The Efficiency Number Your Floor Cannot Reproduce

The plant that ran at 85 percent and could not make rate

A Midwest protein plant had a number it was proud of. The shop-floor productivity dashboard read 84.9 percent efficiency. Leadership repeated it. The schedule was built on it. And when the optimization team asked to see the raw machine data behind it, the plant manager went quiet and stopped sharing the export.

That is the tell. A real efficiency number survives daylight. A manager who reports 84.9 percent and then withholds the underlying data is not protecting a trade secret, he is protecting a configuration. We escalated to ownership to get direct system access, because the only way to settle a disputed efficiency number is to measure the line yourself. The working estimate of actual performance was closer to 50 percent. The plant was not running at 85. It was reporting at 85.

Why the reported number drifts high, every time

OEE is availability times performance times quality. Each of those three is a fraction, and every fraction has a denominator someone gets to define. Exclude planned downtime from availability and the number jumps. Set the performance baseline to a soft demonstrated rate instead of nameplate and it jumps again. Let operators key in the reason codes by hand at end of shift and the micro-stops quietly vanish into running.

None of this requires anyone to lie. It requires a dashboard nobody validated and a definition nobody owns. The number stops measuring the floor and starts measuring the configuration of the tool. The dangerous part is what it does downstream. Capacity decisions are made against the reported number, not the real one. If the dashboard says 85 and the floor runs 50, every plan built on top inherits a 35-point error. The third shift gets approved. The capital request to buy a line gets approved. The constraint that is actually costing you the shift keeps getting funded as already solved, because on paper it is.

A plant that believes it is near world-class has no reason to look for the bottleneck. That belief is the most expensive thing on the floor. We have walked into facilities convinced they were already efficient and found the real constraint sitting in a place no one had instrumented, because the dashboard had pointed everyone's attention somewhere else for years.

Measure the line before you trust the number

You do not argue a reported OEE number down in a conference room. You replace it with measured data. The fastest version of this is mechanical, not analytical. Drop lightweight wireless trackers on the production lines; the install is low impact and takes about a day. Pair that with a 3D scan of the facility so the line exists as a model you can run. Now you are watching where output actually leaks, not where the dashboard says it leaks.

Then you simulate. On one frozen protein line we modeled, a run of roughly 300 simulated scenarios narrowed to five viable operating changes, and inside that set was a pathway worth about 14 percent more throughput and $1.4M in labor. That number did not come from a sharper opinion. It came from measured input the reported dashboard had been hiding. The same hard data that exposes the gap also disarms the argument, because a plant that thinks it is already efficient cannot wave away a tracked shift and a simulated line the way it can wave away a consultant's hunch.

Do this before you spend. The most expensive way to buy throughput is to add a shift against a number that was wrong by 35 points. The cheapest is a day of trackers and a stopwatch on the constraint.

What a validated floor looks like

The reported OEE sits within about 5 points of an independently measured shift, and when it drifts the gap gets investigated instead of repeated. One named person owns the measurement definition, so availability is not quietly re-cut every quarter. Reason codes are captured at the machine, not typed in from memory at end of shift. And no capacity dollar, no new shift, no new line, moves until the constraint has been located by tracked data rather than by the dashboard. On that floor the efficiency number is a tool. On the other one it is a story.

The number was never the cost

The plant was not lying about 85 percent; the dashboard was configured to produce it. The cost was never the number itself. It was every decision made against it, the shift that should not have been added and the constraint that stayed invisible, because nobody put the reported number and the measured one in the same room.