Main idea

To answer a fine-grained question about how the product behaves, you spin up a test environment, prepare a dataset, replay the scenario, and watch whether the product blocks, computes, hides or refuses. Half an hour goes by, sometimes an hour. Product Managers, CSMs, support and developers all perform this gesture, several times a week.

Nowhere is that time recorded. There is no ticket, so nothing in Jira. It matches no roadmap item. It enters no estimate, since only what has been decided gets estimated. It is absorbed by the normal running of the organization, just like walking between two meetings.

Two things make the bill worse and stay out of sight for the same reason. The checking is paid for twice: by the person searching, and by the person they interrupt because they are the only one who knows. And it is only one case of a broader expense — everything that describes the product to someone is a reconstruction, redone in parallel by each department: the documentation, the answers to tenders, the onboarding of newcomers, what support says. Each of them starts again from the code or from someone's memory, and none of those passes shows up in a budget.

Layer added by "I wrote a product's ontology. Three times, I thought I was done." (2026-08-11).

A cost that appears in no instrument cannot be compared with what removing it would cost. That is what makes it durable: it never gets arbitrated, not because anyone judges it acceptable, but because it is never put on the table.

Why it matters

This explains why the projects that would cut this cost — living documentation, a tool for querying the product — don't win their arbitration: they are compared against a visible budget while their gain is drawn from a budget that exists in no spreadsheet.

It also gives the first move to make before tooling anything: make the cost measurable, even roughly, rather than pleading on the intuition that it is high.

Nuances and limits

The invisibility isn't a property of the gesture, it is a property of the instruments in use. An organization that tracks time by activity rather than by deliverable would see this line appear — at the price of a tracking system with a cost of its own.

And part of this checking isn't waste: replaying a scenario teaches you something about the product that reading an answer does not.

Open questions

• How do you estimate this volume without installing a time-tracking system that would itself cost more than it reveals?