Building

How to keep people coming back

You can buy installs, attention and even reviews. Retention is the one number that has to be earned, which is why it’s the one worth optimising.

1. Retention is the only number that can’t be bought.

   Every other metric has a price list. This one only responds to the product being good enough to return to, which makes it the most honest signal you have.

2. Find the aha moment and force it earlier.

   Facebook famously worked to get new users to seven friends in ten days, because that was the point people stopped leaving. Every product has an equivalent threshold, the job is finding it and dragging it forward in the experience.

3. Daily beats better.

   A product with a daily reason to open will out-retain a superior product opened monthly. Pi-Squared gives six questions a day and held roughly 78% week-on-week retention in its launch period, the rhythm did that, not the sophistication.

4. Streaks work, and streaks punish.

   Duolingo built much of its retention on streaks, then had to invent the streak freeze because a broken streak makes people quit entirely. Any mechanic strong enough to pull people back is strong enough to push them away.

5. Variable reward beats constant reward.

   Predictable outcomes stop being interesting. The Hooked loop -- trigger, action, variable reward, investment -- describes most things people open daily without deciding to.

6. Notifications are a loan against goodwill.

   Each one spends trust you have to earn back. A notification that isn’t worth the interruption doesn’t just get ignored; it gets the whole channel switched off permanently.

7. Build the return trigger into the session.

   The best time to give someone a reason to come back tomorrow is while they’re still here today, a result to defend, a puzzle waiting, a friend’s turn pending.

8. Measure cohorts, not totals.

   Cumulative totals only ever go up and will comfort you all the way to zero. A cohort chart shows whether the product got better or you simply added more people to a leaking bucket.

9. Resurrect the lapsed before chasing the new.

   Someone who used it for a month already understands the product. Winning them back is cheaper than any acquisition channel and tells you precisely what broke.

10. If week one is bad, nothing downstream matters.

    Monetisation, growth loops and How to grow all multiply retention. Multiplying a small number is a poor use of a year.