Staffed 27, Running 23, Counted 20

The supervisor knew a number nobody had written down

A consultant spent four days on the floor of a Midwest protein processor, walking the further-processing lines with the area supervisors to confirm his assumptions before building a labor case. The two packaging lines checked out exactly. What he saw was what they ran.

The trim line did not. Every time he had walked it, he counted around 20 people. On the last day he asked the supervisor to confirm the active headcount, and got a different answer: they staff 27 to have 23 on at all times.

Three numbers for one crew. About 20 observed, 23 running, 27 staffed. None of them is wrong. They answer different questions, and only one of them was going into the case.

He had already seen the same thing in the ready-to-eat room without recognizing it. The week before, a person had been stationed right after the X-ray machine, pulling bags off and dropping them into bins. The following visit that station was empty. Nothing backed up. Nobody flagged it. He went back to the supervisor, who confirmed they normally have someone there, and they just did not that day.

Staffing is a buffer policy, and the buffer does not show up on the floor

The gap between 27 and 23 is not sloppiness. It is a policy. Absenteeism, vacation, training, rotation to other lines. The plant staffs above its run requirement so the line still runs at requirement on a bad Monday. That is a defensible choice, and most plants make some version of it.

The problem is that the buffer is invisible from every direction you would normally look at it.

A walkthrough sees the low number, because on any given day part of the buffer is somewhere else. Payroll sees the high number, but as a department total with no split by line or by shift. And the line itself never complains, because absorbing the variation is the entire reason the buffer exists. A position inside the buffer is load-bearing on the worst day of the month and slack on the other nineteen.

That is why the station after the X-ray could disappear for a day and cost nothing visible. The same room ran what the team described as two and a half loaders: two full time, and one splitting time between loading and rework. A half position defined by what it does the rest of the time is a position nobody has ever had to justify out loud.

Meanwhile the census under the case was moving. Between mid June and the site visit, the process area had already dropped about five heads on its own. Two to four more people were slated to shift to new lines starting in October, and an ergonomic lift table was on order that would take trim loading from four positions to two. None of that was the consultant's work, and all of it changed the denominator he was measuring against.

Get all three numbers, then take positions out one at a time

Three things fix this, in order.

Write down all three numbers, per line and per shift. Staffed, scheduled, running. Then make the savings case state which one it is counting against. A case built on the observed number understates the opportunity and leaves real positions on the table. A case built on the staffed number promises reductions the buffer will quietly refill the first week someone calls out.

Change the unit from bodies to rate. Bodies is the wrong unit precisely because the buffer moves and volume moves underneath it at the same time. This plant moved to labor hours per 1,000 pounds of good product, which normalizes for both.

Getting that number is real work, and most plants underestimate it. At a second processor, the operations and reporting leads spent a working session narrowing a data request from twelve months to year to date, and from yield to throughput, just to back into a rate per hour. What they actually needed was run start and stop times, labels printed, weights per combo, and plan versus actual by line. The reporting lead's verdict on the list was specific: daily finished goods output was the hard one, the rest were not. If a plant cannot produce output per day without a project, it cannot tell you what a single position is worth.

Then take positions out one at a time. The plan on the trim line runs a baseline week at 20 active trimmers, then pulls one position, holds, reads labor hours per 1,000 pounds, and pulls the next. The removal sequence is the measurement instrument. Going straight from 23 to a target of 16 tells you only that something broke or nothing did, never which position was surplus, and the argument that follows is unwinnable in either direction.

One distinction is worth protecting while you do this. Separate what is already decided from what is still a hypothesis. The lift table taking trim loading from four to two is not a test. The equipment is bought, the justification is done, and the status is waiting on installation. It sat in the same opportunity funnel as the untested trim reductions, which is an invitation to relitigate it later as though the outcome were still open. Decided work gets documented as captured, with a delivery date. Hypotheses get documented as tests, with a metric and a stopping rule.

What a well-run floor looks like

Staffed, scheduled, and running headcount are posted per line per shift, and the three get reconciled weekly rather than discovered during a site visit. Labor reads in hours per 1,000 pounds of good product. Every position has a named function and a stated absorption rule, so an empty station is either a problem someone reports inside the shift or a position that was already surplus. Reductions move one position at a time with a week of rate data behind each. And anyone who needs daily output by line can pull it the same day without opening a ticket.

The four heads on the schedule

The trim line was never running 27 people, and it was never running 20. Somebody had to ask the second question to find the four heads that exist only on the schedule, and those four heads were the entire argument.