13. Retention and engagement experiments

§The question: why do people leave, and what would make them come back on their own?

§Retention was the foundational validation signal; now it is the highest-leverage growth stage, because, as chapter 6 argued, retention drives revenue, referral and the economics all at once. An hour spent on retention moves more than an hour spent anywhere else in the funnel.

§ 13.1Habit around a natural frequency#

§Products that retain are built around a frequency that already exists in the user's life. Nir Eyal's hook model, trigger, action, reward, investment, describes how a habit forms: an external trigger becomes internal, the action delivers a variable reward, and the user's investment makes the next turn easier and more valuable.

§The critical word is natural. A product used naturally once a month cannot be forced into daily use, and trying produces annoyance, not habit. The retention experiment is to find the product's real frequency, from the retention curve's shape, and build the trigger and reward around it. Duolingo's streak is the most-studied example: it works because language practice rewards daily repetition, so the daily trigger fits the job. The same streak on a monthly product would be noise.

§ 13.2Deliver value repeatedly#

§The activation moment proved value once. Retention requires proving it again, each time. A retention experiment often takes the form: what is the recurring value, and how do we make sure the user gets it each cycle? A reason to return that is the product working, not a notification pretending the product worked.

§This is where engagement and retention are distinguished. Engagement, time in app, sessions, clicks, is a means, and only a good one when it serves the user's job. Engagement that serves the company at the user's expense, dark patterns, manufactured urgency, infinite feeds, buys short-term numbers and long-term churn and reputation. Eyal himself followed Hooked with Indistractable, and the pairing is the ethical point: build habits around value the user wants, not compulsion the user will resent.

§ 13.3Study the users who left#

§The richest source of retention learning is the users who churned, and it is the one teams avoid because it is unpleasant. The experiment is not an A/B test; it is the qualitative read of chapter 4, applied to leavers: what did the users who stopped have in common, when did they stop, and what were they doing, or failing to do, just before?

§Churn clusters. Users who never reached a second value leave early. Users who hit a specific wall leave there. Users whose life changed leave for reasons you cannot fix. Separating these tells you which churn is addressable and which is not, and it comes from talking to leavers and reading their behavior, not from the aggregate rate, which only tells you that churn is happening.

§ 13.4What you leave with#

§The product's natural frequency, read from the retention curve, and experiments that build a habit around it rather than against it. A recurring value the user gets each cycle. A qualitative study of churned users, separating addressable churn from the rest. And engagement metrics kept firmly in their place, as means to retention, never as the goal. The next chapter is the loop that turns retained users into new ones.