Someone tells you their line ran at 90% availability, 90% performance and 90% quality last month. That sounds like a 90% month. It was a 73% month.
This catches people out constantly, and it is not a rounding quirk. It is the whole reason OEE exists as a metric.
Overall Equipment Effectiveness is defined as:
Three numbers, multiplied. Not averaged. So:
Your instinct says 90, because your brain averages three nineties. The formula does something else entirely. Each factor is a filter, and the output has to survive all three. Of the time you planned to produce, you only actually ran 90% of it. Of that running time, you only produced at 90% of the rate you should have. Of those units, only 90% were good. Each stage keeps nine tenths of what the previous stage handed it, so you end up with about 73% of what you planned.
The gap between the 90% people assume and the 73% they get is 17 points. On a line that was supposed to make 1,000 units, that is 271 units you never got, and nobody in the room feels like anything went badly wrong.
Here is a single shift, with the kind of numbers you would actually log. An 8 hour shift gives 480 planned minutes. You lose 60 minutes to a breakdown and a changeover, so run time is 420 minutes. Your ideal cycle time is 1 minute per unit, you made 400 units, and 384 of them passed inspection.
Every individual number there looks respectable. Nobody would flag an 87.5% or a 96%. The line still gave back a fifth of its capacity.
Plenty of teams report uptime and treat it as a proxy for how the line is doing. Uptime is availability, and availability is one of three terms.
In the shift above, uptime is 87.5% while true OEE is 80%. In the three nineties example, uptime is 90% while OEE is 73%. In both cases the number being reported upward is the most flattering of the three, and the losses hiding in performance and quality never reach anyone who could act on them.
If you only measure whether the machine was switched on, you will never see the units it quietly made too slowly, or the ones it made twice because the first attempt was scrap.
The standard breakdown is six categories, two under each factor. Availability loses time to breakdowns and to setup and adjustments. Performance loses it to minor stops and idling, and to running below rated speed. Quality loses it to process defects and to startup rejects before the process stabilises.
Minor stops are the ones that get missed most often. A jam that takes 40 seconds to clear does not get logged as downtime by anyone, but thirty of them across a shift is 20 minutes gone, and it shows up in your performance figure rather than your availability figure. That is usually the first place worth looking when performance is the weak term and nobody can explain why.
The figure usually quoted for world class OEE is 85%, and the classic split behind it is 90% availability, 95% performance and 99.9% quality. Multiply those and you get 85.4%.
Look at that quality number again. 99.9%, not 99%. The benchmark assumes you are essentially not producing scrap.
Another way to see how demanding 85% is: if all three of your factors sat at 95%, a level most lines would be pleased with, you would land at 85.7%. You need roughly 95% across the board just to clear the bar. Typical manufacturing OEE sits closer to 60%, which tells you how much room is normally hiding in a line that everyone describes as running fine.
Because the three factors multiply, a given proportional improvement anywhere moves the final number by the same proportion. There is no mathematically special factor.
The reason to start with your lowest score is more practical than that: it is where the most room is, and it is nearly always easier to find 5 points on a 75% than on a 96%. Your weakest term is both the biggest opportunity and the cheapest one.
So measure all three, honestly, for a couple of weeks before you change anything. Log planned time, downtime, ideal cycle time, total count and good count per shift. Those five inputs give you everything. Then look at which of the three factors is dragging, and point your effort there rather than at whichever problem is loudest.
I built an OEE Dashboard for exactly this, because I got tired of rebuilding the same spreadsheet every time. You log those five numbers per shift and it calculates Availability, Performance and Quality, multiplies them properly, charts the trend, and shows your gap to the 85% benchmark. Every formula is verified against worked examples with known answers.
Get the OEE Dashboard for $19But honestly, the formula matters more than the tool. If you take one thing from this: stop averaging your three numbers, and stop reporting uptime as though it were OEE. Multiply them, and look at what you actually get.