OEE — Overall Equipment Effectiveness — is one number for how well a line is really running: availability × performance × quality. Availability asks if it ran when it should; performance, if it ran at the right speed; quality, if the parts were good. 100% is perfect, and most lines sit well below where they assume.
What OEE actually measures
OEE rolls three separate questions into a single percentage. Availability asks whether the machine was running during the time it was meant to be. Performance asks whether it ran at its rated speed while it was running. Quality asks how many of the parts it made were good. Multiply the three together and you get OEE.
Because it's a product of three factors, a respectable-looking number can hide a weak one. 90% on each sounds healthy, but 0.9 × 0.9 × 0.9 is 73% — and the factor dragging hardest is where your attention belongs.
Why your number's probably wrong
The most common mistake is measuring against the wrong baseline. If 'planned time' quietly excludes the stoppages you'd rather not count, availability flatters you. If the 'ideal' cycle time is set to what you actually achieve rather than what the machine can really do, performance always looks great.
An OEE that never moves is usually a sign it's measuring the target, not the truth. Honest inputs sting at first — and then they start to improve, which is the whole point.
A worked example
Say a line is scheduled for 8 hours but runs for 7 (availability 87.5%), runs at 90% of its rated speed (performance 90%), and 97% of parts are good (quality 97%). OEE is 0.875 × 0.90 × 0.97 = 76%.
That 'lost' quarter of the day isn't abstract — it's output you paid for and didn't get. Putting a pound figure on it tends to focus minds quickly.
What to do with it
Don't chase the headline number — chase the weakest of the three factors. If availability is dragging, you have a downtime problem. If performance is low, you're running slow or stopping in lots of small bursts. If quality bites, you're making scrap.
OEE points you at which question to ask next; it doesn't answer it for you. That's a feature, not a flaw — it turns a vague sense that 'the line could do better' into a specific place to look.