day-N retention

Retention rate answers the most important question about a product: do people come back? The standard product version is day-N retention (also called classic or N-day retention): take a cohort of users who started on the same day, and count the share who are active on exactly day N after that.

$\text{D}N\ \text{retention} = \frac{\text{users from the cohort active on day } N}{\text{users in the cohort}} \times 100\%$

Day 0 is the sign-up day, so D1 is the next calendar day. The usual checkpoints are D1 (did the first session land?), D7 (did a habit start?) and D30 (is the product part of life?). "Active" must be defined as a meaningful action, not any event: for Halves, opening the app or adding an expense by hand counts, while a bill the server created automatically doesn't. A small cohort table shows how it reads:

Day-N retention is strict: a user who comes back on day 6 and day 8 but not day 7 counts as lost for D7. That's fine for products used daily (messengers, games), but for a product used weekly or monthly it is noisy and understates real retention. For those, measure retention in weeks or months ("active at any time in week 2") or use a bracket such as days 7–13. Pick the period that matches the product's natural usage rhythm.

A different formula carries the same name in subscription and B2B businesses: customer retention rate over a period, $\text{CRR} = \frac{E - N}{S}$, where $S$ is customers at the start, $E$ at the end and $N$ new customers gained during the period. It measures what share of existing customers you kept, not how a cohort behaves over time. Say which one you mean.