check the measurement system. A leak tester with a %GRR of 40 % produces noise, and any analysis on its data is wasted. Then measure the baseline long enough to see the normal variation, for example two to four weeks.
Analyze: prove the cause
Collect possible causes with the team, then narrow them with facts and data: where and when the problem occurs and where not, which family of variation dominates, which factors differ significantly. A cause is verified when switching it on and off switches the problem on and off.
Improve: test before rolling out
Find the solution or the settings, assess the new risks, and run a pilot. Compare before and after with a hypothesis test and check that the sample was large enough to see the difference you expect.
Control: make it last
Update the control plan and the work instructions, put the key input or output on a control chart with a reaction plan, train the people and hand the process back to its owner. Confirm the savings after a few months.
Practise the theory with the Green Belt practice test and see all tools in order in the Six Sigma hub.