Main idea

Describing a product's objects, rules and vocabulary had been estimated at a few dozen hours, spreadable area by area. The whole thing took under six hours, in one session. The error is a factor, not a margin.

And it is probably that estimate, rather than the work itself, that explains why nobody had undertaken it before: a project judged out of reach compares to nothing, it doesn't get planned, it doesn't even get proposed. The imagined cost acts as a barrier while never having been measured.

The real figure isn't an average either — one measurement, on one product, once; the sources were already gathered and indexed, and setup time isn't in it. It says as much about the tooling as about the method: the same work by hand would have taken weeks.

Why it matters

This indicates where to look for the projects that were never launched: not among those that were weighed up and refused, but among those that were never put on the table. What they have in common is an old estimate, never revisited since the tooling changed.

And it suggests a cheap countermeasure: attempt the work on a reduced scope rather than estimate it, since the estimate is precisely what turned out to be wrong.

Nuances and limits

The error can run the other way, and often does: an underestimated project opens and never closes. What transfers isn't "it's cheaper than you think", it's that estimating work whose tooling has just changed has no basis left.

And the figure authorises no promise of return on investment: it measures a build, not a use.

Open questions

• What other pieces of work considered out of reach rest on an estimate made before their repetitive part became automatable?