Read week four before you argue about week twelve

Twelve-week return looks serious in a board pack. In a young app, those cells are often a handful of people. Start with the week that actually has a sample.

A weekly cohort table is a grid of people who first arrived in a given week, and whether they came back in later weeks. The far-right columns look important because they are far. They are also the first to become gossip: five returners in week twelve, dressed up as a rate.

In studies we ran in Kuala Lumpur last year, more than one team wanted to debate month-three return while week-four cells were still the first honest picture of habit. Week four is not magic. It is simply where many Malaysian consumer apps still have enough people left to talk about jobs: a second grocery order, a second article, a second ride that was not a refund.

When we mark a table, we write the headcount under the rate. If week twelve is twelve people, we treat it as a list, not a percentage. The argument then moves to week one and week four: which prompt, which failed substitution, which language. That is slower than a single chart. It is also harder to accidentally lie.