Most of your changeover happens while the machine is stopped for no reason

Ask why a batch is the size it is and the answer is almost never an analysis. It is that this is the size the batch has always been, or that smaller batches would mean more changeovers, or that the run has to be worth setting up for.

That last one is the real answer, and it is correct. It just has a number attached that almost nobody has calculated, and the number moves the moment the changeover does.

Why setup forces a batch

A changeover costs time during which nothing is produced. Spread that lost time across the units in the batch and each unit carries a share of it. Bigger batch, smaller share.

---

	Changeover time	90 min
	Run time per unit	2 min
	Batch of 50: setup per unit	1.8 min
	Batch of 200: setup per unit	0.45 min
	Effective time per unit, 50 against 200	3.8 vs 2.45 min

Illustrative figures. At a batch of fifty, nearly half the effective time per unit is the changeover being amortised. At two hundred it is under a fifth. This is the entire argument for a large batch and it is a real argument.

So the pressure to batch is genuine, and anyone dismissing it has not done this arithmetic. Where it goes wrong is that this is only one of the two costs, and the other one is never on the same page.

The cost on the other side

A large batch buys efficiency at the step and pays for it in time everywhere else.

The first unit waits for the last. In a batch of two hundred, the first item finished sits until the two hundredth is done before the batch moves on. At two minutes a unit that is nearly seven hours of an item being finished and going nowhere, which is the arithmetic in a batch size is a changeover time that nobody measured.

The stock is real money. Two hundred units of work in progress against fifty is a hundred and fifty units of material and labor sitting in the process rather than invoiced.

And you find defects two hundred units late. If something is wrong with the setup, a batch of fifty exposes it after fifty. A batch of two hundred exposes it after two hundred, and the other hundred and fifty are already made.

The lever nobody pulls

Given a fixed changeover, there is an economic batch size that balances the two, and arguing about it is where most of these conversations go. That argument accepts the changeover as a fact of nature.

It is not. It is the most improvable number in the whole calculation, and improving it moves the balance rather than trading against it.

Halve the ninety minute changeover to forty five, and a batch of a hundred now carries the same setup per unit that two hundred did before. The efficiency argument for the large batch is gone, and every downstream cost halves with the batch. Nothing was traded. The constraint the argument was built on moved.

And changeover time is usually improvable by a lot, because it has usually never been looked at. Most of it turns out to be preparation that could have happened while the previous run was still going, tools being fetched, settings being rediscovered, and a first-off check that waits for someone. Those are all recoverable without buying anything.

What to do with it

Time a changeover, properly, once. Not the number people quote. Stand and watch one, with a stopwatch, and write down what happens in each interval. The quoted figure is almost always the best case and the real one is usually longer.

Then split what you recorded into two columns: what genuinely requires the machine stopped, and what does not. The second column is nearly always the larger, and everything in it can be moved outside the changeover for the price of some organization.

And recompute the batch afterwards, not before. The economic batch is an output of the changeover time. Fix the changeover first, then the batch size argument answers itself, and it answers smaller.

The batch size is a symptom. The changeover is the thing, and it is almost always the number in this whole conversation that nobody has ever measured.