Why install-week cohorts misread a daily ritual

The first time a founder shows us a retention table, the left-most column is almost always “installed this week.” That column is inherited from advertising reports, where the download is the conversion. In a habit product the download is closer to buying a mat. Nobody would judge a yoga studio by how many people purchased a mat on Monday and stood on one by Sunday, without asking whether they ever unrolled it.

Take a morning stretch product we read in 2025. Install-week Day 7 sat at a number the team was proud of. When we recut the same export with week zero as the first completed stretch — a ten-minute circuit the person had to finish, not merely open — Day 7 among those people was lower, and Day 14 was more stable. The proud number had been carrying people who installed on a Friday night, opened a paywall, and never stretched. They were still being counted as “in the week.”

The correction is not complicated, and it is not a slogan. Name the ritual in one sentence. Find the event that fires only when that ritual is completed. Make that event the door to the cohort. Keep the install-week table if you must talk to an advertising partner; do not use it to decide what happens on day two of the stretch.

There is a practical catch. If your completed-ritual event is wrong — if it fires when a timer ends in an empty room — the new cohort will be just as misleading. That is why we draw the canvas before we cut. The door has to be a real door.

Teams sometimes worry that moving week zero later will make early numbers look worse in a board pack. It will, once. After that you are reading the people who actually tried the habit. Those are the only people a retention conversation can honestly be about.

If this note names the argument in your team, send the ritual and the week people vanish.