I rebuilt onboarding after the PMF module wrecked my assumptions

Going into the PMF module I assumed onboarding was fine, because signups were fine. What the module forced me to actually check — mapping every step a new user took against the moment they got real value, not the moment they finished a tutorial — was a different story entirely.

The gap between "finished onboarding" and "got value"

Sixty-eight percent of new signups completed our onboarding flow. Nineteen percent came back the following week. That gap is where the entire product was quietly failing, and no amount of polishing the tutorial steps was ever going to close it, because the tutorial was optimising for completion, not for the moment someone actually felt the thing was worth their time.

The rebuild removed three onboarding steps entirely and replaced them with one: get the user to their first real result with their own data, not sample data, inside the first session. It's less impressive to demo. It's also the only version that moved the return-rate number, because it stopped optimising for a completed checklist and started optimising for the actual thing people were coming back for.