Main idea

Asking a product team to prove that a need exists leads nowhere if you expect the standard of proof of a demonstration: no interview, no ticket, no usage data is sufficient on its own, and each can be refuted with "that's just one customer" or "that's just a correlation".

What counts as evidence here is of another kind. Three interviews in unconnected companies, a sales objection recurring in another segment, an adoption friction observed by the CSMs and a workaround visible in the usage data demonstrate nothing taken one by one; their convergence, on the other hand, makes a hypothesis hard to dismiss.

So the strength of product evidence lies in the number of independent paths leading to the same point, not in the solidity of the best one among them.

Why it matters

This changes what you ask for before a trade-off. Looking for the irrefutable signal delays things indefinitely; looking for independent sources that meet is feasible and comes to an end.

It also gives the criterion that separates good collection from large collection: adding up ten pieces of feedback from the same channel and the same segment adds almost nothing to the convergence, whereas two pieces from two different worlds strengthen it.

Nuances and limits

Convergence can be manufactured without anyone noticing: signals collected by the same people, from the same accounts, with the same questions, are not independent — they confirm each other because they share an origin.

And it remains a probability, not a certainty: several sources can converge on a false reading, especially when each of them repeats what the others already say.

Open questions

• How do you assess the real independence of two signals, when the collecting is done by an organization that shares its interpretations continuously?