# Reading day-7 retention without fooling yourself

> Day-7 retention is the number every game studio quotes. Here are four ways it lies, and how Cadence shows the real curve.

- URL: https://posthog-example.notra.site/blog/day-7-retention
- Published: 2026-08-19
- Authors: Ines Duarte
- Tags: guides

Every publisher call starts with the same question: what’s your D7? It’s a good number to ask for. It’s also easy to get wrong without noticing.

## Classic or rolling

Classic D7 counts players who opened the game exactly on day 7 after install. Rolling D7 counts players who opened it on day 7 or any day after. Rolling is always higher, often by 5 to 8 points for a casual puzzle game.

Neither is wrong. Mixing them is. Cadence labels every retention chart with the method, and the default is classic because that’s what most publishers mean.

## Four ways the number lies

- **Install cohorts that are too small.** Below about 500 installs a day, D7 swings by several points from noise alone. Group by week instead.
- **Paid and organic in one cohort.** A burst of cheap paid installs drags D7 down for a week, then it “recovers”. Split by acquisition source first.
- **Time zones.** A day is a calendar day in the player’s time zone, not UTC. Counting in UTC moves late-evening sessions into the wrong day.
- **Reinstalls.** A player who deletes and reinstalls is not a new player. Cadence matches them on the device ID you send and keeps their original install date.

> **Note**
>
> Cadence shows a confidence band around each point once a cohort has fewer than 1,000 players. If the band is wider than the change you’re looking at, wait another week.

## What a healthy curve looks like

For a free-to-play mobile game, a D1 around 40% and a D7 around 15% is a solid start. The shape matters more than the points. A curve that flattens after day 14 means you have a core audience. A curve that keeps sliding means you don’t yet.
