Skip to main content

Retention

Overview​

A Goal Retention metric measures how many users return to complete a goal again after a defined period of time. It tells you whether users continue performing the same action after a delay, rather than only measuring their initial behavior.

You can track retention either:

  • from the moment a user is first exposed to the experiment, or
  • from the moment the user first completes the goal itself

This makes it possible to measure longer-term engagement and whether your experiment influences users’ likelihood to come back and repeat an action.

Examples​

context.track("purchase", {
  price: 1000,
  order_number: "0000982532",
  product_id: "ABC123",
  category_id: "ZYZ123"
});

Imagine you want to measure how many users who first purchased make another purchase within 7 days.

You can create a Purchase Retention (7 days) metric by:

  • Selecting the purchase goal
  • Setting the Retention Period to 7 days
  • Choosing whether the retention window starts from:
    • the user's first exposure, or
    • the user's first purchase

Then the metric counts users who satisfy:

  • they completed the initial purchase
  • and they completed another purchase after 7 days, within the configured retention window
caution

If your metric uses CUPED, the lookback window must be equal or larger than the retention period.

More examples

  • Checkout Recovery (24 hours): Users who returned to complete a purchase within 24 hours after first adding to cart.

  • Content Return (3 days): Users who read an article again within 3 days after their first reading.

  • Subscription Renewal (30 days): Users who returned to perform a renewal action after a 30-day cycle

Good to know​

  • Great for measuring long-term impact rather than immediate conversions.
  • Helps identify behaviors such as repeat purchases, content revisits, subscription renewals, or delayed engagement.
  • The metric counts users, not events — a user either meets retention criteria or does not.
  • Filters on the goal event apply before evaluating retention (for example: “retention among users who purchased category ZYZ123”).
  • Retention metrics are often more stable and less noisy than total repeated activations, since each user contributes at most once.
  • Changing the retention period alters the meaning of the metric and will requires a new version.
  • Useful for experiments where the impact is delayed: onboarding experiences, notifications, emails, recommendations, reminders, etc.