Most plants have no idea what a changeover costs them. They know the machine was stopped, roughly how long for, and that it happens often. That is precisely why SMED works: it turns a vague irritation into a measured number, and the number is almost always worse than expected.
What SMED actually claims
SMED stands for Single-Minute Exchange of Die — a changeover completed in under ten minutes. Shigeo Shingo formalised it at Toyota in the 1950s, on presses that took hours to switch. The ambition is not the point; the method is. And the method rests on one distinction that most people miss on first reading.
Internal and external: the distinction that carries everything
An internal operation can only happen while the machine is stopped: removing the tool, fitting the new one, adjusting it. An external operation could happen while the machine is still running: fetching the tool, preheating it, staging the fasteners, preparing the first-off gauge.
In an unimproved changeover, a large share of the elapsed time is spent on operations that had no reason to be internal. The operator stops the press, then walks to the tool store. That walk is external work being done at internal cost — and it is usually the cheapest minute you will ever recover.
The four steps, in order
1. Measure without judging. Film one changeover end to end. Not a representative one — the next one. The video removes the argument about what really happens, and it is the only way to see the waiting, the searching and the second trip to the store.
2. Separate internal from external. List every operation and mark each one. This step alone, changing nothing else, typically recovers a third of the time: everything that was external but done with the machine stopped simply moves before or after.
3. Convert internal into external. Preheat the mould off-line. Pre-set the tool on a setting fixture. Standardise clamp heights so no shimming is needed. This is where the engineering effort goes, and where the remaining large gains are.
4. Streamline what is left. Quick clamps instead of bolts, positive stops instead of measurement, one-turn fasteners. Useful — but attempted first, before separation, it produces expensive fixtures that save little.
Where SMED pays, and where it does not
SMED pays when changeover time constrains your batch size. If you run large batches only because switching is painful, every minute recovered buys flexibility: smaller batches, less stock, shorter lead times. That is the real return, and it is far larger than the recovered machine hours alone.
It pays much less on a machine that is not a bottleneck. Recovering forty minutes on equipment that already idles half the day changes nothing but the report. Start with the constraint.
What an ERP contributes — and what it does not
No software performs a changeover. What it does is make the effort targetable and durable, which is where most SMED programmes fail: the workshop improves for three months, then drifts back because nobody is measuring any more.
Three things need to be recorded. First, the changeover time per machine-tool pair — an aggregate average hides exactly the pairs worth attacking. Second, the split between internal and external, so an improvement can be attributed. Third, the trend over months, because a changeover time that creeps back up is the normal outcome without a scoreboard.
In AVIA ERP, shop-floor clock-ins distinguish changeover from production, so the time is captured as a by-product of normal work rather than through a separate measurement campaign. The figures feed the OEE calculation, and the tool counters trigger preventive maintenance by cycle count rather than by date — a mould that breaks mid-programme costs far more than any changeover.
Where to start on Monday
Pick the machine that constrains your output. Film the next changeover. Split the operations into internal and external, and move the external ones out. You will have your first result before you have bought anything — and a number to argue with, which is the point.