DMAIC without the belt: running the five phases in a lean operation
The Lean Six Sigma improvement cycle, stripped to what a mid-market team can actually execute alongside its day job.
DMAIC — Define, Measure, Analyse, Improve, Control — is the improvement cycle underneath most process work, including ours. It has a reputation for ceremony: charters, tollgates, belts, a statistician. In a company with 400 people and no dedicated continuous-improvement function, that version does not survive contact with the quarter.
Here is the version that does.
Define — one page, one number
The only artefact that matters is a problem statement with a number in it. Not “the close is slow” but “the month-end close takes 11 working days against a target of 5, and the delay costs us X in delayed billing.”
Two failure modes to avoid: defining the solution in the problem statement (“we need an automated reconciliation tool”), and defining a scope larger than one process owner can authorise. If closing the problem requires three departments to agree, it is a programme, not an improvement.
Measure — count 30, not 3,000
Teams stall here waiting for perfect data. You do not need it. Thirty observations of a real transaction, timed by hand, will tell you where the time goes with enough confidence to act. Instrument properly later, once you know which step is worth instrumenting.
Measure three things: how long each step takes, how long work waits between steps, and how often work comes back. That last one — rework — is the figure most operations cannot produce on demand, and it is usually where the recoverable cost is concentrated.
Analyse — find the constraint, then stop
The analysis phase attracts more effort than it repays. You are looking for one thing: the step that sets the pace of the whole process. Everything upstream of the constraint is producing inventory it cannot use; everything downstream is idle waiting.
A quick test: if you doubled the capacity of this step, would total throughput increase? If no, it is not the constraint, and improving it produces a tidier process with the same output.
Improve — smallest change that moves the number
Rank candidate changes by effort against the measured cost of the problem, and take the smallest one that moves the number. Resist the redesign. A process that improves 30% next month beats one that improves 80% after a nine-month implementation, both in cash terms and — more importantly — in whether the organisation still believes in the programme.
This is also the phase where automation belongs, and only here. Automating in the Measure phase locks in the current process. Automating after Analyse locks in the improved one.
Control — the phase everyone skips
Improvements decay. Someone leaves, a workaround creeps back, an upstream system changes. Control is what stops the gain being temporary, and it needs exactly three things:
- A number reported weekly to the process owner — the same number from the Define phase.
- A documented standard for the new process, written by the people who run it.
- A trigger that says what happens when the number moves the wrong way, and who acts.
If you only have capacity for two phases, make them Measure and Control. Organisations rarely lack ideas for improvement; they lack the measurement to know which idea worked and the discipline to keep it working.
Where this fits with automation
DMAIC and automation are not alternatives. The cycle tells you what to change and whether the change held. Automation is one option inside the Improve phase — frequently the right one, occasionally the expensive way to preserve a process that should have been deleted.
That order is the whole argument: process first, technology second.
Our four-phase methodology is DMAIC adapted to a fixed-fee engagement — Discover, Analyse, Design, Deploy — with the Control mechanics built into the deployment phase.