Blog / Downtime
Downtime · Guide

How to Count Changeovers in OEE: The Clear Rules

OL By OEE Lab Editorial |Updated 2 November 2026 |6 min read
Short answer

Changeovers count in OEE as an availability loss, even though they are planned, because the line is scheduled and is not making product. Keep them inside scheduled time, code them apart from breakdowns, and report them as their own loss.

Where a changeover sits

In Overall Equipment Effectiveness (R. C. Hansen, 2001), a product changeover is a planned stop that is part of normal running, and it stays in scheduled time. Nakajima's six big losses call the same loss setup and adjustment.

"Planned" does not mean "excluded". Excluded time is when you did not plan to produce at all, while a changeover is a step on the way to producing.

That rule is simple, but people argue about it every week. So here are the arguments, each with numbers.

The example line

Example: a line has 480 scheduled minutes and an ideal rate of 20 units per minute. It does 3 changeovers of 35 minutes each and has 25 minutes of other stops, so it runs 350 minutes.

It makes 5,880 units and 5,700 are good. Availability is 72.9%, performance is 84.0%, quality is 96.9% and OEE is 59.4%.

The good units are worth 285 minutes at the ideal rate, and 285 ÷ 480 = 59.4%. Changeovers alone take 105 ÷ 480 = 21.9% of the shift.

Argument 1: "They are planned, so exclude them"

Take the 105 minutes out and scheduled time is 375, so OEE reads 285 ÷ 375 = 76.0%. That looks much better, and nothing has changed.

The real damage shows up later. Say a SMED project cuts each changeover from 35 to 20 minutes, so the line gains 45 minutes and makes 6,636 units, about 6,433 of them good.

Counted, before SMED59.4%
Counted, after SMED67.0%
Excluded, before SMED76.0%
Excluded, after SMED76.6%
The same SMED project on the same line, with about 13% more good units either way. Exclude changeovers and the win almost disappears from OEE.

When excluded, OEE moves by 0.6 points after a real gain of about 13% in good units. A metric that cannot see your best improvement project is not doing its job.

Argument 2: "Sales gives us small batches, it is not our fault"

That may be true, and OEE does not assign blame. It measures how much of the scheduled time made good product at full speed.

Show the changeover loss as its own line in the report. If the cause is batch size, that is the evidence you need for a talk with planning, as changeover frequency explains.

The changeover cost calculator puts a money value on that talk.

Argument 3: "Count only the overrun past the standard"

Some plants set a standard of, say, 30 minutes and book only the extra 5 as a loss. That makes the standard invisible, and it can never improve.

Track actual vs standard as a separate measure if you find it useful. In OEE, all 35 minutes are lost production time.

Argument 4: "Where do start-up scrap and slow running go?"

This is the honest hard one. After the line restarts, it may run slow and make scrap for a while, and you have to decide where those minutes go.

Suppose each 35-minute changeover is 25 minutes stopped and 10 minutes of slow restart. You can book it in two ways, and both are defensible.

Whole 35 min as a stop25 min stop, 10 min run
Run time350 min380 min
Availability72.9%79.2%
Performance84.0%77.4%
Quality96.9%96.9%
OEE59.4%59.4%
Same line, same units. The factors shift, and OEE does not. Each uses the 5,880 total units and 5,700 good units.

OEE stays at 59.4% because the good units and the scheduled time did not change. So this argument changes the labels and never the result, and you can see why in time-based vs count-based OEE.

Pick one rule and keep it. Our advice: book the stop while the machine is stopped, and let slow running and scrap fall into performance and quality, because machine signals see it that way.

Then keep the full 35 minutes as your changeover clock for improvement work, from last good unit to first good unit at rate. That method is in how to measure changeover time, and the restart losses are in startup losses.

Argument 5: "Cleaning between products is not a changeover"

If the cleaning is needed to run the next product, it is part of the same loss and stays in scheduled time. Give it its own code if you want to see it apart.

A long cleaning planned at least a week ahead, with no product scheduled, can be excluded time. See sanitation downtime for the detail.

What does not count as a changeover loss

  • A planned stop with no next order. If nothing is scheduled, that time is excluded. See OEE with no orders.
  • Waiting for the new tool or material. Still a stop, but code it as waiting, since the cause and the fix are different.
  • A size change with the line still running. Nothing stopped, so it falls into speed loss or waste.

How to code changeovers so the loss is useful

Split the time into a few phases on the log: clear, swap, adjust, wait. Record how often as well as how long, because 3 changeovers of 35 minutes and 7 of 15 minutes both cost 105 minutes and point to very different fixes.

1
Code them apart from breakdowns
Use one code family, for example CO-clear, CO-swap, CO-adjust, CO-wait.
2
Record count and minutes
105 minutes from 3 changeovers, not just 105 minutes.
3
Compare by product pair
A to B may take 15 minutes while A to C takes 60. Planning can use that.
4
Rank it with the other losses
Put it in the same downtime Pareto, then decide what to attack first.
A short routine that turns one loss number into a list of actions.

The six big losses show how it sits with everything else, and the OEE loss tree gives the code families.

Bookings that distort the number

  • Moving changeovers to excluded time to reach a target.
  • Writing a standard time on the log when the clock said something else.
  • Counting trial units as good because they were later shipped without being tested.
  • Letting each crew pick its own start and end marker.
See it live in Fabrico

A changeover written from memory is usually rounded to the standard. Fabrico takes the stop live from the machine, and operators tag it as a changeover at the line, so OEE per line and shift shows the real minutes. See how Fabrico tracks changeover stops.

What people ask next

Is changeover time planned downtime or unplanned?

It is planned, and it still lowers availability because it sits inside scheduled time. See planned vs unplanned downtime for how the two types are treated.

Are changeovers in availability or performance?

The stop is in availability. If the line restarts slowly, that slow time can fall into performance and the scrap into quality, depending on your booking rule. OEE comes out the same.

Can I show OEE with changeovers excluded?

You can show it as a separate measure, clearly named, to see running quality without changeovers. Do not call it OEE, and do not compare it with other lines' OEE. Use the OEE calculator for the standard number.

How do I compare lines with many changeovers and few changeovers?

Compare the loss split, not only the total OEE. A line with many short runs has more changeover loss, so look at performance and quality apart, and see OEE with mixed products.

How much can changeover time be reduced?

It varies a lot by plant. Shingo (1985) reported setups falling to about a fortieth of their original time, from his own cases. Single-plant studies report cuts of around 40-50% (Brito et al., 2017; Sousa et al., 2018). See SMED changeover reduction.

See OEE per line and shift on your own machines

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.

Book a Fabrico demo

Related guides: Does Planned Maintenance Count in OEE or TEEP? · Tool Change Downtime: Measure It and Cut It

← More OEE Lab guides · How to measure changeover time →

OEE essentials: OEE calculator · What is OEE · Six big losses · Availability · TEEP · Cost of downtime · OEE benchmark · Free OEE tracker · Free OEE software · How to track downtime · How to implement OEE · Light MES vs full MES