Predictive Audiences in Google Analytics 4

Google Analytics 4 documents predictive metrics such as purchase probability, churn probability, and predicted revenue. Purchase probability concerns an active user’s chance of triggering a purchase event within a specified future window; churn probability estimates future inactivity, while predicted revenue estimates purchase-event revenue over a defined horizon. The metrics are not available for every property or every user. Google’s current documentation specifies minimum positive and negative examples over a recent period and sustained model quality; these requirements can change, so practitioners should verify current help pages. Correct event implementation matters. Purchase events should include required value and currency parameters for predicted revenue, and noisy or duplicated events can degrade predictions. Audiences apply thresholds such as a percentile, so a label like “likely purchaser” reflects a relative cutoff as well as a prediction. Some users may have no score because the model cannot calculate one. That missingness should not be interpreted as a zero probability. Marketers should use audience definitions transparently, check retention and consent settings, and avoid implying certainty to business stakeholders. Campaign performance should be measured against appropriate baselines; a predictive audience can have higher conversion rates because it selects already-likely buyers, without the advertising causing the purchases. GA4 provides audience and exploration uses, but cross-product exports and audience sizes may differ. Check eligibility, data collection, thresholds, and measurement before activation.