A team can say what problem the feature solves, who has it, how big that group is and what those people use today. Nobody can say what would have to be true in six months for the feature to count as having worked. Which question is open?

The principle — a team that cannot say what success would look like cannot later tell whether it happened.

Why the key is correct — the last question on Cagan's list is How will we measure success/make money from this product? (metrics/revenue strategy) It is asked before the work so that the bar is set by the argument for building, not by whatever the numbers turn out to be.

Why the others are wrong — deferring the measure until data exists lets the outcome pick its own definition of success. Reaching for market size ignores the scenario, which says the group is already sized. And an assessment completed after shipping has no power over the decision it was written to inform.

Remember this — decide what would count as working before you build the thing that has to work.

Sources — Marty Cagan on assessing product opportunities, Silicon Valley Product Group.