Find the First Line Before You Buy the Second

The conversation that almost cost a second oven

A Chicago baking manufacturer pulled me into a call last fall to scope a capacity expansion. He runs about 7.5M pounds a year on a single oven line, and he is in late-stage conversation with a West Coast prospect that would add another 4 to 6M pounds annually. His instinct, reasonable on its face, was to build a parallel oven and stand up a multi-level conveyor system to feed it. He had already commissioned a 3D scan of the facility. The scan was finished. The capital case was being drafted.

Then he said the sentence that always tells me what is really going on: we measure in pounds per hour, not OEE. He had pulled the metric off a benchmarking exercise with peers and was rolling it out at the next monthly meeting. He had been with the company long enough to know the people on his floor; he had not been with the data long enough to know which numbers were lying to him. He told me, directly, that he has 20-year veterans on the line whose paperwork dates do not match the dates the pallets actually shipped, whose cash conversion cycle drifts a week off because the invoices follow eight days behind the trucks. That data integrity gap is not a back-office problem. It is the same gap that is hiding capacity inside his existing oven, and it is the reason a parallel line would be the wrong first move.

Why the throughput you priced is not the throughput you own

Ghost capacity is the gap between the throughput your spec sheet promises and the throughput your plant actually produces over the course of a real week. It is the difference between rated and realized. Every operations leader has been told this; very few have measured it on their own line. The reason is mechanical. Throughput is not governed by any single rated component. Throughput is governed by how rated components interact under live conditions: changeover sequence, micro-stops, supplier carton tolerance, labor allocation, the way the upstream layout forces a downstream operator to walk an extra eight feet per cycle. The math at the equipment level says you have 60 units per minute. The math at the system level says you get 38. Pounds per hour, measured weekly and averaged across an idealized shift, will not surface that gap. It will reassure you that you are at capacity.

I saw the interaction effect in its most literal form a few weeks ago on a refrigerated-foods client. They had transitioned to a new corrugated supplier, and the new boxes were jamming at a rate of one per hundred. The supplier's first recommendation was to enlarge the master case. That is a six-figure annualized spec change across the SKU set, plus the design and validation tail. I pushed back hard. The defect rate of one in a hundred is exactly the kind of number that points at a system interaction, not a design flaw. It could be carton-tolerance variance at line speed. It could be a sequence issue inside the case packer. It could be an orientation glitch that only surfaces above a certain throughput threshold. None of those are solved by making the box bigger; some of them get worse. The right next step is a larger-quantity production trial that reproduces the defect under load, then a diagnostic on what specifically the line is doing when the jam happens. The wrong next step is the spec change the supplier proposed, because the supplier is solving for their own ghost, not yours.

The labor savings are a capacity story in disguise

The most expensive form of ghost capacity is the version that shows up as a labor line item. I am closing a proposal right now for a frozen-pie manufacturer that runs 130 million units a year and is scaling toward 140. Our team modeled $35.4M in projected savings over the engagement. The largest single line was 48 full-time employees in palletizing and 16 more in the front-end crust automation. On a finance deck, those are labor numbers. On the floor, they are capacity numbers wearing a labor costume. Forty-eight people in palletizing are not slow because they are slow. They are slow because the upstream sequence, the product flow, the pallet layout, and the case configuration force a manual choreography that automation can compress by more than half. Once the choreography compresses, the line that was rated for X actually produces X. The labor budget falls because the ghost capacity the labor was absorbing comes back into the asset.

This is the second mechanism that gets missed. Operations leaders look at a high-labor station and conclude they have a labor problem. CFOs look at the same station and budget overtime. Neither sees that the station is the visible cost of an invisible throughput loss upstream. The 10 percent pallet waste in the same facility is in the same category: on the spreadsheet it is a yield number, but it is functionally a downtime number, because the pallets that get rejected are the pallets the line had to slow down or stop to remake. None of this shows up in pounds per hour, because pounds per hour averages over the rework.

Three moves before the capital ask

If you are inside a quarter where someone is asking for a second line, a major retrofit, or an automation spend, three diagnostic moves come before the capital case. Run them in order; do not skip the first one because the executives have already decided.

First, instrument the existing asset for one full production cycle. Wireless sensors on the line, real OEE math, segregating planned downtime from unplanned, changeover loss from quality loss, micro-stops from blocked-and-starved. One full cycle is enough to see whether the rated-versus-realized gap is 5 percent or 25 percent. The Chicago bakery in the lead story is about to do exactly this, before the second oven line gets specified, and the answer to whether the West Coast prospect needs new asset or recovered asset will land in the first 30 days of data. We have repeatedly seen double-digit percentages of nameplate capacity sitting inside an existing line on comparable builds; that recovered capacity funds a lot of West Coast.

Second, when a supplier hands you a structural recommendation in response to a defect, ask them what system interaction they have ruled out. A one-percent jam rate is a behavior, not a design. Make them prove the defect is in the box geometry before you let them ship you a different box. Eight times out of ten the trial under load tells you the line is operating outside the carton's tested window, and the cheaper fix is operational, not structural. The other two times you have the data to negotiate the spec change with the actual root cause in hand, instead of with the supplier's first guess.

Third, when you cost out a labor program, walk the labor-heavy station and trace the sequence that created the labor demand. If the answer is upstream layout, the labor savings are a capacity story and you should book the value against both lines on the P&L. If the answer is genuinely manual operation that automation cannot reach, the labor savings are a cost story and you should value them lower. Most operations leaders book the cost-story value on everything. That leaves money on the table at proposal time and makes the second-line decision look more urgent than it is.

The line you already own

The second oven, the bigger master case, the 48 fewer palletizers: each is the visible shape of a buried question. The buried question is what your existing system would produce if you measured it correctly and ran it inside the window it was designed for. Find that number first. It is almost always larger than the operating team believes and almost always smaller than the spec sheet promises, and the gap between the two is the budget for the next year.