The activation magic number

Some products have a known early-usage threshold that separates users who stick from users who churn. Cross it during onboarding and retention jumps. Fall short and the user is almost certainly gone. The famous examples get quoted as folklore - a social network’s “reach X connections in the first Y days”, a messaging tool’s “send N messages”, a file-sync tool’s “store one file on one device” - but the numbers matter less than what they represent: a measurable activation bar you can design onboarding to push people across.

The reason this is a mastery signal rather than a trivia point is the method. The number isn’t guessed, it’s found.

How you actually find it

You’re looking for the early behaviour that best separates retained users from churned ones. The process:

1. Pick a retention outcome that matters - still active at day 30, or converted to paid.
2. Pull candidate early behaviours - actions taken in the first session or first week.
3. For each, look at retention split by whether the user did it, and how much. You’re hunting for the behaviour where the retention curves fork hardest.
4. Find the threshold where the curve bends - the point past which more doesn’t help much and short of which retention collapses.

The output is “users who did X, N times, in their first W days retain at three times the rate of those who didn’t”. That’s your activation target, and onboarding’s entire job becomes getting more users across it.

The statistical health warning is the obvious one: this is correlation, not proof of cause. The behaviour might cause retention, or engaged users might just do both. The only way to know is to intervene - push more users across the bar and check whether their retention actually rises, ideally against a holdout. Plenty of teams optimised hard for a magic number that turned out to be a symptom of engagement rather than a driver of it, and moved it without moving retention.

Why it reshapes onboarding

Once you have a real activation metric it becomes the input metric the onboarding funnel optimises for, sitting upstream of revenue. Conversion to paid is too far away and too slow to test against directly. Activation rate is close, fast, and predictive, so it’s what tests target.

It also tells you what to cut. Anything in onboarding that doesn’t move users towards the activation behaviour is decoration. The setup-wizard step that makes the product look thorough but delays the user reaching value is actively harmful, because every day before activation is a day they can churn.

Where it breaks

• Treating a benchmark from another company as yours. “Seven friends in ten days” was that company’s number for that product. Yours is different and has to be derived from your own data.
• Optimising the metric instead of the value. If you nag users into hitting the number, you can move the metric without delivering the underlying value, and retention won’t follow.
• Picking a metric that’s really just engagement. Without an experiment to confirm the link is causal, you might be optimising a thermometer rather than turning up the heat.