Retention analysis tells you how often users come back after their first visit. Instead of one number, the retention heatmap is a grid where each cell is the percentage of a cohort that returned on a given later day.
Anatomy of the heatmap
- Rows (cohorts) — each row is a group of users who first visited on the same day (or week / month, depending on the granularity selector).
- Columns (intervals) — Day 0 is the cohort's first visit; Day 1 is the next day; Day 7 is one week later, and so on.
- Cells (return rate) — the % of users from that cohort who came back at least once during that interval.
The far-left column (Day 0) is always 100% — by definition every user in the cohort visited on their first day. The further right you go, the more telling the numbers.
What "good" retention looks like
There's no universal benchmark — it depends on your product. But here are common patterns:
- Content / news sites: high Day 1 retention, sharp drop after Day 7, long tail at Day 30+.
- SaaS products: Day 1 typically 30-50% for healthy products; if Day 7 stays above 20%, you have product-market fit signals.
- E-commerce: Day 30 retention is more relevant than Day 1 — repeat purchases tend to space out.
The colour gradient encodes retention strength. Darker green = higher retention. A row that turns very pale after Day 1 is a churn signal.
Reading the rows top-to-bottom
If the most recent cohorts (top of the table) have stronger retention than older ones, your product is improving. If the opposite, recent product changes may have hurt onboarding.
Cohort definition
By default a cohort is "users who had their first session on day X". You can switch to:
- First conversion — cohort by the day they completed a goal.
- First purchase — cohort by the day they made their first purchase (revenue plans).
- Custom event — cohort by the day they sent any custom event.
Switching cohort definitions can produce dramatically different heatmaps — pick the one that maps to your activation moment.
Sample size warnings
If a cohort has fewer than 20 users, the cells will display the absolute number rather than the percentage. Below 20, single-user behaviour can swing the percentage wildly.
Plan limits
| Plan | Max cohort window |
|---|---|
| Free | Not available |
| Pro | 30 days |
| Scale | 90 days |
| Enterprise | 365 days |