Two forklifts that never moved

November 2026 · Lisbon

The first fleet we ever measured properly belonged to an automotive plant in northern France. Twenty-four assets: eighteen forklifts and six tugger trains. We put readings on all of them and then we did the hardest part of the job, which is nothing. We waited. For seven and a half months, every movement, every minute.

Before the measurement, the plant would have told you the fleet was about right. Not wasteful, not tight. Everyone says this. It is almost never a lie — it is an estimate, delivered with confidence, by people who are good at their jobs and have no data.

Here is what the readings said.

At the busiest moment of the busiest shift in seven and a half months, nineteen of the twenty-four assets were in use at once. That is the peak. The all-time high. Which means at the plant's most demanding moment, five machines stood still — and on an ordinary day, far more.

Two forklifts never moved at all. Not once, in seven and a half months. They were rented, maintained, insured, inspected and parked. Their utilization was exactly zero, to the decimal.

When we applied the real shift tables to the data — not a model, the actual rosters — the fleet was oversized by about €301,000 a year.

Nobody had done anything wrong. That is the uncomfortable part. The fleet had grown the way fleets grow: a busy quarter here, a new line there, a rental extended because returning it felt risky. Each decision was locally reasonable. Nobody could see the whole, because the whole had never been measured. The plant was not run by careless people; it was run on estimates, like almost every industrial operation on earth.

The sequel is my favorite part. At another plant, armed with this kind of reading from day one, the fleet went from ten forklifts to four. Same output. The change paid for itself in under a year — not through heroics, just by replacing a guess with a number and acting on it.

I have come to think of this as the general case, not the exception. Less than a fifth of what happens inside a typical operation is visible in real time. The rest is reconstructed afterwards from memory, spreadsheets and walking the floor. Factories are run by estimates the way medieval medicine was run by humors: sincerely, systematically, and blind.

The fix is not intelligence. The people were already intelligent. The fix is instrumentation. Measure first, for long enough that the data survives every "yes, but" — seasonality, the busy week, the special order. Then put the finding in euros, because euros are the only unit everyone in the building respects. Then let the number argue. A number that took seven and a half months to collect does not lose many arguments.

We turned that audit into a company. But the lesson is bigger than my product, and it costs nothing to apply: somewhere in your operation, two forklifts are not moving. You are paying for them this month. You will pay for them again next month. The only question is whether you would rather keep estimating — or know.