A Recipe Change Is a Capacity Decision in Disguise

A Recipe Change Showed Up as a Capacity Cut

A protein co-packer that runs a national quick-serve chain's steak program got the word: the product was moving to a pre-marinated process. New tumbling step, an adobo step, heavier packaging. Capital for the transition was already approved. The piece nobody had pinned down was the throughput cost.

There was a number, and it was not small. Based on the plant's own experience converting a prior carne asada item, the pre-marinated change was expected to cut line run-rate by 12 to 15 percent. The line speed on the spec sheet did not move. The recipe did, and the recipe quietly took 12 to 15 points of capacity with it. On a line already running close to 5,000 pounds across an 8 to 8.5 hour day, that is 600 to 750 pounds of daily output gone, and the question of whether the plant could still meet seasonal demand at constant volume was sitting open while the capital request was already closed.

The Ceiling Lives in the Transition Matrix

This is the trap of the nameplate rate. A spec sheet gives you one number per format and invites you to multiply it by hours. Reality is not one number; it is a transition matrix. Every format carries its own setup, its own film and label change, its own added steps. Add a SKU and you do not add one row to the schedule, you add a new column to every changeover that SKU now sits next to. The changeover graph grows superlinearly with SKU count. The pre-marinated steak did not slow the cutters; it added steps and weight that the line was never balanced for, and the line rate fell out the back.

The corollary holds in the other direction too. At a beverage division that had been flat for a decade, one line carried the bulk of the run and was heavily automated. The tail lines, the infrequent, unpredictable SKUs, never built enough consistent volume to justify automation investment, so they stayed manual and stayed expensive per unit. Same mechanism, read backward: the high-volume core can absorb its changeover load, the tail cannot, and the tail is where the schedule actually leaks. Treating every SKU as equal load is the error in both cases.

The way out is to stop arguing about averages and model the real line. When a digital twin was built around a sausage producer's plant, fed a full year of production, labor, downtime, and throughput data rather than a clean 30-day slice, the model tracked the facility to 98.2 percent of baseline. The value was not the average rate. It was the variation. With a model that accurate, the operator could finally ask the only question that matters at the ceiling: if demand holds, if you grow, if you double in size, where does this facility actually break? One line in that same plant only modeled to 88 percent, and the reason was instructive: six or seven weeks of bad OEE data. You cannot find a ceiling on dirty numbers. The model is only as honest as the floor data underneath it.

Model the Format Change Before You Sign the Capital

The fix is sequencing, not software. Three moves put the throughput ceiling and the format change in the same model, where they belong.

First, build the transition matrix, not the rate table. List the run-rate of each format on its own, then list the cost of every change between them: minutes lost, scrap on the transition, the steps a new process adds. The pre-marinated steak was a new process bolted onto an existing line; its 12 to 15 percent was knowable in advance because a prior conversion had already shown the shape of the loss. That number existed before the capital decision. It just was not in the capacity model.

Second, price each SKU on its changeover load, not its average throughput. The minor SKU that runs twice a month looks cheap on a per-unit average and is brutal on the schedule, because it drags its full setup behind a fraction of the volume. The tail SKUs at the beverage division were not failing on line speed; they were failing on the economics of their own changeovers. Knowing which SKUs carry their load and which ones tax the line is the difference between adding capacity and adding cost.

Third, run the format change as a capacity scenario before the capital closes, not after the trial. This is exactly what the digital twin was built to do: try the change virtually, at year-of-data accuracy, and watch where the line bends. The pre-marinated transition had an approved budget, a Q4-into-2027 timeline, and an unmodeled 12 to 15 percent line-rate hit. The model could have run before the check cleared. The order was backward.

The Lens Flip

The capital for the steak transition was approved as a recipe change. It was a capacity decision. The 12 to 15 percent was not a surprise; it was a known number from a known precedent, and it never made it into the capacity model. The ceiling does not announce itself on the spec sheet. It hides in the changeover, and the only way to see it is to put the format change and the throughput model in the same picture before you commit the spend.