Key takeaways
  • SMED separates work that must happen while the machine is stopped (internal) from work that can happen while it runs (external).
  • Most first-pass improvements come from moving work to external and from preparation, not from equipment.
  • Long changeovers cause large batches, which cause inventory and slow response. Changeover time is a lead-time problem disguised as a setup problem.
Why changeover time matters more than it looks
Every minute of setup is an argument for a bigger batch

When a changeover takes four hours, nobody runs small batches, so inventory grows and lead time lengthens. Halving setup time does not just recover capacity, it removes the reason for batching.

The four steps

  • Observe and record. Film a real changeover. Not a demonstration, a normal one, ideally on a difficult shift.
  • Separate internal from external. List every element and mark whether it genuinely requires the machine stopped.
  • Convert internal to external. Pre-stage tooling, pre-heat, pre-set, pre-position. This is where most early gain lives.
  • Streamline what remains. Quick-release fasteners, standard shut heights, eliminate adjustment through fixed stops.

What the video always shows

Two things, reliably. A large share of the stopped time is spent looking for something, and adjustment after the first piece takes longer than the physical change. Both are solvable without capital: a shadow board and a kit trolley address the first, fixed stops and recorded settings address the second.

Adjustment is the hidden cost

Trial-and-error adjustment after a change is frequently the single largest element. It is also the most reducible: record the correct settings per product, use positive stops rather than measurement, and the first piece should be right rather than approximately right.

Sustaining the gain

Changeover time reverts quietly. The countermeasure is a documented changeover procedure with the standard time on it, a kit prepared to a checklist before the machine stops, and the actual time recorded each run. When the recorded time drifts upward, the procedure has stopped being followed and you will know within a week rather than a year.