Measure the Line, Then Use Their Headcount

The number you argue about is not the number you measured

Every labor reduction on a manual line meets the same question in its first review, and it is never about the time study. The question is: compared to what?

Get that wrong and the study does not matter. You can put a stopwatch on a hand-paced station for four days, build the work content cleanly, and still lose the room, because the person across the table is quoting a different starting headcount than you are and both of you are right.

This part of a labor case gets treated as bookkeeping. It is not bookkeeping. The baseline is the most contested number in the project, and whoever sets it decides what counts as a win.

Three numbers, all of them true

Any manual line carries at least three legitimate headcount figures at once.

There is the roster, what the plant staffs to. Plants staff deep on trim, cut, sort, and hand-pack stations because absence is real and a short line stops. There is the active count, how many people are actually at the stations on a given shift. And there is work content, how many people the volume requires at a rate the line can hold.

Round illustrative numbers make the shape clear. A plant staffs thirty to a cut line so it can keep twenty-six running through call-outs, and the work content at a defensible pace supports twenty-one. Nobody is lying. The HR system says thirty, the supervisor says twenty-six, the study says twenty-one. Those values are invented for the example; the spread is the real thing, and the spread is where the argument lives.

A fourth number sits in the production log: the all-in rate. Take the line's own output records, divide by the hours the crew was on the clock, and you get units per person-hour for the shift. Then measure the same operator doing the same work with the work in front of them. The second number is routinely several times the first.

That gap is the finding, and it is almost never operator speed. It is waiting for feed. It is a station starved while the one upstream is blocked. It is handling and staging nobody assigned to anyone. It is break coverage that empties three positions at once because the line has no rotation. It is a conveyor set to a pace chosen for a headcount the line no longer runs. The stations interact, and the interaction sets the rate. Add people to a starved station and the all-in rate gets worse, which is exactly what the plant will point at the first time you ask for a reduction.

Set the target off their number

Here is the rule I use, and it runs against what the arithmetic suggests.

Measure the line properly. Then state your target against the plant's own stated headcount, not against your measured potential.

If the study says the work supports twenty-one and the plant says it runs twenty-six, do not open with twenty-one against a roster of thirty. That is a nine-head claim built on two figures the plant does not accept, and it will be argued as a nine-head claim for a year. Open with twenty-six, because the plant said it, and name the endpoint you believe in. Then step there.

One person on the plant side owns that baseline and signs it, by name. Not the improvement team, and not a committee. If the number belongs to the people running the line, the reduction is theirs to defend. If it belongs to you, every step is yours to relitigate.

Four things follow, and each matters more than the size of the number.

Name the endpoint up front. If the target is twenty-one, say twenty-one on day one. Do not take one head out, then another, and let the destination emerge. An open-ended reduction reads as a squeeze with no floor, and it gets cancelled the first week production misses, because nobody agreed to where it ends.

Discount the rate and write the discount down. Nobody runs all shift at the rate you measured while they were actively working. There has to be an allowance for the real day. Pick the utilization you think is fair, state it as a number, put it in the document, and let people argue with the number instead of with the conclusion. A target built on an undiscounted rate is a target you defend once and never again.

Take the free position first. On most manual lines there is a station already empty some days with nothing downstream noticing. Someone catching bags off a conveyor into bins. A second loader who is only loading half the time. That reduction needs nothing built, nothing bought, and no argument about work content. It also proves the method before you ask for anything expensive.

Then let the test close the gap. Step down one position at a time, hold each level long enough to read real data, and use the line's own output records to establish that the previous level held. Every step that holds is evidence you did not have to argue for.

Keep the states separate while you do it. A step that holds for two weeks is an implemented change. It becomes realized labor cost only when the paid hours for that line actually drop in the payroll data, and those are different dates. Book the second one, not the first.

What a defensible labor baseline looks like

One headcount figure for the line, pulled from twelve months of actual paid hours rather than the week you were sampling, with operations, finance, and the improvement plan all quoting the same figure. The rate target is stated as a measured rate times a written-down utilization, never as a maximum. Every reduction step names the level, the hold period, and the output metric that confirms it. The endpoint is on paper before the first person moves. A supervisor can reconcile the plan's headcount against the schedule with no translation step.

Twelve months matters more than it sounds. A one-week sample is wrong in both directions and you cannot tell which. Fourteen people on the line the week you look might mean the line runs on fourteen, or it might mean four people called out. Pull the trailing twelve and the seasonality, the overtime, and the true staffing pattern all arrive together.

Give up the credit fight

The last piece is discipline, not analysis.

By the time a labor case gets underway, a plant is usually already taking heads out. Some of those reductions were in motion before anyone was engaged. There will be an argument about who gets credit for them, and it will run in the background whether you join it or not.

Do not join it. Fighting for credit on reductions already in flight costs you the one thing you actually need, which is an agreed starting number. Concede those heads. Take the plant's current stated headcount as the baseline, whatever it is, and build forward from there. You give up a number you were never going to collect cleanly, and you get a baseline nobody can withdraw.

Whoever owns the baseline owns the result. That is worth more than the credit.