Executive summary: OEE multiplies availability by performance by quality to express what share of scheduled production time an asset spent making good parts at full rate. The multiplication is the point and the trap: three respectable-looking factors compound into a mediocre score, which is why a line everyone considers well run frequently measures in the low sixties. This guide covers the formula, a worked calculation, what a good score actually is, and why OEE only means anything at the constraint.
- What is OEE?
- How do you calculate the three factors?
- What does a full calculation look like?
- Why do three good factors produce a bad score?
- What is a good OEE score?
- Why does OEE only matter at the constraint?
- How do you keep the measurement honest?
- What are the most common OEE mistakes?
- OEE: operator FAQ
- About the Stagnation Assassin
What is OEE?
Overall Equipment Effectiveness is availability multiplied by performance multiplied by quality. It expresses the percentage of scheduled production time an asset spent producing good parts at its full rated speed. A score of 100 percent would mean producing only good parts, at full rate, with no stops, throughout all scheduled time.
The measure came out of total productive maintenance practice in Japanese manufacturing and has become the standard way to express equipment effectiveness in a single number. Its value is that it captures three genuinely different kinds of loss in one figure, and its danger is exactly the same thing: a single number hides which of the three is hurting you.
What makes OEE useful is that it counts everything against the same denominator. A machine can be available all shift, running at full speed, and producing scrap. Another can produce perfect parts slowly. A third can run perfectly for the four hours it was actually working. Traditional metrics flatter all three. OEE does not, because a loss anywhere in the chain drags the whole result down.
I have led transformations at Berkshire Hathaway, Illinois Tool Works, and Whirlpool, and OEE is one of the few metrics I consistently ask for early, with one strict condition attached: I want it for the constraint, calculated honestly, and I am not particularly interested in it anywhere else. That condition does most of the work, and I will come back to why.
How do you calculate the three factors?
Availability is run time divided by planned production time. Performance is actual output divided by what the asset should have produced at rated speed during that run time. Quality is good units divided by total units produced. Each is a ratio between zero and one, and OEE is their product.
Availability
Take planned production time, which is scheduled time minus planned downtime such as breaks and scheduled maintenance. Subtract all unplanned stops, including breakdowns and setup time, to get run time. Availability is run time divided by planned production time. Changeovers count as availability loss, which surprises people and is correct, because during a changeover the machine is not producing.
Performance
Take the total count produced during run time and compare it to what the ideal cycle time says should have been produced in that same run time. Performance is actual divided by theoretical. This factor captures speed losses and micro stoppages, which is why it is usually the hardest of the three to measure and the most commonly fudged.
Quality
Good units divided by total units produced. Anything requiring rework counts as a quality loss even if it is eventually sold, because the capacity consumed producing it and reworking it is gone. Counting reworked parts as good is the single most common way plants inflate this factor.
One structural note. Planned downtime sits outside the calculation by convention, which means OEE measures effectiveness during time you intended to produce. That convention is reasonable and it creates a loophole: a plant can improve OEE by reclassifying unplanned downtime as planned. If your OEE improved sharply without any physical change, check whether the definitions moved.
What does a full calculation look like?
Take a line with 72 percent availability, 90 percent performance, and 96 percent quality. Multiplied together, OEE is 62 percent. That means 38 percent of scheduled production capacity is being lost, sitting inside an operation most plants would describe as functioning normally, because each individual factor sounds acceptable.
Build it from the floor up.
Availability. The line is scheduled for 480 minutes. Planned downtime, meaning breaks and scheduled maintenance, takes 30 minutes, leaving 450 minutes of planned production time. Unplanned stops, breakdowns plus changeovers, consume 126 minutes. Run time is therefore 324 minutes. Availability is 324 divided by 450, which is 72 percent.
Performance. Ideal cycle time is 30 seconds per unit, so 324 minutes of run time should yield 648 units. Actual output was 583 units. Performance is 583 divided by 648, which is 90 percent. That missing 10 percent is speed loss and micro stoppages, most of it too brief to appear in any downtime log.
Quality. Of the 583 units produced, 560 passed first time and 23 required rework or were scrapped. Quality is 560 divided by 583, which is 96 percent.
OEE. Multiply: 0.72 times 0.90 times 0.96 gives 0.62, or 62 percent. In plain terms, of 450 minutes of intended production, the line delivered the equivalent of about 280 minutes of perfect production.
Why do three good factors produce a bad score?
Because OEE multiplies rather than averages. Three factors of 90 percent each produce an OEE of 72.9 percent, not 90 percent. Losses compound, so a modest shortfall in each of the three dimensions creates a total loss far larger than any individual number suggests, which is why OEE consistently surprises management.
