Most remarketing audiences are built on past behaviour: users who visited a product page, added to cart but didn’t buy, or purchased in the last 30 days. These are useful — but they’re reactive. You’re targeting users based on what they did, not what they’re likely to do next.
GA4’s predictive audiences flip this model. Instead of targeting users who added to cart, you target users who GA4’s machine learning model predicts are likely to purchase in the next 7 days — even if they haven’t added to cart yet.
This guide explains how predictive metrics work, how to build the most useful predictive audiences, and how to activate them in Google Ads.
What Are Predictive Metrics?
GA4 generates three predictive metrics using machine learning models trained on behavioural data from your property:
Purchase Probability
The probability that a user who was active in the last 28 days will complete a purchase event in the next 7 days.
Churn Probability
The probability that a user who was active in the last 7 days will not be active in the next 7 days.
Predicted Revenue
The estimated revenue a user is expected to generate in the next 28 days.
These aren’t static scores — they’re recalculated daily for each user based on their recent behaviour on your site.
Official reference: GA4 predictive metrics
Eligibility Requirements
Predictive metrics require your GA4 property to meet minimum data thresholds before they activate:
- At least 1,000 users who triggered the
purchaseevent in the past 28 days - At least 1,000 users who did not trigger
purchasein the past 28 days (for comparison) - Both groups maintained for at least 28 consecutive days
In practice: your GA4 property needs to have been tracking purchases reliably for at least a month, with enough volume for the model to find patterns.
If your property isn’t eligible yet:
- Check GA4 Admin → Audiences — if predictive audience templates aren’t visible, you haven’t met the threshold
- Focus on fixing your purchase event tracking first (volume and consistency matter more than complexity)
- Keep waiting — the model activates automatically once thresholds are met, with no manual intervention required
The Built-In Predictive Audience Templates
GA4 provides several ready-made predictive audience templates:
Likely 7-Day Purchasers
Users in the top percentile for purchase probability in the next 7 days. This is the primary audience for remarketing to warm prospects.
Best use: Remarketing campaigns targeting users who haven’t purchased yet but show high intent signals.
Predicted 28-Day Top Spenders
Users predicted to generate the most revenue in the next 28 days. Often overlaps with existing high-value customers who show repeat purchase signals.
Best use: VIP loyalty campaigns, upsell campaigns for premium products.
Likely 7-Day Churning Users
Users who were recently active but are predicted to become inactive. High churn probability = the user is likely to leave and not return.
Best use: Win-back campaigns before users go cold. Email re-engagement campaigns for returning customer segments.
Likely 7-Day Churning Purchasers
A subset of churning users who have previously made a purchase. These are lapsed customers at risk of not returning.
Best use: Post-purchase retention campaigns, loyalty programme invitations.
Building Custom Predictive Audiences
Beyond the templates, you can build custom audiences that combine predictive metrics with behavioural conditions. This is where the real power lies.
Audience 1: High-Intent Non-Buyers (Best for Google Ads)
Target users who show high purchase probability but haven’t converted yet — your warmest unconverted prospects.
Configuration:
- GA4 → Admin → Audiences → New audience → Create custom audience
- Add condition:
Purchase probabilityin top 10% (use the slider) - Add condition:
Lifetime purchases= 0 (excludes existing customers) - Audience name:
High Purchase Probability - Non Buyers - Membership duration: 7 days (refresh frequently as the model updates daily)
In Google Ads: Bid higher for this audience in your existing search campaigns using bid adjustments (+30–50%). These users are more likely to convert than general site visitors, so your ROAS improves if you allocate more spend to them.
Audience 2: High-Value Prospect Lookalike Seed
Your predicted 28-day top spenders make an excellent seed list for Google’s similar audiences / customer match expansion. Export this audience to Google Ads and use it as the basis for lookalike targeting.
Configuration:
- Predictive audience:
Predicted 28-Day Top Spenders - Export to Google Ads via the GA4 → Google Ads link
- In Google Ads → Shared Library → Audience Manager → find the imported audience
- Create a Similar Audience (or Customer Match expansion) from this seed
The insight: you’re not just remarketing to people who are likely to spend — you’re finding new users who look like your predicted high spenders.
Audience 3: At-Risk Customers for Email Suppression
Users with high churn probability who are existing customers — use this to trigger proactive retention outreach.
