DMAIC
Define, Measure, Analyse, Improve, Control: a five-phase route for problems whose causes are not yet known.
About 6 minutes to read.
Key idea
Some problems announce their own fix: the printer is broken, so mend the printer. Others are stubborn. Deliveries are late, but nobody agrees why, and every fix so far has faded within a month. DMAIC is built for the stubborn ones. It slows the team down at the start, so that it speeds up at the end.
Each phase ends with a tollgate, a short review with the sponsor, where the team shows its evidence and agrees to move on. Tollgates stop a project leaping from a hunch straight to a solution.
The five phases
- Define: agree the problem, the goal, the customer and the scope, in a charter everyone signs up to.
- Measure: decide exactly what to measure, check the measurement can be trusted, and record how the process performs today.
- Analyse: list the possible causes and test them with data until the few that really matter stand out.
- Improve: design changes that remove those causes, try them on a small scale, and confirm they work.
- Control: put measures and routines in place so the gain holds, then hand the process back to its owner.
Example: late deliveries
A distributor’s first project
A building supplies distributor promises next-day delivery and keeps only 78 per cent of those promises. In Define, the team writes a charter: raise on-time delivery to 95 per cent within four months, for orders placed before 2pm. In Measure, they find that “on time” meant different things to sales and transport, and agree one definition before collecting data.
Analyse shows that most late orders were picked after the 3pm loading cut-off because picking lists printed in one batch at noon. Improve tries rolling print runs every half hour on one depot. On-time delivery climbs to 96 per cent. Control adds a daily chart and a check at the cut-off, and the depot manager takes ownership.