What it is.
Step one is a definition you can compute. You write churn as an event filter, so for example unique users who did an action in the last 30 days and stopped in the last 15. That gives you a number and, more usefully, the users behind it.
Step two is the lifecycle chart, which splits your active users into new, returning, resurrecting, and dormant each period. A rising dormant band is churn arriving before it shows in revenue.
Step three is the diagnosis. Save the churned users as a list, then watch session recordings from that list to see what they hit before they stopped.
What you get.
- A churn definition written as an event filter, so two people compute it the same way.
- Lifecycle charts splitting active users into new, returning, resurrecting, and dormant.
- A saved list of the users who churned, ready to filter anything else by.
- Session recordings from that list, showing what happened before they left.
- Actions you can reuse as the churn definition across dashboards.
How to set it up.
Write your churn definition as an event filter, naming the action that means active and the window that means gone.
Set the two windows deliberately, since 30 days of activity and 15 days of silence means something different for a weekly product.
Build a lifecycle chart on the same action and read the dormant band over time.
Save the churned users as a reusable list so you can filter anything else by them.
Open session recordings filtered to that list and watch five sessions end to end.
Turn what you see into one change, then re-check the same lifecycle chart a month later.
Use cases
Write the definition down
Express churn as an event filter so two people compute it the same way.
Turn a percentage into names
Read the lifecycle split and get the accounts behind the number.
Compare intake months
Split the chart and see which intake month leaves fastest.
Best for
A churn number nobody can explain
One percentage tells you nothing. A saved list of the people who left, with their recordings attached, tells you where they stopped.
Products with irregular usage
Writing churn as your own event filter beats a vendor's built-in definition of inactive.
Before you build win-back
Watching five churned sessions usually changes what the win-back email says.