Retention lessons from building a daily puzzle (the streak is not the feature)

Daily puzzles look like a solved design problem: one board, one day, come back tomorrow. Having built one, I think most of the retention advice about them is aimed at the wrong mechanic.

The streak is not the feature

Everyone adds a streak counter, and it does work — for the people who already play. It does nothing for the person who missed day one. A streak rewards consistency you already have; it does not create the first visit.

What created repeat visits in my numbers was simpler: the puzzle being solvable in a sitting. If a player can finish in three minutes on a phone, they come back. If it takes fifteen, they say "later" and never do.

Make the first session end in success

The most useful metric I tracked was not day-2 retention. It was: what fraction of first sessions ended with a solved board? Players who solved on day one came back at a far higher rate than players who did not, regardless of how long they played.

That changed the design. The early boards got easier than I wanted them to be. It felt like giving something away. It measurably helped.

Sharing is a growth loop, not a retention feature

The shareable result grid is usually filed under growth. In practice it also drives retention, because seeing a friend's result is a reminder that today's board exists. But it only works if the shared thing is legible without explanation — a grid of coloured squares with no instructions, not a score out of a thousand.

What I would cut

• Accounts. Not needed on day one. Cost me a signup wall's worth of first sessions.
• Tutorials. The board teaches itself faster than a modal does.
• Timers. They add anxiety and subtract return visits, at least for a casual puzzle.

The short version

Ship one daily number puzzle, keep it solvable in a sitting, and let the first session end in a win. The streak counter can come later — it is a reward for the habit you have already built, not the thing that builds it.

---

Written after rebuilding the same idea three times. The third version was the smallest and the only one people kept playing.