Ghost Capacity Lives Between the Oven and the Box

The capacity that isn't there

A packaged-meats processor handed us thirty days of line data and asked what their sausage line could really do. We asked for a full year instead. Thirty days shows you an average; a year shows you the variation, and the variation is where the money hides. We built a digital twin of the plant and ran it against their own numbers. On most lines the model matched the floor inside 2 percent, 98.2 percent accuracy against actual production. One line came back at 88 to 89 percent, and when we flagged it the operator remembered the cause immediately: six or seven weeks of intermittent trouble that had quietly poisoned the data months earlier. The line had been telling a story about its capacity that wasn't true, and nobody had reconciled it.

That is ghost capacity. It is the throughput your forming or cook line claims it can hit, on a board everyone trusts, that your packaging line can never actually absorb.

Why the constraint hides downstream

Meat and protein plants score themselves on the loud end of the line. The oven, the former, the cook step, those are where the OEE dashboards point, because that is where the expensive equipment runs. But shippable cases come off pack-off, and pack-off has a different rhythm than the cook line feeding it. When the upstream step overproduces, the surplus does not become output. It becomes work in process sitting in front of a packaging operation that absorbs product at its own variable rate.

Watch what the plant does to cope. At one two-site protein processor, one plant ran double-wall boxes justified by humidity, and thicker oven film to stop product slipping on the racks. Pack-off was inconsistent enough that pallet builds varied, weight distribution wandered, and the answer to wandering pallets was a stronger box. The packaging spec had quietly become a shock absorber for upstream variability. Double-wall corrugate runs on the order of 20 to 40 percent more than a right-sized single-wall box, and because the spec drifted per machine and per site, the SKU count fragmented and stripped whatever pricing leverage a single standardized box would have bought. They were paying twice: once for the overbuilt material, once in lost scale.

The tell is always the same. When packaging is overengineered, the root cause is rarely packaging. It is upstream variability that someone, years ago, solved with a thicker box instead of a tighter process. Strict packaging test minima and higher edge-crush specs feel like rigor, but they can mask the real cause and lock plant flexibility in place. The honest move is performance specs that set a floor while preserving supplier room, not a single hard number that everyone overshoots to be safe.

How to find your ghost

You find ghost capacity by modeling variance, not by walking the floor and trusting the gauges. A precooked-sausage co-manufacturer we piloted with put it in the right order: phase one, replicate current OEE and variability in a twin until it matches; phase two, run the high-volume experiments you would never risk on a live line; phase three and four, wire in live data and let the model prescribe line adjustments. None of it works without the unglamorous inputs, the CAD layout, the line-monitoring extracts, bagger fill rates, bags per minute, case and pallet timestamps. Get those, validate them, and the variance band reveals itself.

Then run the test this week. Pull a year of throughput by line, not a month. Lay the cook-line output rate next to the pack-off rate and look at the spread, not the mean. If the cook line's reported capacity sits well above what pack-off can clear on a bad week, the surplus is ghost. Next, audit your packaging spec against your pallet consistency. If you are buying double-wall to survive uneven pallets, you have found upstream variability wearing a packaging costume, and the cheaper fix lives at pack-off, not at the box plant. Last, count your box and film SKUs; every avoidable variant is leverage you handed back to your supplier.

What a well-run line looks like

The digital twin matches every line inside 2 percent of the floor, so capital arguments run on modeled throughput instead of the loudest opinion in the room. Packaging is specified to performance minima that the pallet actually requires, not to the worst pallet the line has ever produced. Pack-off variability sits inside a single-digit band, which means the cook line is paced to what packaging can clear rather than run hot to feed a WIP pile. Box and film SKUs are consolidated far enough that the plant carries real pricing leverage, and a year of clean line data, not thirty days, is what every throughput claim traces back to.

The number worth believing

The cook line will always volunteer a bigger number than the plant can ship; that number is the easy one to celebrate and the expensive one to believe. Put the cook rate and the pack-off rate in the same model, on a full year of data, and the ghost stops being free.