GA4 Predictive Audiences: How to Use Them for Your Campaigns

GA4’s Machine Learning at the Service of Your Campaigns

GA4 integrates machine learning models capable of predicting your users’ future behavior. Instead of targeting audiences based on what they did (visitors from the last 30 days, cart abandoners), you can target audiences based on what they are likely to do. This is a paradigm shift for media buying.

Three predictive metrics are natively available in a properly configured GA4 setup:

• Purchase probability: likelihood that an active user will make a purchase in the next 7 days.
• Churn probability: likelihood that a user active in the last 7 days will not return in the next 7 days.
• Predicted revenue: expected revenue from an active user over the next 28 days.

Prerequisites: The 1,000 Sample Threshold

Predictive audiences are not available by default. GA4 requires a minimum volume to train its models:

• At least 1,000 users who triggered the positive condition (purchase, return) AND 1,000 users who triggered the negative condition (no purchase, no return) over a 7-day period.
• The purchase event must be properly implemented with the value parameter.
• Model quality must remain stable over an extended period. GA4 may disable predictions if quality degrades.

In practice, this means only sites with substantial traffic and regular transaction volume can benefit. An e-commerce site with fewer than 500 transactions per week will struggle to meet the thresholds.

Creating and Configuring a Predictive Audience

In GA4, go to Admin > Audiences > New audience. Predictive audiences appear in the suggested templates if your property meets the conditions:

• Likely 7-day purchasers: users with a high probability of purchasing within 7 days.
• Likely 7-day churning users: recently active users who are at risk of not returning.
• Predicted 28-day top spenders: users likely to generate the most revenue.

You can combine these predictive conditions with standard conditions. For example: “Likely purchasers” AND “source = organic” to target future organic buyers with a specific remarketing campaign.

Export to Google Ads: Targeting and Exclusion

Once created, predictive audiences sync automatically with your linked Google Ads account. Two main strategies:

Targeting likely purchasers: increase your bids or create dedicated campaigns for users GA4 identifies as close to conversion. Cost per acquisition is mechanically lower since you concentrate budget on your hottest prospects.

Excluding likely churners: remove from your remarketing campaigns users GA4 considers lost. There is no point spending budget on visitors who are unlikely to return. Reallocate that budget toward acquisition or likely purchasers.

Limitations and Practical Tips

Predictive audiences are not a silver bullet. GA4’s model remains a black box: you do not know which signals it uses or how it weights its predictions. Systematically test predictive audience performance against your manual audiences before reallocating significant budgets. Also monitor audience size: if it drops below 1,000 users in Google Ads, the targeting will be too narrow for bidding algorithms to work effectively.