Predictive metrics and audiences

GA4 trains models on your own property's data to compute three predictive metrics per user:

• Purchase probability: chance of a purchase (or in_app_purchase) event in the next 7 days, for users active in the last 28 days.
• Churn probability: chance an active user will not be active in the next 7 days.
• Predicted revenue: expected revenue from purchases in the next 28 days from users active in the last 28 days.

Eligibility is strict. The property needs the purchase (or in_app_purchase) event, at least 1,000 returning users who triggered it and 1,000 who did not over a 7-day window in the last 28 days, and the model must keep passing quality checks; otherwise the metrics are unavailable and Admin > Audiences shows the predictive templates as ineligible. Churn probability needs only the activity data, so it is often available before purchase probability.

You use them in two places:

• Predictive audiences (Admin > Audiences > Suggested > Predictive): Likely 7-day purchasers, Likely 7-day churning users, Predicted 28-day top spenders, Likely first-time 7-day purchasers, Likely 7-day churning purchasers. Each is a prebuilt condition on a percentile of the metric (top 10% by default) that you can edit before saving.
• As conditions in your own audiences and in Explore user segments: Purchase probability > 80th percentile; and as metrics in User lifetime (Purchase probability, Churn probability, Predicted revenue).

Typical activation: export Likely 7-day purchasers to Google Ads with higher bids, show a retention offer to Likely churning purchasers, or exclude Predicted top spenders from discount campaigns.

The metrics refresh daily and are modelled, so treat them as rankings rather than precise probabilities, and expect them to disappear if traffic drops below the thresholds.