OEE stands for Overall Equipment Effectiveness. It’s a single percentage that answers one question: of all the good parts a machine could have made in the time it was scheduled to run, how many did it actually make. Everything else, the formula, the three factors, the benchmarks, is detail underneath that one question.
The short version
OEE multiplies three numbers, each its own percentage, each answering a narrower question:
- Availability, was the machine running when it was supposed to be
- Performance, was it running at the speed it should have been
- Quality, were the parts it made actually good
Multiply the three together and you get OEE. A machine that runs 90% of its scheduled time, at 85% of its rated speed, making 98% good parts, has an OEE of 0.90 times 0.85 times 0.98, which is about 75%. Not 90%, even though 90% is the number that probably jumps out first. That’s the part people underestimate: three strong-looking numbers multiplied together produce a noticeably weaker one, because losses in each factor compound instead of average out.
The formula, factor by factor
Availability compares run time against planned production time. Planned production time is the time the machine was actually scheduled to make parts, shift hours minus scheduled breaks and planned maintenance, not the full 24 hours in a day unless you’re actually running around the clock. Run time is planned production time minus every stop, whether that’s a jam, a changeover, waiting on material, or an operator away from the machine.
Availability = Run time / Planned production time
A machine scheduled for an 8 hour shift that’s actually producing for 6.8 of those hours has an availability of 6.8 divided by 8, or 85%.
Performance compares actual output speed against the machine’s ideal speed for the part it’s running. This is the factor people get wrong most often, because it needs a real ideal cycle time, not a padded one. If a part’s real ideal cycle time is 30 seconds and the machine actually averaged 36 seconds a part over the run, performance is 30 divided by 36, or about 83%.
Performance = (Ideal cycle time x Total count) / Run time
Quality compares good parts against total parts made. A run of 500 parts with 15 scrapped or reworked gives you 485 good parts, so quality is 485 divided by 500, or 97%.
Quality = Good count / Total count
A worked example, one machine, one shift
Take a single CNC running an 8 hour shift, 480 minutes scheduled.
Run time is 420 minutes, which is 25,200 seconds. At a 45 second ideal cycle time, the machine could have made 560 parts in that window. It actually made 480. Performance is 480 divided by 560, or about 85.7%. Of those 480, 460 were good, so quality is 95.8%. Availability was 420 out of 480 scheduled minutes, or 87.5%.
Multiply the three: 0.875 x 0.857 x 0.958 is about 0.718, or 71.8% OEE. That’s a solid number for most shops, well above the roughly 60% industry average, and nowhere near the 85% textbook target, which is fine. The OEE benchmarks guide covers what other plants actually run and why 85% isn’t the number to chase out of the gate.
Why the 85% target misleads more shops than it helps
The 85% figure comes from a specific recipe: 90% availability, 95% performance, 99% quality, multiplied together. It was never meant as a universal pass mark, it’s what Toyota-style single-part, high-volume manufacturing lines were measured against decades ago. A high-mix job shop running twelve different parts a week on the same machine has setup time baked into its schedule as a fact of life, not a failure, and that alone can cap availability well under 90% no matter how well the shift runs.
Treating 85% as the bar every machine should clear leads to one of two bad outcomes. Either people loosen the standard, an ideal cycle time gets padded, a stop gets reclassified as planned, until the number reads 85% and means nothing. Or people chase a target that was never realistic for their production mix and burn morale getting nowhere. Neither helps. The useful target is your own real number today, plus meaningful improvement, not a number borrowed from a different kind of factory.
The three questions OEE actually answers
Splitting the score into three factors matters because each one points at a different kind of problem, and the fix for each is different.
A low availability score points at scheduling and setup: too much unplanned downtime, changeovers that run long, machines waiting on material or an operator. The fix usually lives in maintenance, scheduling, or material flow, not the machine’s programming.
A low performance score points at the machine running slower than it should while it’s actually cutting: a program that hasn’t been optimized, a feed rate backed off out of caution and never revisited, wear that’s crept in gradually. The fix usually lives in the process itself.
A low quality score points at the part, not the uptime: tooling wear, a fixture that’s drifted, a process that’s marginal against the tolerance. Scrap and rework cost capacity just as surely as a stopped machine does, and it’s the factor most likely to hide, because a machine making bad parts still looks like it’s running fine on a glance across the floor.
Knowing which factor is dragging the score down tells you who should be in the room when you fix it. A plant that only tracks the single blended OEE number loses that diagnostic power entirely, it can tell you something’s wrong but not what.
Where the numbers actually come from
The accuracy of an OEE number depends entirely on where its inputs come from. Run time and stop time need to come off the machine’s own control or a sensor watching it, not a memory of roughly how the shift went. Ideal cycle time needs to be the part’s real, unpadded standard, not a number loosened until the score looks acceptable. Good and scrap counts need to reflect what actually happened at the machine, instead of only what final inspection caught. Get any of those three wrong and the multiplication still produces a clean-looking percentage, it’s just not describing your floor.
That’s also why a first automated OEE reading often comes in lower than whatever number the shop was already using. It’s not that anything got worse. It’s that a control watching every second catches losses a person glancing at a machine a few times a shift never could. Your first accurate month covers what that drop means and why the lower number is the one worth trusting.
Quick recap
- OEE = Availability x Performance x Quality, three percentages multiplied, not averaged
- Availability = run time / planned production time
- Performance = (ideal cycle time x total count) / run time
- Quality = good count / total count
- A worked shift of 87.5% availability, 85.7% performance, and 95.8% quality lands at 71.8% OEE, not the 90% the first number suggests
- 85% is a specific historical benchmark, not a universal target, especially for high-mix, low-volume work
- The three factors matter because each points at a different fix: scheduling, process, or the part itself
- The number is only as accurate as its inputs: a real ideal cycle time and stop data that comes off the machine, not a memory of the shift
OEE FAQ
What does OEE stand for? Overall Equipment Effectiveness. It measures what share of a machine’s fully productive time was actually spent making good parts at the right speed.
What is the OEE formula? OEE = Availability x Performance x Quality. Availability is run time divided by planned production time. Performance is ideal cycle time times total parts, divided by run time. Quality is good parts divided by total parts.
How do you calculate OEE by hand? Track four numbers for a shift: planned minutes, downtime minutes, total parts made, and good parts made. Availability is (planned minus downtime) divided by planned. Performance needs the part’s ideal cycle time, multiplied by total parts, divided by run time in the same units. Quality is good parts divided by total parts. Multiply all three.
What is a good OEE score? 85% is the world-class benchmark, but most manufacturers run closer to 60%. See the OEE benchmarks guide for the full breakdown by range.
Why is my calculated OEE different from what I expected? Usually because one factor is weaker than it looks in isolation, and the three compound. A machine that seems to run “most of the time” and make “mostly good parts” can still land at 70% OEE once availability, performance, and quality are all multiplied together.