How to Calculate and Improve OEE (Formula + Example)

The single number is useful for tracking. The three components are what you actually fix. (For the one-line definition, see the glossary entry on OEE.)

The OEE formula

OEE breaks production into three loss buckets, each scored 0–100%:

Availability = Run Time ÷ Planned Production Time Captures stop-time losses — breakdowns, changeovers, and unplanned downtime. (Planned Production Time is your scheduled time minus planned breaks.)

Performance = (Ideal Cycle Time × Total Count) ÷ Run Time Captures speed losses — slow cycles, minor stops, and running below the equipment's ideal rate.

Quality = Good Count ÷ Total Count Captures quality losses — scrap, rework, and any part that doesn't pass the first time (your FPY).

OEE = Availability × Performance × Quality

Because the three multiply, a weak link hits the whole score hard. Three respectable-looking 90s still land you at 73%.

What's a good OEE?

The benchmark most cited is 85%, the level Seiichi Nakajima defined as world-class in the foundations of Total Productive Maintenance. Hitting it means roughly 90%+ Availability, 95%+ Performance, and 99.9%+ Quality — and in practice only about 3% of manufacturers sustain it (Evocon; Tractian, 2025–2026). The typical plant runs closer to 55–65%.

One caution before you benchmark: if you track OEE by hand, your number is probably flattering you. Manually logged OEE tends to run 8–12 percentage points higher than reality (Symestic, 2026), because short stops and small speed losses never make it into the spreadsheet. A line "reporting" 78% on paper is often in the mid-60s once measured automatically. And benchmarks vary by industry — a high-volume automotive line and a high-mix aerospace shop shouldn't be held to the same target. Compare against your own past, and against similar operations, not a generic ideal.

How to improve each component

You don't improve "OEE." You improve one of its three parts — so find your weakest and start there.

Availability (stop-time losses): cut unplanned downtime with predictive maintenance, shrink changeover time with structured setup reduction (SMED), and hunt down the recurring minor stops that quietly add up.

Performance (speed losses): find where the line runs below ideal cycle time, eliminate the micro-stoppages too brief to log by hand, and address the causes of slow running — tooling, material feed, operator wait.

Quality (defect losses): reduce scrap and rework by catching defects early — in-line checks or computer-vision inspection — and attack startup scrap after changeovers, a common hidden source.

A realistic improvement pace is two to four OEE points per quarter, sustained. Chasing 85% in a year from a 60% baseline usually just frustrates the team.

The catch: you can't improve what you can't see accurately

Every lever above depends on one thing: knowing your real losses, as they happen. That's why OEE and accurate data are inseparable. A number assembled by hand at the end of a shift hides the micro-stops and speed losses that are often the biggest opportunity — and points you at the wrong problem. OEE measured automatically on connected, real-time data shows the line as it actually is, the moment it happens, so a supervisor can act on the current shift instead of reviewing a flattering summary of the last one.

That accuracy is a product of the connected data foundation underneath — and surfacing it live is the job of manufacturing analytics. If your OEE is hand-tracked today, fixing the measurement is usually the highest-ROI first move, because it's the prerequisite for fixing everything else.