OEE answers a simple question: of all the time you planned to make parts on this machine, how much actually produced good ones?
Think of a taxi on a ten-hour shift. Some of those hours it sits at the rank waiting, or in the garage with a flat tyre. That’s availability. When it does drive, it often crawls through traffic at a fraction of the speed the road allows. That’s performance. And now and then it takes a passenger to the wrong address and has to drive them again, for free. That’s quality.
The driver can feel busy all day and still end up with half a shift of paid fares. OEE is the meter that shows that gap, and, more usefully, splits it into three piles you can attack separately: stops, slow running and bad parts.
That split is where the value is. A 68% on its own tells you very little. Knowing that most of the missing 32% is changeovers tells you exactly where to send your best engineer next week.
How to calculate it
The formula, step by step
Here is the shift from the example above. The order matters: each factor uses the time left over by the previous one, so a minute is never counted as lost twice.
| Step | Calculation | Result |
|---|---|---|
| Planned production time | 480 min shift − 30 min scheduled breaks | 450 min |
| Run time | 450 − 58 min of stops (breakdowns, changeovers, waiting for material) | 392 min |
| Availability | 392 ÷ 450 | 87.1% |
| Performance | 268 parts × 1.2 min ideal cycle ÷ 392 min | 82.0% |
| Quality | 257 good ÷ 268 produced | 95.9% |
| OEE | 87.1% × 82.0% × 95.9%, or 257 × 1.2 ÷ 450 | 68.5% |
Notice the shortcut in the last row. OEE is just good parts × ideal cycle time ÷ planned production time. The three factors don’t change the result, they explain it. If someone shows you an OEE without the breakdown, ask for it.
Try it with your own shift
Everything stays in your browser.
Where the data comes from
Five inputs, three systems, at least three owners
The formula is the easy part. The hard part is that its five inputs usually live in different places, are owned by different people and are updated at different speeds. Every gap between them ends up in your OEE.
| Input | Where it usually lives | What goes wrong |
|---|---|---|
| Planned production time | Shift calendar, ERP work-centre calendar | The calendar says two shifts, the floor runs one and a half. |
| Stops and their reasons | MES, machine PLC signals, operator terminal, paper log | Short stops never get recorded by hand. “Other” becomes the most common reason. |
| Ideal cycle time | ERP routing, set by engineering | Set once at launch, never updated after the process improved, or copied from the machine brochure. |
| Parts produced | Machine counter, MES, work order confirmations | Confirmed at the end of the shift from memory, or at the end of the order in one lump. |
| Good parts | Quality inspection, scrap declarations | Scrap is recorded in the MES and never reaches the ERP, or the other way around. |
Sloth lives in the second row. Operators don’t log a three-minute jam when they’re trying to catch up, so those minutes quietly move from availability into performance, and you go looking for a speed problem that is really a reliability one. If “other” or “not specified” is one of your top three stop reasons, your Pareto chart is fiction.
Greed lives in the third and fifth rows. Ideal cycle times belong to engineering, scrap belongs to quality, the counter belongs to production. Unless someone decides which system holds the truth for each input, every department will have its own OEE, and the weekly meeting will be spent arguing about whose is right.
How it goes wrong
Four sins OEE invites
OEE is not a bad KPI. It’s a very good one that’s very easy to misuse, because it’s a single tidy percentage that feels like it sums up a machine. These are the four mistakes I see most often.
03Lust · falling in love with one number
Raising OEE on a machine that isn't the bottleneck
Two machines in sequence. Machine A runs at 45% OEE, machine B at 90%. The obvious move is a project to improve A. But A can make 100 parts an hour and B, downstream, only 50. A is at 45% because it’s waiting for B. Push A harder and you get a pile of work in progress between the two, more cash tied up, and not one extra part shipped.
Read OEE only next to throughput at the bottleneck. On a machine that isn't the constraint, a low OEE is often exactly what should happen.
07Pride · built to be admired, not used
A definition that keeps the number green
There are two quiet ways to make OEE look better without making anything better. The first is to call changeovers, cleaning and waiting for material “planned” stops: every minute you label as planned leaves the denominator, and availability jumps. The second is a slow ideal cycle time. Set it 20% slower than the machine can really run and performance looks great forever.
A useful alarm: if performance ever comes out above 100%, the ideal cycle time is wrong. The machine didn’t beat physics.
Write the definition down, publish it next to the number, and let the operators check the cycle times. They know the real ones better than anyone.
A 68% you can explain is worth more than an 85% nobody can check.
06Envy · measuring against the wrong benchmark
Chasing the "world-class 85%"
Almost every article on OEE quotes 85% as world class.1 That figure comes from high-volume, repetitive production with long runs of the same part. If you make small batches of many different parts, you have many changeovers by design, and your availability will never look like a car-parts line. Setting 85% as your target is buying a Tour de France bike for your Sunday ride.
Benchmark the same machine against its own history first, and against similar lines with a similar mix second. The trend is the number that matters.
05Wrath · deciding fast on data you haven't understood
Reacting to a red day
On a line with a varied mix, daily OEE swings a lot. One long changeover or one missing batch of material can knock fifteen points off a shift. If every red day triggers a phone call, people learn to avoid the number, or to start bending its definition. That’s how Wrath turns into Pride.
Look at the weekly trend and ask which of the three factors moved before asking who is responsible.
Gluttony Should you track it?
Not on every machine
Track OEE where a lost hour is a lost hour for the whole plant: the bottleneck, and expensive automated equipment that’s supposed to run most of the time. There it pays for the effort of collecting good data.
Don’t put it on every machine because the MES makes it easy. On a manual workstation or a machine with plenty of spare capacity, OEE mostly measures how often it waits, which you already knew. One more red tile on the dashboard makes the important one harder to find.
The eighth sin Who should own it
One owner for the number, another for the rules
The production manager should own the bottleneck’s OEE and what gets done about it. The definition (what counts as a planned stop, which cycle time is “ideal”) should be owned by someone whose bonus doesn’t depend on it, such as continuous improvement or the controller.
And the people who feed it have to get something back. If the operator who logs a stop reason never sees that reason come up in a meeting, the reasons will drift to “other” within a month. A terminal at the machine with a short list of reasons, and a weekly look at the top three losses with the people who recorded them, does more for your OEE than any dashboard.
Ask yourself
Before you decide your OEE is fine
- Which machine is your bottleneck, and is it the one whose OEE you talk about most?
- Who set the ideal cycle times in your ERP, and when were they last checked on the floor?
- What share of your stops is coded as “planned” or “other”?
- If OEE went up ten points next month, what would you expect to see in throughput and on-time delivery? Would anyone check?
Footnotes
-
The 85% figure is usually traced to Seiichi Nakajima’s work on Total Productive Maintenance (TPM) in the 1980s: roughly 90% availability × 95% performance × 99.9% quality. ↩