Your Capital Case Is Built on the Wrong Hour

The Capital Case That Quoted the Wrong Number

A Tier-1 meat processor walked into a labor-line optimization project for one of their plants quoting a $30-plus per-hour labor rate. The plant returned its actual data: $26.02. Leadership had been running the case math, the headcount math, and the "what if we automated this" math on a number that was 15% high.

The downtime report came back too. It said 1%. Same description in every row. The partner walking through the data with me called it: "this feels like they just write the same thing every time." A capital case had been moving forward on labor numbers that were inflated and downtime numbers that were fictional. The data did not support the decision the case wanted to make. The decision was already in motion.

This is the failure mode the rest of this is about. Average-hour math is the math most plants use to justify capital. The marginal hour, where overtime kicks in, where a thin shift backfills with a temp, where a missing operator stalls the whole line, costs 1.5 to 2x the average. The capital decision flips when you model the marginal hour. Most operators never get there.

Why the Marginal Hour Is the One That Matters

Sit with a plant's data for a few days and the labor distribution tells you something the line manager will not. A three-line packaging operation we modeled recently looked fine on paper. Ten people on shift, lines averaging 5 to 6 parts per minute against a proven peak of 20. The headcount looked staffed. The output looked acceptable. The plant manager would tell you the line was running.

Underneath, every line was eating 1.5 hours of daily overtime to make demand. Line 1 started at 8:00 AM when it should have started at 7:00. Line 2 started at 7:00 when it should have started at 6:30. Thirty to sixty minutes a day, every day, came out of the back of the shift, where the hour costs 1.5x.

When we modeled the worker utilization properly, current state was 10 people at roughly 40% utilization. Future state was 6 people at 80% utilization, with the overtime gone. The capital case for headcount reduction was real. The capital case for buying the next piece of automation was more interesting still: it was being justified on average-hour savings, and the actual savings, once you killed the overtime, were nearly double.

The mechanism is straightforward once you see it. A facility that runs at 90% utilization on average hours and absorbs 10% of demand in marginal hours is paying more for that 10% than the average suggests. Overtime is one form. Cross-shift backfill is another. So is the shift handoff where the first hour out of break is half-productive because the prior operator already left. The marginal hour does not show up in the labor-rate spreadsheet. It shows up in the variance.

The Modeling Move That Builds Confidence

Across these projects the pattern repeats. A Chicago-area baker producing 7.5M lbs/year measures throughput in pounds per hour, which is the most average-hour version of an operational metric you can build. Ovens are the hard constraint. He wants to add 4 to 6M lbs of capacity to capture a West Coast co-brand prospect, which means a second oven line and a multi-level conveyor system. A capital decision worth millions, queued behind a measurement framework that cannot tell him whether his current 7.5M lbs is constrained by oven capacity or by operational discipline he has not yet fixed.

The right sequence on a capital decision this size is not "buy the oven, then optimize." It is "model the marginal hour, then decide what the oven solves for." A digital twin overlaid on the facility CAD, with wireless sensors measuring true OEE for two weeks, tells you whether the current ovens are running 90% utilized or 60% utilized. If they are 60% utilized in the marginal hour because of changeover, startup, and feeder starvation, the second oven line is solving the wrong problem.

The 18-person trimming line on a different meat-processor's site told the same story. Conveyor speed was not set as the pace-setter. People filled the gap the conveyor left. The simulation showed 10 to 13 trimmers at 100 seconds of touch time per piece would deliver the same output the 18 trimmers were delivering. The constraint had been mis-identified. The capital decision they were about to make, more conveyor capacity, would have absorbed the labor inefficiency. Right-sizing the labor first turned a capital ask into a $250 to $300K annual operational gain.

What to Do This Week If You Are Inside a Capital Decision

First, ask what the labor rate in the case actually traces to. If the case quotes a single number, treat that number as suspect until you have seen the payroll distribution behind it. The plant above had a 15% gap between the quoted rate and the actual rate. Capital cases built on the quoted rate were optimistic by that margin.

Second, ask how many marginal hours the operation is currently absorbing. Overtime hours per week. Backfill shifts per month. Shift handoffs where the first hour is half-productive. Multiply by 1.5 to 2x the average rate. That is the number the capital case should be saving against, not the average.

Third, treat downtime data with the same scrutiny you treat financial data. A 1% reported downtime number, repeated across every row in the export, is not data. It is a status update. Push for credible measurement before you let the case proceed. The data quality is the binding constraint on the decision, not the analytics.

Fourth, separate the constraint from the symptom. If the operation is running overtime to make demand, the question is not "do we have enough capacity?" The question is "what is consuming the capacity we have, and would a capital project absorb that loss or solve it?" The second oven, the next conveyor, the additional packaging line, each of these is a way to absorb operational losses you could have eliminated for a tenth of the spend.

The Lens-Flip

Average-hour math makes the capital case look attractive. Marginal-hour math tells you whether the capital actually buys you the throughput, or whether it just buys you permission to keep your operational losses. Most capital decisions get made before that second model is built. The ones that hold up under the marginal hour are the ones worth signing.