Lean and Six Sigma are two distinct methods, born in two different companies thirty years apart. They are sold today under a single label, which blurs the only useful question: is your problem a lead-time problem or a variation problem?
Two methods, two enemies
Lean came out of Toyota. Its enemy is waste — anything consuming time or material that the customer would not pay for: waiting, transport, inventory, rework, overproduction. Its question is “where does the time go?”
Six Sigma came out of Motorola. Its enemy is variation — the fact that the same operation does not give the same result twice. Its question is “why aren't the parts identical?”
A plant that delivers late but builds right has a Lean problem. A plant that delivers on time but scraps eight percent has a Six Sigma problem. Confusing the two means applying the wrong remedy — and failed programmes almost always start there.
DMAIC, and why the first step decides everything
Six Sigma runs on five phases: define, measure, analyse, improve, control. The temptation is to rush to the fourth. It is the first that decides the outcome.
Define means writing the problem as one sentence, with numbers and boundaries: “scrap on part 4412 rose from 2% to 6% since March, on press 3 only.” Not “we have quality issues.” An unbounded problem is never solved, only discussed.
Measure requires checking the instrument first. A gauge R&R study regularly reveals that part of the observed spread comes from the measurement system rather than the process. Improving a machine on the strength of a badly calibrated caliper is an expensive exercise.
Analyse means looking for the cause, not the culprit. Five whys and Ishikawa exist for that: they force you below the first level of explanation, the one that always names an operator.
Improve only comes next, and controlis the step everyone skips. Without a control chart and a review, the process drifts back within months — that is the rule, not the exception.
What Cp and Cpk are actually worth
Cp compares the width of your tolerance to the spread of your process. Cpk also accounts for how far off centre you are. A high Cp with a low Cpk describes a precise and common situation: your machine is capable, but it is set off target. That is good news — a setting costs less than a machine.
The common error is computing these on a sample taken the same day, on the same shift, from the same material lot. The spread you measure is then that of a moment, not of the process. Cover shifts, lots and setups — otherwise the number reassures without guaranteeing anything.
Where the effort usually fails
It rarely fails on method and almost always on data. A Six Sigma project needs reliable records, dated, tied to their lot, station and operator. When those records live in a folder on the shop floor, the measure phase takes three months and the momentum dies before analysis begins.
It also fails through ambition. A first project should cover ground you fully control: one part, one machine, one defect. Programmes that open with “improve overall quality” produce nothing but a committee.
What an ERP contributes — and what it does not
No software runs an improvement project. What it removes is the month of collection: if inspection records are entered at the station, tied to the work order, the batch and the operator, the measure phase becomes a query rather than a campaign.
It also removes the relapse. A control chart living in the system flags the drift; a printed chart pinned to a board goes yellow. That is the whole difference between a control phase performed and the same phase declared closed.
In AVIA ERP, records feed the control charts and the capability indices, non-conformities open a CAPA that must be verified effective before it can be closed, and OEE separates availability, performance and quality — the three losses Lean sets out to tell apart.
Where to start
Take the defect that costs you most over twelve months, not the one that annoys you most. Bound it to one part and one machine. Check your gauge before measuring anything. Within two weeks you will know whether your problem is lead time or variation — and it is that answer, not the label, that picks the method.