Retention analysis

The Retention report shows how many visitors return in the periods after their first visit, grouped into cohorts by their start date.

How to read it

• Each row is a cohort (everyone who first visited on a given day/week/month) with its size.
• Each column is a period since that first visit (0, 1, 2, …).
• Each cell is the percentage of that cohort that returned in that period — darker means higher retention.

Controls

• Period — daily, weekly or monthly cohorts.
• Segment — restrict retention to a specific segment to compare audiences.

The retention curve

Alongside the cohort grid, the curve averages retention across cohorts into a single line. A curve that flattens after the first few periods is healthy — it means you have a core of visitors who keep coming back. A curve that keeps falling towards zero means you are renting attention, not building it.

Churn risk

The Churn risk table scores individual visitors on how likely they are to stop returning, so you can act before they've gone rather than after.

• User ID, Last Seen, Days Away and Sessions give the raw picture.
• Recency, Freq Decline and Engagement are the components behind the score — how long since they visited, whether their visit frequency is dropping, and how deeply they engage when they do.
• Score combines them into a single number, and Risk buckets it: Low, Medium, High or Churned.

Filter by risk level to get a working list. High risk is where intervention is still worth something; Churned is a retrospective, useful for asking what those visitors had in common.

Click through to a person's profile to see what they actually did before they drifted. Churn risk depends on visitors being recognisable across visits, so it is far more useful once you identify them.

Period 0 is always 100% (the cohort's first visit). The interesting signal is how quickly the later columns fall off — and whether they ever level out.