The upper limit is (shift minutes - minutes needed to make the demand) / changeover minutes. Below that limit, the best number balances changeover cost against the stock that bigger batches create. In the example below, a 450-minute shift can take 5 changeovers but runs cheapest at 3.
There is a pull in both directions. Planners want short runs so stock stays low and customers get what they ordered, while the floor wants long runs because each change eats the shift.
Both sides are right, and neither can win an argument without numbers. So here are the numbers, in two questions: how many changeovers fit, and how many pay.
Question 1: how many fit?
Example: a line runs 5 products on a 450-minute shift and must average 30,000 units a shift. Its real net rate is 100 units a minute, and each changeover takes 30 minutes.
Use the real rate, including stops and speed loss, and not the ideal rate. With the ideal rate you would overstate what fits, then miss the plan.
Count the whole changeover too, from last good unit to first good unit at rate. How to measure changeover time has the method, and the tuning at the end is the part most often missed (see adjustment losses).
| Changeovers per shift | Changeover min | Run min | Capacity | Spare units | Batch size |
|---|---|---|---|---|---|
| 1 | 30 | 420 | 42,000 | 12,000 | 30,000 |
| 2 | 60 | 390 | 39,000 | 9,000 | 15,000 |
| 3 | 90 | 360 | 36,000 | 6,000 | 10,000 |
| 4 | 120 | 330 | 33,000 | 3,000 | 7,500 |
| 5 | 150 | 300 | 30,000 | 0 | 6,000 |
| 6 | 180 | 270 | 27,000 | -3,000 | 5,000 |
Five is a ceiling, not a plan. With zero spare units, any breakdown turns directly into a late order, so leave yourself some room.
The production capacity calculator gives you the same sum for your own line.
Question 2: how many pay?
Inside the limit, two costs pull against each other. Changeover cost rises with each change, and stock cost falls as batches shrink.
Example: each change costs €100 in scrap, labor and materials, not counting lost output, since at 3 changeovers the line has spare capacity. Holding one unit in stock costs €0.012 a shift.
Each of the 5 products holds about half a batch on average. So average stock = 5 x batch / 2, which is the usual simple estimate.
The low point is where the two costs match, at 3 changeovers and batches of 10,000. There is also a formula that lands in the same place.
Notice how flat the curve is near the bottom. Going from 3 to 4 costs just €25 a shift, so service to customers can decide between them.
When lost output belongs in the cost
The €100 above leaves out lost output, and that is only right while the line has spare capacity. A freed minute on a machine that already has time to spare earns nothing.
On a constraint, every changeover minute is a minute of lost sales. Add the margin on those units to the cost per change, and the best number falls.
See bottleneck analysis for how to tell. The changeover cost calculator works out the full cost of a change.
What a shorter changeover changes
Example: a team cuts the changeover from 30 to 15 minutes, and the cost per change falls to €50. Now run both questions again.
| 30-min change, €100 | 15-min change, €50 | |
|---|---|---|
| Most that fit | 5 | 150 / 15 = 10 |
| Best number | 3 | 4 |
| Batch size | 10,000 | 7,500 |
| Changeover time per shift | 90 min | 60 min |
| Cost per shift | €600 | 200 + 225 = €425 |
That is the real point of SMED: it changes the tradeoff instead of picking a side of it.
Is a cut that size realistic? Shingo reported setups falling to about a fortieth of their original time, and that was the method's author on his own cases.
Published single-plant studies report cuts of around 40 to 50% (Brito et al., 2017; Sousa et al., 2018).
The OEE trap
A product changeover stays inside scheduled time, so more changeovers lower availability. How to count changeovers in OEE explains why they are not excluded.
Under the 30-minute change, 1 changeover takes 6.7% of the shift, and 5 take 33.3%. So the best case for OEE falls from 93.3% to 66.7%, with no other losses at all.
A team judged only on OEE will push for long runs, and stock will build up where nobody on the line can see it. Judge the schedule by total cost and delivery, then use OEE on mixed products to read the results fairly.
Practical rules for your schedule
- Group by family. A size change inside one family is often far shorter than a full change. Run similar products in sequence.
- Use your real spread. Plan with your typical changeover, then check it against your worst ones.
- Set a run length floor. A run shorter than its own startup settling time makes little good product.
- Vary by product. A fast seller may run every shift, and a slow seller once a week. Five products need not change equally.
- Check demand against takt time. Your batch plan has to meet the pace of orders.
Where the sums go wrong
- Counting only stopped time. The real change is longer, so the limit is lower than you think.
- Using the ideal rate for capacity. The line never runs there, so the plan fails by the second day.
- Guessing the holding cost. Ask finance for it. Cash, space and scrap risk all belong in it, and the result moves with this number.
- Freezing the answer. Rerun the sums when demand, product count or changeover time moves.
These sums need real changeover times and real net rates. Fabrico takes stops, speed and counts live from your machines, and operators tag the cause at the line, so real stop minutes and rates come from the line, with OEE per line and shift. See how Fabrico tracks changeovers and stops.
What people ask next
How many changeovers per shift is normal?
We found no neutral benchmark, and it depends on the product mix, the changeover time and the demand. Work out your own limit and best number with the two sums above. Your line's history is the better reference.
What is the formula for optimal batch size?
The classic form is the economic batch: the square root of (2 x demand x cost per change / holding cost per unit). In the example, per product and per shift, that is the square root of (2 x 6,000 x 100 / 0.012), which is 10,000. Your plant may also have minimum runs and shelf-life limits.
Do more changeovers always lower OEE?
They lower availability if the changeover stays in scheduled time, which is the standard method. A plant can still choose more changes, since lower stock may be worth it. See how to count changeovers in OEE.
Should we cut changeover time or cut the number of changeovers?
Cut the time first if you can, because that improves both sides of the tradeoff. Cutting the number only moves cost into stock. SMED is the method to start with.
What if we run out of capacity at the changeovers we need?
Look for shorter changes, family sequencing and less loss in the run itself. If the numbers still fail, you need more hours, or other lines to share the work. How to increase production capacity lists the options in order.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.
Related guides: OEE and Seasonal Demand: Reading the Swings · Cost Per Unit: Formula and How OEE Changes It · Net Operating Time: Definition vs Run Time