OEE multiplies three ratios: availability (was the equipment running), performance (was it running at rate), and quality (was the output good first time). The result is the share of planned production time that produced sellable output at the intended speed.
They multiply rather than average, which is why the figure looks severe — 90% on each gives 73% overall. That severity is the useful part. Three tolerable losses compound into an intolerable one, and the arithmetic shows it in a way three separate percentages never do.
The single number is close to meaningless on its own. What earns its keep is the split: availability-dominated loss points at maintenance and changeovers, performance loss at minor stoppages and slow running, quality loss at process capability. Three different projects, and the ratio tells you which one you have.
It degrades badly when used as a target. Planned downtime gets inflated, equipment gets run past what the process will hold, rework gets counted as good — all of which raise the number and none of which improve anything. It is a diagnostic. Treated as a KPI it becomes fiction, and comparing it between machines or sites was never valid in the first place.
