Your Time Study Measured The Task, Not The Day

Four Days, One Hundred Cuts, Two Different Rates

An operator spent four days on the trim floor of a protein plant with a stopwatch. About 100 cuts sampled. Roughly 20 pounds per cut. Average cycle time just under 53 seconds. That works out to 1,300 pounds per hour per trimmer while the trimmer is actively trimming. Good sample, clean method, a number you could defend in a room.

Then he asked the plant for its own data. Every flat cut on the line: when trim started, when it ended, what it produced. He assumed 21 people on the line, which the supervisors agreed was directionally right, and asked the sheet a simple question. All in, what is this line running?

Three hundred seventy-seven pounds per hour.

He ran it again for smaller cuts, because the plant pushed back that flats were not representative and they also run tri tips and short cuts. At 15 pounds per cut, still a large gap. At 10 pounds, still a large gap. The shape did not move. The measured rate and the delivered rate were separated by roughly 71 percent of the paid hours, and nothing about product mix explained it away.

Both numbers are correct. They are answers to different questions, and almost every labor case in manufacturing gets built without noticing which one it used.

The Stopwatch Starts When The Work Starts

That is the whole mechanism. A time study begins timing at the moment the work begins. It cannot see waiting for product to arrive at the station. It cannot see knife sharpening, the walk, the short gap while upstream catches up, the two minutes a supervisor pulls somebody for something else. The measured rate is a property of the task. The all-in rate is a property of the system the task sits inside. A stopwatch is structurally incapable of measuring the second one, and no amount of additional sampling fixes that, because you are sampling the wrong interval.

This runs in the other direction too, and it is worth checking both. In the ready-to-eat room of the same plant, the crew was asked how long it takes to unload one rack. Two people, six minutes. One person, twelve. That is linear, which tells you something specific: there is no fixed setup buried in that task, so it is pure labor trade. Plug the production schedule into it and you get about six hours of straight loading time. A task that behaves like that is a scheduling decision, not a headcount, and treating it as a headcount is the same error with the sign flipped.

The general rule: a rate you measured at the station tells you what a person can do. It never tells you what a line does. If you cannot state the ratio between the two, you do not have a labor model, you have a stopwatch reading and a hope.

Name The Discount Before You Promise The Number

Two failure modes sit on either side of this, and both are common.

Publish the 1,300 as the opportunity and the plant produces the 377 within a week, and the entire case dies on contact. You will not get a second hearing, and you should not.

Publish the 377 as capacity and you have just written every current loss into the baseline as physics. The waiting, the seam with upstream, the position that exists because it has always existed, all of it becomes permanent by arithmetic.

The move is to name the discount out loud, as a discount, before anybody commits to a number. Pick a utilization factor. Say plainly that you picked it and that it is not yet measured. State the headcount it implies. In this case the working assumption was around 80 percent, which put the endpoint at roughly 16 active trimmers against a line staffed well above that. Nobody pretended 80 percent was derived. It was declared, in the open, as the thing the test exists to measure.

Then stage the test so it actually measures the discount. Remove one position. Define the break-even before you remove it, so the room agrees in advance what "this worked" looks like: daily production still complete, labor hours not up, downstream work in process still fed. Hold it a week. Then take the next one.

One more rule, and it is the one that decides whether any of this survives contact with the plant. Declare the endpoint at the start. A test that advances one position per week can be cancelled after two weeks with everyone's original position intact, and the plant leader who spent a year insisting there was no opportunity will cancel it. Going in with "we believe this line runs at 16 and the test is to get to 16" is a different agreement than "let us try removing one and see." The first is a plan with a stated hypothesis you can be wrong about in public. The second is an experiment that ends the moment it gets uncomfortable.

What A Well-Run Labor Case Reads Like

Three numbers on the page: measured rate, all-in rate, and the utilization factor between them. The factor carries a date and a source, and that source is a staged test on the actual line, not a benchmark from a textbook or a different plant. Each step of the test has a break-even defined before the step runs, in the plant's own units. The endpoint headcount is stated in the first meeting, not discovered in the fourth. Every station on the line is either measured or explicitly listed as unmeasured, and nobody has to guess which.

The Distance Is The Opportunity

The stopwatch told you what a trimmer can do. The payroll told you what the line does. Neither number is the opportunity by itself. The distance between them is, and you only get to keep the part of it you can name.