Lean Six Sigma in manufacturing

A line rarely fails in the way the board hears about it. Output is down, so the conversation becomes about people and shifts, when the actual losses are sitting in three places nobody is counting: the changeover that takes ninety minutes because the tooling is fetched rather than staged, the machine that runs fast but produces work the next station has to correct, and the hours the constraint spends waiting for material while everything upstream of it runs flat out.

The last of those is the most expensive misunderstanding in manufacturing. Keeping every machine busy feels like efficiency and is usually the opposite: work piles up in front of the bottleneck, lead time inflates, and the inventory hides the very problems that caused it. An hour lost at the constraint is an hour lost by the whole plant. An hour saved anywhere else is, in strict terms, worth nothing.

The second failure is quality that is inspected rather than built. Inspection at the end finds defects after all the value has been added to them, which is the most expensive possible moment to find one. The method moves detection upstream, towards the step that causes the variation, and then removes the cause rather than adding another check.

A first manufacturing project is nearly always the constraint, because that is where an hour is worth the most. Define states the loss in units the plant already recognises: parts per shift, or scrap value per week. Measure puts a data collection sheet at the station and runs it for a fortnight, because shift to shift variation is exactly what the monthly report averages away.

Analyse separates the losses into buckets and tests the favourite theory against them. This is where a manufacturing project usually surprises the people who have worked there longest: the changeover everyone complains about turns out to be a smaller loss than the twelve short stoppages a shift that nobody logs because each one lasts ninety seconds.

Improve is the part manufacturing already knows how to do: stage the tooling, convert internal setup steps to external ones, standardise the way the changeover is run with the people who run it. Control is the part that gets dropped. A run chart at the station, owned by the shift leader, reviewed weekly, is the difference between a gain that holds and a gain that appears in a slide deck once.

Low volume and high mix work is the honest limit. The statistical half of the method wants observations, and a job shop making eleven of a thing has no distribution to look at. The Lean half still applies in full, and the Six Sigma half largely does not until you aggregate across similar jobs, which takes judgement rather than software.

The second limit is the plant where the real constraint is not on the floor at all. If the line is waiting because purchasing cannot get material, the project belongs in purchasing, and every hour spent optimising the line is wasted effort dressed as progress.

The third is programmes bought as training. A plant that certifies forty green belts and frees none of them to finish a project has spent money on certificates. It is the commonest way manufacturing improvement fails, and it has nothing to do with the tools.