Capital Approves What It Can See. The Constraint Lives Downstream.

The asset you can finance isn't the one that governs throughput

A Tier-1 protein processor was about to spend $1.4M on four auto-loaders. The pitch wrote itself: faster line feed, fewer touches, a hard-dollar labor and OEE story the CFO could underwrite. Five-year amortization. Clean math.

We built a digital twin of the facility from a year of actual production data. The model hit 98.2 percent accuracy against measured plant performance, then it gave back a number nobody wanted to hear. The cook system caps the facility at 18 million pounds per year, roughly three times current volume. Buy the auto-loaders, push the front end as fast as the line will run, and the system still stops at the cook ceiling. Capital approved on the visible variable produces savings on paper and queues on the floor.

Why the wrong variable keeps winning the capital fight

Capital committees buy what they can see. Line speed is visible. A faster conveyor is photographable. A four-person headcount reduction shows up in the next labor variance report. So the asset that gets funded is almost always the one closest to whatever bottleneck the floor manager talks about loudest, not the constraint that actually governs throughput.

The interaction problem is harder to defend on a one-page IRR memo. It requires the sponsor to argue that the meaningful number isn't the speed of any single station. It's how station 3 behaves when station 5 takes a 12-minute changeover, how scheduling stacks two short runs back-to-back, how a downstream cook batch waits 40 minutes for a wash-down. Nobody buys a wash-down with capital. So nobody models it. And throughput keeps living below the rated capacity of every single asset on the line.

Two patterns repeat across the work we see.

The first is the expansion that gets framed as "simple equipment relocation." We did a feasibility review on a Midwest meat processor's move into a sister facility to produce a contract brand. The initial scope: move the equipment, paint the floor, run by June. The actual scope, once anyone drew the system: $6.3M of new electrical (600-amp switch gear), process and refrigeration piping, three new storage tanks, structural modifications, 9 to 12 months minimum. The June deadline was already dead the day the request landed. Nobody had drawn the system. They had drawn the asset.

The second is the line-speed bet on top of a broken measurement system. A processor was running 7 weeks of OEE data that, on inspection, was structurally impossible (it showed greater than 100 percent, a data error). The "we're already efficient" conviction sat on top of a measurement system that couldn't prove it. The interaction problems weren't hidden, they were unmeasured. Several Tier-1 processors we've started with this year were convinced they were near world-class going in. The data, once captured properly, almost always relocates the bottleneck somewhere the leadership team wasn't looking.

Run the constraint pass before you write the capital ask

Three questions every capital justification should answer before it leaves the plant floor.

Where does the next binding constraint sit after you spend? If you approve the proposed dollars, what's the new ceiling in physical units, and where in the system does it live? For the protein processor: 18M lbs/yr at cook. That number reframes the entire conversation. The auto-loaders don't get killed. They get phased. Phase 1 runs a 1 to 2 line trial with the new equipment, operational workshops, and staffing optimization, and proves the savings model on a piece of the floor you can recover from. Phase 2 cuts older lines or shifts to two-shift operation based on what Phase 1 showed. Phase 3 only scales toward growth scenarios when demand actually crosses thresholds that justify cook investment. The capital still flows. The risk gets metered.

Does your OEE data survive a sanity check? If your numbers exceed 100 percent, if your operators disagree with the dashboard, if maintenance and production report different uptime for the same shift, you don't have a measurement system. You have a number-generator. Spending capital on top of a broken measurement system is how plants end up "world-class" on paper and 70 percent utilized in reality. Fix the measurement before you fix the asset. Wireless trackers and 3D scans now make this an afternoon of work on a single line, not a six-month study across the plant.

Is cash tied to operational verification or to delivery? A contract packager moving on a 23.57 percent ROI conveyor automation project for year-end install wanted equipment ordered, delivered, and commissioned by mid-December to hit capital approval timing. The right pricing structure held the last 25 to 33 percent of vendor payment until the line hit its promised throughput in production, not the day the truck unloaded. Equipment that's installed isn't equipment that's working. Tie cash to the outcome the capital was approved against.

What this looks like in a deal

When the next capital ask lands on the desk, the first 30 minutes shouldn't be on IRR. They should be on the system map. Where's the constraint today. Where does it move to after you spend. What does the OEE data actually say when somebody puts a tracker on every line for a week. How does cash stage against operational proof rather than against the calendar.

If those answers aren't clean, the IRR doesn't matter. You're going to fund the asset and import a different bottleneck six weeks after commissioning.

The savings were real on paper. So were the queues at the cook. Nobody put the two in the same model.