Retention

How it works

Start with everyone who first opened the app on one day. For each later day, count how many of them were active, and divide by the size of the starting group. Plotted over time this gives a retention curve that typically falls steeply at first and then flattens.

Two choices change the result, so state them whenever you quote a figure.

• What counts as coming back. An app open is the usual default. A stricter definition uses a meaningful in-app event, such as an order or a finished article.
• Which days count. Classic retention asks whether the user was active on exactly day N. A looser version, often called unbounded or rolling retention, asks whether they were active on day N or any day after it. The looser version never gives a lower number.

When it matters

Retention matters for every app, and most for those whose revenue depends on repeat use. A content app lives on daily habit, and a delivery app earns its profit from the second and later orders. In both, a campaign that brings users who leave after a day has bought nothing.

How it relates to attribution

Once each new user is assigned to a source, retention can be read per channel and per campaign. This is often the earliest sign of traffic quality: revenue takes weeks to appear, while a weak day-1 or day-7 rate shows up within days. Retention is measured on a cohort, the group of users who started together, and it is one of the main inputs to LTV.