Every factory I have ever walked into tracks OEE. And in about eight out of ten cases, the number on the board is wrong. Not slightly wrong. Fundamentally, systematically, comforting-wrong. The kind of wrong that makes everyone feel good until a customer audit or a margin compression event reveals the gap between the number and reality.

I have audited OEE calculations in automotive plants in Germany, aerospace suppliers in the UK, electronics assembly in Slovakia, and family-run machine shops in Poland. The mistakes are always the same. Here they are.

The availability lie

Availability is supposed to measure the time your equipment is actually running versus the time it could be running. Simple enough. Except nobody agrees on what "could be running" means.

In one aerospace plant, the scheduled production time was 120 hours per week per machine. But the maintenance window — 12 hours — was excluded from the denominator. So the machines could never achieve less than 100% availability during production time by definition. Their OEE was artificially inflated by ten percent before anyone even measured anything.

In another plant, changeover time was classified as "planned downtime" and excluded. Forty-minute die changes, six times a day, across eight presses. That is three hours and twelve minutes of lost production time per press per day that simply vanished from the calculation. Their OEE said 87%. Their real OEE was closer to 72%.

If your availability definition excludes changeovers, maintenance, meetings, warm-ups, and breaks, you are not measuring availability. You are measuring how good your accounting team is at reclassifying lost time.

The performance fiction

Performance measures actual cycle time against ideal cycle time. The problem is that nobody's "ideal cycle time" is actually ideal. In most plants, the ideal cycle time is the nameplate speed from the equipment manufacturer — a number that was achieved once, in laboratory conditions, by a test engineer, with a perfect part, on a brand-new machine, before it had worn at all.

In reality, your machine will never run at nameplate speed on serial production. The materials vary, the tooling wears, the operator is not a robot, and the machine has accumulated twenty thousand hours of thermal and mechanical stress since it was new. If your performance metric uses the nameplate speed as the ideal, your performance number is fiction.

I have seen plants where the real sustainable cycle time was 30% slower than the nameplate. They were running at "98% performance" while actually running 30% below the rate they thought they were achieving. When we recalibrated the ideal to a demonstrated sustainable cycle time — validated over a two-week study — their performance dropped to 72%. That is the number they should have been managing to all along.

The quality illusion

Quality rate is the most manipulated OEE component because it is the easiest to manipulate. If you define quality as "parts that passed final inspection," you are measuring inspection effectiveness, not process quality. Reworked parts count as good parts. Parts that were caught and fixed in-process count as good parts. Parts that were downgraded and sold at a lower margin count as good parts.

In one automotive electronics plant, the first-pass yield was 89%. But after rework — an extensive, labour-intensive rework process that consumed 15% of production capacity — the quality rate was reported as 99.2%. The OEE calculation used the 99.2% figure. The plant manager was proud of it. But the rework labour, the rework materials, the rework energy, the rework floor space, and the rework management attention were all real costs that the quality rate had hidden.

Real quality rate is first-pass yield. Period. If you want to include reworked parts in your quality metric, you need a separate metric for that — rework rate — and it should be visible, costed, and targeted for reduction. Hiding rework inside quality is the most common OEE sin I see.

The denominator game

The single biggest OEE manipulation is the denominator — planned production time. I have seen plants that exclude lunch breaks, shift handovers, tier meetings, 5S time, TPM time, training time, cleaning time, and "warm-up time" from planned production time. One plant excluded the first thirty minutes of every shift as "stabilisation time." Another excluded every Monday morning as "ramp-up." By the time they were done excluding, the denominator was so small that achieving 90% OEE was mathematically guaranteed.

Here is the standard I use: planned production time is every minute the plant is open and staffed. If people are on the clock and the equipment is available, it counts. The only exclusions are scheduled non-production days, plant closures, and capital projects that physically prevent production. Everything else is planned production time. If that makes your OEE look bad, good. It should. That is the starting point for improvement.

The spreadsheet theatre

Most OEE tracking I have seen is not real-time. It is a weekly exercise where a production engineer sits at a desk, opens a spreadsheet, and tries to reconstruct what happened on the shop floor last week from shift reports, maintenance logs, and handwritten notes. By the time the number is calculated, the conditions that produced it have changed. The operator who ran the machine has moved to a different line. The tooling has been changed. The material lot is gone.

OEE is a diagnostic metric, not a report. If you are not looking at it today, for yesterday's performance, you are doing historical accounting, not process control. The value of OEE is in the trend and the reaction to the trend, not in the absolute number. A plant that tracks OEE in real time, reacts to deviations within the shift, and uses the data to drive improvement will outperform a plant with a higher OEE number that is calculated weekly and reviewed monthly.

What I recommend

Stop trusting your OEE number until you have audited every assumption behind it. Walk the floor. Time the changeovers with a stopwatch. Check the ideal cycle time against a demonstrated sustainable rate. Pull the first-pass yield data, not the final inspection data. Recalculate the denominator from first principles.

When you are done, your OEE will almost certainly be lower. That is the point. The real number is the one you can act on. The inflated number is the one that makes you feel good while your competitors eat your lunch. I would rather have an honest 65% OEE that everyone is working to improve than a fictional 85% that everyone is working to defend.

OEE is not a score. It is a tool. And like any tool, it is only useful if it tells you the truth.