OEE done right: measure equipment performance without fooling yourself.

Has your business ever been caught in this spot: the month-end report shows an OEE as high as 85%, and yet actual output still misses the plan? The paradox is so familiar that many factories quietly stop asking why. In truth, OEE is one of the easiest metrics to “dress up”: choose the denominator cleverly and round a couple of numbers, and a line that’s leaking capacity can still produce a lush green report. In this article, we’ll dissect three ways OEE quietly fools its reader, look at how to measure honestly from machine data at the source, and finally how to turn a dry score into concrete work right there on the shop floor.

What OEE measures — and why a single number is easy to mislead with

OEE (Overall Equipment Effectiveness) packs three questions into one ratio: does the machine run when it should (availability), does it hit the design rate while running (performance), and is the product right the first time (quality)? OEE = availability × performance × quality. Because it’s the product of three ratios, a figure like 85% sounds high but can be assembled from three very different components — and that assembly point is exactly where the truth tends to get lost. Don’t rush to chase a benchmark before you’re sure the way you measure is honest.

Three components, three ways to fool yourself

Each component has a “back door” to inflate the number without anyone deliberately cheating:

• Availability: taking “planned stops” — cleaning, mold changeovers, start-of-shift meetings — out of the denominator, so the time the machine is truly available gets overstated.
• Performance: using a “theoretical” cycle rate rounded down to be easy to hit, or ignoring slow runs and micro-stops of under a few minutes.
• Quality: counting only scrap that must be discarded, while forgetting rework, downgraded goods, or the trial run at the start of a batch.

The biggest trap is in the denominator

The question “availability over which time base” determines most of the honesty. If you count only against scheduled run time, all the losses from poor scheduling — missing orders, missing staff, waiting on materials — vanish from view. A filling line might show “82% OEE” for a shift; but counted against the entire time the factory is open, the real number is far lower because the first half of the shift had no orders to run. Both calculations are “correct,” but they answer two different questions — and mixing them is the fastest way to reassure yourself.

Measure from the machine, not from memory

OEE recorded by hand at the end of a shift is almost always optimistic, because people remember long stops and forget the short, repeated ones. When data comes straight from the machine in real time — every stop, every cycle, every defect time-stamped — the “small stops” that have long been invisible finally show up. Often it’s exactly those 30–90 second petty stops, accumulated across a shift, that leak the most capacity — not the big incident everyone sees and everyone records.

The six big losses — a framework to trace causes

A low OEE number is only useful when you can split it into actionable loss types. The classic “six big losses” framework helps do that:

• Breakdowns and unplanned machine stoppages.
• Changeover and setup time between batches.
• Idling and micro-stops.
• Running below design rate (reduced speed).
• Scrap and rework during stable production.
• Scrap and downgrades at startup and batch trial runs.

Turn the metric into action

OEE is only worth measuring when every lost percentage point traces back to which machine, which shift, and what reason. That requires OEE not to live alone in a report, but to sit on the same data layer as maintenance and quality: a repeated string of micro-stops on the same station leads straight to a preventive-maintenance ticket; a cluster of defects traces back to the exact incoming material batch. When MES records the machine cycle and QMS records defects on the same data foundation, OEE stops being an end-of-month score and becomes an input to work that needs doing within the shift.

Measuring OEE honestly — a quick checklist

• Agree on one definition of the denominator, and state clearly which time base it counts against.
• Take the design cycle rate from the manufacturer’s specs, not from an “easy to hit” average.
• Count every stop, including those under a minute, and tag a reason code to each one.
• Include rework and downgrades in quality losses, not just discarded goods.
• Compare OEE across shifts and across identical machines; don’t just look at one aggregate number.
• Prioritize the trend over time over an absolute score to show off.

The goal of OEE isn’t a pretty number to post on the board, but a map that points to exactly where capacity is leaking. A 70% OEE you trust and can trace is worth far more than a 90% OEE no one dares to look at closely. Start by agreeing on the definition and bringing data in from the machine — the improvements will reveal their own path.

Related articles