Configuration:
- Add condition:
Churn probabilityin top 20% - Add condition:
Lifetime purchases≥ 1 - Audience name:
At-Risk Customers - Retention
Use case: Export to your email marketing platform (via Measurement Protocol user import or GA4 → Firebase → CRM integration). Trigger a retention email sequence for users in this audience — a re-engagement offer, a survey, or a loyalty reward.
Audience 4: Predicted Revenue Tier Segmentation
Segment users by predicted revenue into tiers for differentiated marketing treatment:
- Tier A: Predicted revenue > £200 → premium service, personalised outreach
- Tier B: Predicted revenue £50–200 → standard email campaigns
- Tier C: Predicted revenue < £50 → cost-efficient channels only
Build one audience per tier using the predicted revenue metric range selector.
Activating Predictive Audiences in Google Ads
Step 1: Ensure GA4 → Google Ads Link is Active
GA4 Admin → Product Links → Google Ads Links → confirm your Google Ads account is linked with “Remarketing” enabled.
Step 2: Find the Audiences in Google Ads
Google Ads → Tools → Shared Library → Audience Manager → Your audiences tab. GA4 audiences appear here once synced (allow 24–72 hours for initial population).
Step 3: Apply to Campaigns
For Search campaigns:
- Campaign → Audiences → Add audience → select your GA4 predictive audience
- Set to Observation mode first — this lets you see how the audience performs without restricting who sees your ads
- After 2–4 weeks of data, evaluate CPCs, CVR, and ROAS for users in vs. out of the audience
- If the audience shows strong positive performance, switch to Targeting mode or apply bid adjustments
For Display and YouTube campaigns:
- Predictive audiences work particularly well for Display and YouTube remarketing
- Create dedicated campaigns targeting only your high purchase probability audience with tailored creative
- Compare CPA against your broad remarketing campaigns — predictive audiences typically show 20–40% lower CPA when properly activated
Step 4: Exclude Purchasers
Always exclude recent purchasers from your “likely to buy” remarketing campaigns to avoid wasting budget on people who’ve already converted:
- Campaign → Audiences → Exclusions → add “Purchasers (last 30 days)” audience
Measuring Predictive Audience Performance
In GA4: Audience Overlap Report
GA4 → Explore → Funnel exploration or Segment overlap technique:
- Create segments for “in predictive audience” and “completed purchase”
- Measure conversion rate within the predictive audience vs. all site users
Benchmark: If your overall conversion rate is 2%, users in the “Likely 7-Day Purchasers” audience should convert at 5–15% — if they’re not, your event tracking may have issues that are undermining the model.
In Google Ads: Audience Insights
Google Ads → Campaigns → select a campaign using predictive audiences → Insights → Audience insights
This shows conversion rate, CPC, and ROAS broken down by audience segment — crucial for deciding whether to increase or decrease bid adjustments.
Common Mistakes with Predictive Audiences
Mistake 1: Sending predictive audiences straight to Targeting mode Start with Observation. Predictive audiences can be narrow — if you put Search campaigns in Targeting mode, you may severely limit your eligible impressions. Observation mode lets you collect data without risk.
Mistake 2: Not refreshing audience membership duration Set membership duration to 7 days maximum for “Likely 7-Day Purchasers” — this matches the model’s prediction window. A 30-day membership duration means you’re including users the model predicted 3 weeks ago, whose purchase probability has almost certainly changed.
Mistake 3: Treating predictive audiences as a replacement for conversion tracking Predictive audiences improve the efficiency of spend for users who are already on your site. They don’t replace the need for proper conversion tracking, proper GTM setup, and high-quality landing pages.
Mistake 4: Using predictive audiences for brand-new properties If your property has only been tracking purchases for 2 months, the model doesn’t have enough history to make reliable predictions. Give it 3–6 months of clean data before expecting strong predictive model performance.
Predictive audiences are one of GA4’s most powerful features, and they’re underused — partly because they require solid underlying event tracking to work, and partly because the connection to Google Ads isn’t obvious. If your property has the volume to qualify, the ROAS improvement from well-configured predictive remarketing is one of the highest-return analytics investments available.
We set up and activate GA4 predictive audiences as part of our Google Analytics and Google Ads integration service. Book a free consultation to discuss your property’s eligibility and how to structure predictive remarketing campaigns.