This is the arithmetic that changes how people think about their plant, so it is worth sitting with. Run the sequence:
- 95, 95, 95 percent compounds to 85.7 percent OEE, which is genuinely excellent performance
- 90, 90, 90 percent compounds to 72.9 percent, which sounds strong and is ordinary
- 85, 85, 85 percent compounds to 61.4 percent, meaning nearly 40 percent of capacity is gone
- 80, 80, 80 percent compounds to 51.2 percent, meaning the asset is delivering barely half of what it was scheduled to
Nobody looks at an 85 percent availability figure and panics. Nobody looks at 85 percent quality and calls an emergency meeting, though they probably should. Yet three separate 85s mean the plant is throwing away close to four tenths of the capacity it already owns and pays for.
This compounding is also good news, and it is the part people miss. Because the factors multiply, small gains multiply too. Moving each factor by five points, from 85 to 90 across all three, takes OEE from 61.4 to 72.9 percent. That is an 11.5 point gain in OEE, which translates to roughly 19 percent more output from the identical asset, achieved through improvements that individually look minor.
Three factors at 85 percent each produce an OEE of 61.4 percent, not 85 percent, because OEE multiplies. Improving each factor by just five points lifts OEE to 72.9 percent, which is roughly 19 percent more output from the same machine. Compounding cuts both ways, which is why modest gains on all three beat a heroic effort on one.
What is a good OEE score?
Roughly 85 percent is the widely cited world-class benchmark for discrete manufacturing, built from about 90 percent availability, 95 percent performance, and 99 percent quality. Typical plants measure around 60 percent, and scores near 40 percent are common in operations that have never measured rigorously. Context matters more than the benchmark.
Those reference points are useful and frequently misused, so treat them carefully.
The 85 percent world-class figure comes from total productive maintenance practice and assumes discrete, repetitive manufacturing. It is a demanding target and a legitimate one in that context. In a high-mix operation with frequent changeovers, an 85 percent OEE may be structurally unreachable without changing the product mix itself, and chasing it can drive genuinely bad decisions such as running longer batches to protect the metric.
The more important benchmark is your own trend. An OEE of 62 percent improving steadily toward 75 is a far better story than a static 78 percent, because the trend tells you whether the organization is learning. I would rather run a plant with a rising mediocre score than a flat good one.
The genuinely useful comparison is between your OEE and your own theoretical ceiling given your product mix and changeover requirements. That number is specific to your operation and it is what tells you how much room is actually left, as opposed to how you compare to a benchmark drawn from a plant that runs three products where you run three hundred.
Why does OEE only matter at the constraint?
Because improving OEE on a machine with spare capacity produces a better number and no additional output. The system produces what the constraint produces. Raising a non-constraint from 65 to 85 percent OEE means it now has more idle time, not that the plant ships more, which is why plant-wide OEE averages mislead so reliably.
This is the condition I attach whenever I ask for OEE, and it eliminates most of the metric’s failure modes at a stroke.
Consider what a plant-wide average OEE actually tells you. It blends a constraint at 62 percent with a dozen non-constraints at 80 percent and reports something around 78 percent, which is reassuring and meaningless. The one number that governs your output is buried inside an average designed to obscure it. I have watched plants report improving average OEE for four consecutive quarters while shipping the same volume, because every gain came from machines that already had capacity to spare.
There is a subtler trap too. A non-constraint with high OEE may be actively harmful, because high OEE at a non-constraint often means it is producing more than the constraint can absorb. That shows up as work in process, consumed working capital, and hidden quality problems. In that situation the good OEE number is evidence of a problem, not of performance.
So the discipline is: measure OEE rigorously at the constraint, use it to target where capacity is being lost, and treat non-constraint OEE as diagnostic information only, never as a performance target. The moment non-constraint OEE becomes something people are rewarded for, you have built an incentive to overproduce.
How do you keep the measurement honest?
Automate data capture where possible, define planned downtime once and freeze the definition, count reworked units as quality losses, and use the true ideal cycle time rather than a negotiated standard. Every one of these is a place where OEE gets quietly inflated, usually without anyone intending to cheat.
Four specific defenses.
Freeze the planned downtime definition. Since planned downtime sits outside the calculation, reclassifying unplanned stops as planned improves OEE without changing anything physical. Write the definition down, date it, and require sign-off to change it. A sudden OEE improvement with no corresponding physical change is almost always a definitional shift.
Use the true ideal cycle time. Performance compares actual output against theoretical output at rated speed. If the rated speed used in the calculation is a conservative negotiated standard rather than the machine’s genuine capability, performance will look excellent while real speed losses stay invisible. Use the demonstrated best cycle time, not the routing standard.
