You do not need a faster crew. You need a better work list.
For years I thought the way to get more out of my tissue culture lab was to make everyone faster. That works, but it is slow, and it has a ceiling. Then the data showed me something I had been walking past every day.
A tissue culture lab does not ship plants. It makes plantlets, tiny cultures in jars, which are a manufacturing stage well upstream of a finished plant. Everything below is counted in plantlets, not plants.
A fast worker with the wrong assignment
Take one technician. I will call him Tech A. On the cutting step, where we divide and transfer cultures, he is one of the slowest people in the lab, near the bottom of forty. On the pre-rooting transfer, the setover where plantlets are moved onto rooting medium, he is one of the fastest in the building. He is not a slow worker. He is a fast worker with the wrong work in front of him.
We had another technician, Tech B, who was the exact mirror image. Fast at cutting, slow at the pre-rooting transfer. For a long time both of them split their time across both steps, which felt fair and normal.
Worth saying plainly, because it is what makes this cheap: nobody has to move. Techs stay at their own hood. What changes is which plants we deliver to which hood. It is a scheduling decision, made on paper, the week before.
| Technician | Cutting step | Pre-rooting transfer |
|---|---|---|
| Tech A | slow (bottom quartile) | fast (top third) |
| Tech B | fast (above median) | slow (below median) |
Two people make a clean story. Across the whole crew it looks like this, one number for each person at each operation. Read down a column to see who is fast at that step. Read across a row to see where each person is strongest. The gold outline marks each person's best operation, and notice they land in different columns.
Try it: swap the two assignments
Say Tech A is assigned cutting and Tech B the pre-rooting transfer. Each has the work they are worst at. Press the button and trade the assignments. Nobody moves. Same two people, same hours, same payroll.
Both operations speed up at the same time, for free. When I worked it out on the real numbers, every hour I reassigned was worth roughly ninety more plantlets.
Made in a single conversation. Paid out for the entire year. And this was not a hunch. I could only see it because we log how many plantlets each person makes at each operation, and how long it takes. Plantlets per hour, per person, per step. Most labs never measure this. They know roughly who is fast, but not who is fast at what.
One person is a hint. The whole crew is the number.
If one swap is worth that much, what happens if I stop guessing and optimize the whole crew at once? This is a classic assignment problem, the kind linear programming was built for. You give it every technician's real rate at every operation, the amount of work each operation needs, and the hours each person has. It returns the assignment that produces the exact same output in the fewest labor hours.
I ran it on one crop line, on the last three months of real work, with the operation speeds defined the way our production network actually models them. That includes a detail most people miss: the same cut takes a different amount of time depending on the routing decision behind it. A finer cut to stretch short stock is slower than dumping a surplus, and the model knows the difference.
The model only moves people between jobs they have already proven they can do, so it is a floor, not a ceiling. Cross-training would raise it.
And it schedules the work that exists. It does not invent easier work, and it does not choose the routing decisions, which are dictated by inventory. Even with those guardrails, an eighth of the labor was sitting on the table.
The manager's side of the ledger
Here is the part that makes this worth doing. The hard part is not the decision. The hard part is having the numbers. Once a lab is recording plantlets made per person per operation, the analysis is a few days of work. Pull the rates, run the assignment, make the calls. A manager can do the whole thing in less time than a single production week takes.
And the decisions are one-time. You change a few assignments, and that is it. The gain does not have to be re-earned every week. It shows up in every week that follows, at the same output and the same payroll, until the crew or the work changes.
So the trade is lopsided in your favor. A few days of thinking, done once, against months of saved labor. On a single crop line that was worth hundreds of hours a quarter. Multiply it across every line a lab runs, every quarter, and the few days a manager spends here may be the highest-return work they do all year.
Two ordinary ideas
None of this is exotic. It is two ordinary ideas put together. Measure how fast each person is at each specific operation. Then let the math, not habit, decide who gets which work.
So here is my question for anyone running a production lab. Do you actually know your per-technician, per-operation rates? Not who is generally good, but the real plantlets per hour for each person at each step. Because if you do, the gain is already sitting in your schedule, waiting for you to claim it.