Activation event

Signup is a promise; activation is the receipt. Activation is a measurable criterion that says a new user reached the point where your product did something for them. Sometimes that is one tracked event, and often it is a composite: N of an event within a window, or two different events both completed. The famous ones are oddly specific: Facebook's 7 friends in 10 days, Slack's 2,000 messages sent inside a team, Dropbox putting one file in one folder on one device. Specific is the point, because a vague definition cannot be counted. Whatever the shape, it has to evaluate to a clean yes or no per user, so store it as a derived 'activated' flag with a timestamp rather than re-deriving it in every report.

You find yours by comparing the first-week behavior of users who stuck around against users who vanished, then looking for the action where the two groups separate hardest. Once you have it, it becomes the target of your onboarding, the last step of your signup funnel, and usually an input to the North Star metric.

Gotcha: that comparison finds correlation, not cause. Retained users also update their avatar more often, and shoving new users toward the avatar screen will not retain anyone. Sanity-check a candidate by asking whether the action plausibly delivers value, then test it properly: randomize new signups into an onboarding that pushes toward the criterion versus one that does not, or hold back a slice, and compare retention between the arms. Shipping the change and eyeballing the next cohort proves nothing, because seasonality, traffic mix, and everything else you shipped that month moved too. Second trap: pick an activation bar so high that almost nobody clears it and the metric stops moving enough to teach you anything.