Count rework as a quality loss. Parts that get reworked and eventually shipped feel like good parts and are not. They consumed capacity twice. Counting them as good is the most common inflation in the quality factor and it hides exactly the losses that are stealing constraint time.
Automate capture where you can. Manual logging systematically misses short stops, because nobody records a ninety second interruption. Those micro stoppages live in the performance factor and are frequently the largest single hidden loss. Automated counting is what makes performance measurable rather than estimated.
What are the most common OEE mistakes?
Four recur: measuring plant-wide averages instead of the constraint, comparing OEE between dissimilar assets, treating OEE as a performance target for operators, and improving whichever factor is easiest rather than whichever is costing the most constraint capacity.
Mistake 1: the plant-wide average
Averaging OEE across dissimilar assets produces a number that cannot be acted on and that hides the constraint inside it. Report the constraint’s OEE as the headline figure. Everything else is supporting detail, and rolling it into one average destroys the only signal that mattered.
Mistake 2: comparing machines to each other
OEE is a comparison of an asset against its own potential, not against other assets. A machine with frequent changeovers by product design will always score lower than a dedicated single-product line, and that gap says nothing about how well either is run. Compare each asset to its own trend and its own ceiling.
Mistake 3: making it an operator scorecard
The moment OEE becomes a number operators are judged on, it stops being a diagnostic and starts being managed. Stops get logged as planned, rework gets counted as good, and the metric drifts upward while output stays flat. Use OEE to find losses, not to rate people, and say so explicitly when you introduce it.
Mistake 4: attacking the easiest factor
Teams improve whichever factor they know how to improve rather than the one costing the most. If availability is 72 percent and quality is 96 percent, quality work is the wrong project no matter how skilled your quality team is. Break the constraint’s losses down first, then route the work to the factor with the largest recoverable loss.
My own mistake, early on, was accepting reported OEE at face value. A plant told me its constraint ran at 84 percent OEE. Two days of observation put the real figure closer to 60, once micro stoppages nobody logged and reworked parts counted as good were both included. Nobody had lied to me. The measurement system had simply been built with comfortable definitions and never challenged. Measure it yourself once before you trust the number, and the second measurement will be reliable because everyone now knows what the definitions mean.
A packaging operation running 62 percent OEE was facing a $15M request for two new lines with an 18 month lead time. Fixing changeovers, micro stoppages, and first-pass quality delivered the entire required output increase for about $600,000 in six months. The presses were never bought, and the capacity had been sitting inside the existing lines the whole time.
OEE: operator FAQ
What is the OEE formula?
OEE equals availability multiplied by performance multiplied by quality. Availability is run time divided by planned production time, performance is actual output divided by theoretical output at rated speed, and quality is good units divided by total units produced. For example, 0.72 times 0.90 times 0.96 gives an OEE of 62 percent.
What is a good OEE score?
Around 85 percent is the widely cited world-class benchmark for discrete manufacturing, built from roughly 90 percent availability, 95 percent performance, and 99 percent quality. Typical plants run near 60 percent. In high-mix operations with frequent changeovers, 85 percent may be structurally unreachable, so your own trend matters more than the benchmark.
Why is my OEE so low when all three factors look fine?
Because OEE multiplies rather than averages, so losses compound. Three factors at 90 percent each produce 72.9 percent, and three at 85 percent produce 61.4 percent. That compounding works in reverse too: improving each factor by five points can add roughly 19 percent more output from the same asset.
Should you measure OEE on every machine?
Measure it rigorously at the constraint and treat everything else as diagnostic only. Improving OEE on a machine with spare capacity creates more idle time rather than more output, and plant-wide OEE averages hide the constraint inside them. High OEE at a non-constraint can even signal harmful overproduction.
About the Stagnation Assassin
Todd Hagopian is a Fortune 500 transformation executive who has generated $3B+ in shareholder value across Berkshire Hathaway, Illinois Tool Works, Whirlpool, and JBT Marel, where he serves as VP of Global Product Strategy. Known as The Stagnation Assassin, he is the author of two published books: The Unfair Advantage: Weaponizing the Hypomanic Toolbox and Stagnation Assassin: The Anti-Consultant Manifesto. His blog is published in 15+ languages and read by operators worldwide. Bring him to your stage via the speaking page or connect with him on LinkedIn.
Next step: measure the one that counts
Your plant-wide OEE average is hiding the only number that governs your output. Book a 20 minute session and I will help you calculate your constraint’s OEE with honest definitions, then break its losses into the three factors so you know exactly which one to attack first. Start here.