The Line That Reported 140% Capacity

A Line That Reported 140%

A multi-plant protein processor pulled up its real-time OEE system to settle a capacity argument. One line read 140%. Zero unplanned downtime. Planned downtime, 1.26%. On paper, the line was a machine that never stopped and never stumbled.

The team standing on that floor had watched it stop for hours that same week.

A line cannot run above 100% of itself. When the reported number says it can, the number is not measuring the line; it is measuring whatever the plant decided to feed the system. In this case the internal read was that true OEE sat somewhere around 50 to 60%, and the gap was either a misconfigured sensor map or a plant manager engineering the data to keep corporate off the floor. Either way, the dashboard had stopped being an instrument and started being a story.

That gap has a name. It is ghost capacity, and it is the most expensive number in a plant because it is the one nobody can see.

The Hours You Already Spent

Here is the mechanism, and it is one sentence: rework consumes the same line capacity as first-pass production, and the schedule almost never counts it.

When a pallet gets reworked, the line runs it again. When a quality hold gets dispositioned three days late, the rework run lands on top of next week's first-pass schedule. When a sample has to be re-validated, the validation run eats real line hours. None of that work appears as a separate demand on the plan. The planning model sees a line with open hours and books first-pass volume into them, while the floor is already spending those hours running the same product a second time.

So the line looks loaded because it is loaded. It is just loaded with work the schedule treats as free. That is why the capex case for a second line always looks airtight. The math is clean: demand exceeds rated capacity, add capacity. The math is also wrong, because rated capacity was never the real number. The line was running at 50 to 60% of itself and reporting 140%, and the difference was getting burned on rework the model never charged for.

This is why ghost capacity is so hard to kill. The one system designed to reveal it, OEE, is the exact place it disappears. Inflate the number and the loss vanishes. Average the number across formats and the loss smears into noise. Misconfigure the downtime categories and hours of stoppage become 1.26% planned. You cannot find ghost capacity in the reported numbers, because the reporting is where it goes to hide. You find it by mapping the floor.

What It Looks Like When You Map It

A prepared-foods co-packer running a box-conversion project shows the rework loop in the open, without any data games at all.

A pallet of first-round test product sat on a quality hold. A QA tech scanned the pallet ID, saw the hold flag was not real, and shipped it to a customer. The samples were gone. The original test pallets for that same item had already been lost. So the item went back into the queue for a third pass: new pallets, a different facility, someone driving them around to run transit tests again. One item, one validation, three runs against the line and the lab.

Then there is the box that will not size on a bench. Product comes off an auto-bagger with air in the bag. The air settles over time, so a box sized to today's settled inventory is too small to pack as product comes off the line. The only honest validation is run-and-pack, on the line, at speed. Which means the resize cannot be checked with a sample on a table; it has to consume a real production slot. The team's instinct was right: move the run up so the resize can be validated sooner, because the validation is not free, it is a line event.

None of this is failure. The plant culture here was healthy, willing to flag a fumble and iterate. But every one of those loops is a first-pass run the schedule never booked. Multiply it across a top-ten conversion list expanding into die cuts, and you have a steady tax on capacity that no plan accounts for and no OEE number will show you.

Stop Spending the Line Twice

Three moves, and you can run all three this week.

First, distrust any OEE number you did not instrument yourself. Above 100% is not a high score; it is a broken measurement. Insist on raw production data, historical and recent, before you accept a single capacity figure. Lack of transparent data is not a reporting inconvenience; it is the cover that lets ghost capacity survive. If the plant cannot show you the raw downtime log, the OEE number is a narrative, not a measurement.

Second, charge rework to the line. Build one column in your capacity model for re-runs, re-tests, and held-pallet dispositions, and book those hours against the same line as first-pass output. You are not adding work; you are finally counting work that was already happening. The week that column fills up is the week your second-line capex case quietly falls apart, because the capacity was there the whole time.

Third, validate the way the line actually runs. The co-packer's air-in-bag problem is the general rule: if a spec only holds at benchtop and fails run-and-pack, you have not validated it, you have deferred the rework to production. Pull the validation onto the line early, on purpose, so the rework loop happens once on your schedule instead of three times on the floor's.

The protein processor did not need a second line. It needed to stop spending the first one twice and to stop letting the OEE system tell it otherwise. The capacity was never missing. It was already booked, against work nobody was counting.