Activation metrics are diagnoses, not goals

Slack didn't discover that two thousand messages caused retention. They discovered that teams that retained had sent two thousand messages. The activation metric is forensic. Pushing teams toward the number doesn't build retention; it builds the appearance of retention, which is what charts look like when the underlying fit isn't there.

The number is a diagnosis

Every famous engagement threshold — Slack's two thousand messages, Facebook's ten friends in seven days, Notion's collaborative workspaces — was discovered by looking at teams that stuck and asking what they had done that the teams who left had not. The number was downstream of the survival.

If you treat the number as the cause, you optimize for the number. If you treat the number as a symptom, you optimize for whatever caused users to generate it in the first place — usually, the product being the right answer for the work the user was trying to do.

Nudging the number is dressing up the disease

There are well-worn ways to nudge a metric upward. Email reminders. Streak counters. Notification badges. In-app nudges that say "you haven't done X yet." Many of them work, in the narrow sense that they make the number go up.

What they don't do is make the user care. A user who sends a message because the app asked them to hasn't formed a habit. They've responded to a nudge. The next nudge has to do more work than the last one. The chart goes up and the cohort goes down, and you find out at the next quarter end when the renewal rate prints.

The version of this that works is the one where the product is the obvious place to do something the user was already doing. Slack works for teams who needed to chat anyway. Linear works for teams who needed to track issues anyway. In those cases, the activation threshold is just counting what was happening anyway. Elsewhere, it's counting what marketing got people to do once.

The engagement work is the fit work

When activation looks weak, the temptation is to redesign onboarding. Sometimes that helps. More often the onboarding is fine, and the people walking through it are not the people for whom the product is the obvious place to do the thing.

The work that drives engagement, then, is mostly the work that finds fit. Talking to the users who came back. Watching what they did before they came back. Trimming the product, the onboarding, and the cohort definition until the path from signup to the second use is short and obvious. The metric improves because the underlying thing improves. The metric was not driving anything; it was just easier to see.

Loops are recognized, not designed

Engagement loops exist. The product creates value, the value creates use, the use creates more value. Slack, Linear, Notion all have versions of this.

The loop is a property of the fit, not a feature you add. Teams that needed group chat got the messaging loop. Teams that didn't, don't have it now. You can't graft a loop onto a product that doesn't have one for the cohort you've chosen.

The honest move when no loop exists is to figure out which job the product would loop on, for which cohort, and reshape until the loop appears. That's not an engagement strategy. It's a product strategy. The engagement chart follows.

The shortest path to the activation number is the same as the shortest path to the cohort that generates it. Skip the nudges. Listen to the people who already get it. Make the product more of the thing they get.

The number was always reporting on the cohort. It wasn't driving them anywhere.