Savings Above the Constraint Never Reach the Bank

The objection that ended the meeting before it started

A continuous-improvement lead cut in during the first five minutes of a labor and lines review, before anyone reached the recommendation slides. His point was blunt. If the study saves half a person in three different areas across three shifts, he cannot tactically execute that; you cannot send half a person home. And if a piece of automation gets the day's production done in six hours instead of eight, he is not sending anyone home at hour six, and he cannot add volume, because he is already at his cook limit. "What's the breadcrumbs to get me to where I have money in the bank," he asked, "not just a theoretical improvement of two hours of daily productivity."

He had just described the single most common way an operations improvement evaporates. The number on the slide was real. The cash behind it was not.

Savings above the constraint are ghost capacity

Every plant has one step that governs how much it can ship in a day. In a prepared-foods facility it is usually thermal: the cook, the retort, the oven, the pasteurizer. That step runs at a fixed rate, and nothing upstream of it can move a case faster than it clears.

When you speed up a line that feeds the constraint, you do not increase output. You increase the queue in front of the constraint. The line finishes its work earlier and then waits. Total throughput for the day is unchanged, because the cook still meters the flow at the same rate it did yesterday. The improvement is real in isolation and worth zero at the plant gate. Call it ghost capacity: hours or heads freed above the ceiling, credited as savings, invisible in the P&L.

The fractional-person problem is the same disease in a different form. A study that shaves fifteen minutes here and a third of an operator there is describing capacity that cannot be collected. You cannot lay off a third of a person. Those pieces only become money when enough of them stack into a whole position a supervisor can actually remove, inside one shift, on one line, in one department.

The two questions that separate real savings from theoretical

The fix is not more analysis. It is a discipline applied to the analysis you already have. Two questions, run against every line item on a savings summary.

First, does this saving sit above or below the binding constraint? A saving below the cook, on a step that is genuinely limited by people rather than by the thermal rate, cashes out. Ten operators at fifty thousand dollars, three of them genuinely not needed once the work is level-loaded, is a clean hundred and fifty thousand dollars, because it sits inside the labor envelope, not upstream of a ceiling. A saving above the cook is queue, and queue is not money. Cross it off, or move it to a separate column labeled honestly as headroom for future volume.

Second, can a real person execute this on a real shift? A saving is only bankable when it collapses into an action a plant leader can take: remove this position, on this line, on this shift. Reframe the goal accordingly. The question is almost never "how do we finish two hours early." It is "are we overstaffed across the full eight hours." Level-loading the same window with fewer people banks the labor; finishing early with the same crew banks nothing, because the crew still clocks eight hours and the constraint still sets the ceiling.

That reframe also decides where capital belongs. If the cook is the ceiling and you want more total output, the only spend that moves the number is spend on the constraint itself, and it has to clear an ROI a manufacturing buyer will actually accept. On the floor that means a stated payback, often a seven-year horizon, laid next to the total capital before a single quote is pulled. Automation anywhere else buys speed the plant cannot sell.

What a well-run savings case looks like

Every claimed saving names the constraint it sits relative to, above it or below it, and the ones above it are not counted as cash. Fractional saves are pooled into whole removable positions before they reach the summary, each tied to a specific line and shift. Capital proposals target the binding step and carry a payback number before anyone requests a quote. And the headline reads in dollars a plant leader can act on, not in hours of daily productivity that never leave the floor.

The savings were on the line, not at the gate

The two freed hours were real on the line. They were also worthless at the gate, because the cook never sped up and no one went home early. Put every proposed saving next to the constraint before you write it down, and most of what looked like money turns out to be queue.