What Are GA4 Predictive Audiences? A Plain English Guide

GA4 can predict which visitors are likely to buy in the next week. It can also predict who’s about to stop visiting your site entirely.

This isn’t speculation—it’s machine learning applied to your actual user data. And as of December 2025, predictive audiences are available to all advertisers, not just large accounts.

Here’s what they are, how they work, and whether your site can use them.

What Are Predictive Audiences?

Predictive audiences use machine learning to identify users based on what they’ll probably do next, not just what they’ve already done.

Traditional audiences look backwards: “People who purchased in the last 30 days” or “Users who viewed the pricing page.”

Predictive audiences look forward: “People likely to purchase in the next 7 days” or “Users likely to stop visiting.”

GA4 analyses patterns in your existing data—how users behave before converting, what signals indicate someone is leaving—and applies those patterns to current users.

The Three Predictive Metrics

GA4 offers three predictive capabilities:

1. Purchase probability

The likelihood that a user who was active in the past 28 days will make a purchase in the next 7 days.

Use this to identify high-intent users who might just need a nudge—a reminder email, a limited-time offer, or a retargeting ad.

2. Churn probability

The likelihood that a user who was active in the past 7 days will not return in the next 7 days.

Use this to catch users before they leave. A well-timed win-back message to someone showing churn signals is more effective than trying to re-engage someone who’s already gone.

3. Predicted revenue

The expected revenue from a user over the next 28 days.

Use this to prioritise your highest-value prospects for premium treatment—better ad placements, dedicated support, exclusive offers.

How Does GA4 Predict This?

GA4’s machine learning models analyse your historical data to find patterns.

For purchase prediction, it looks at users who bought and users who didn’t, then identifies what behaviours preceded each outcome. Maybe users who view 5+ products and visit the FAQ page are more likely to convert. The model learns these patterns.

For churn prediction, it examines users who returned versus those who didn’t. Perhaps users who only visited once and spent under 30 seconds are high churn risks.

The model applies these learned patterns to current users, scoring each one with a probability.

You don’t configure the model. GA4 builds it automatically from your data.

Requirements to Use Predictive Audiences

Here’s where many sites get disappointed. Predictive audiences require significant data volume:

Minimum thresholds:

For purchase probability:

• At least 1,000 users who purchased in a 7-day period
• At least 1,000 users who didn’t purchase in that same period
• This pattern must hold over 28 days

For churn probability:

• At least 1,000 users who returned (didn’t churn) in a 7-day period
• At least 1,000 users who didn’t return in that same period
• Again, sustained over 28 days

Event tracking requirements:

• Purchase events must be configured correctly (purchase or ecommerce_purchase)
• Event data must flow consistently for at least 28 days

Model quality:

Even with enough data, GA4 may not generate predictions if the model quality is too low. This happens when user behaviour is too random to predict reliably.

Checking If Your Site Qualifies

To see if predictive audiences are available:

1. Go to GA4 → Admin → Audiences
2. Click “New audience”
3. Look for “Predictive” in the audience templates

If predictive options appear, you meet the requirements. If they’re greyed out or missing, you don’t have enough data yet.

You can also check the model quality:

1. Go to Admin → Predictive metrics
2. Review the status for each metric

GA4 will tell you if the model is active, training, or unavailable due to insufficient data.

How to Create a Predictive Audience

If your site qualifies:

Step 1: Navigate to Audiences

In GA4, go to Admin → Audiences.

Step 2: Create new audience

Click “New audience” and look for the predictive templates.

Step 3: Choose your template

Available templates include:

• Likely 7-day purchasers
• Likely first-time 7-day purchasers
• Predicted 28-day top spenders
• Likely 7-day churning users
• Likely 7-day churning purchasers

Step 4: Customise if needed

You can adjust thresholds. For example, target only users in the top 10% purchase probability rather than anyone above 50%.

Step 5: Save and export

Name your audience and enable Google Ads export if you want to use it for targeting.

Practical Ways to Use Predictive Audiences

Likely to purchase:

• Serve targeted ads with urgency messaging (“limited stock”)
• Send email reminders about items they viewed
• Offer a small incentive to close the deal
• Prioritise for sales team outreach (B2B)

Likely to churn:

• Trigger re-engagement email sequences
• Show “we miss you” display ads
• Offer loyalty points or discounts to return
• Identify common churn causes to fix product issues

High predicted revenue:

• VIP treatment and early access
• Higher ad bids (they’re worth more)
• Personal outreach from account managers
• Exclusive loyalty tier placement

Limitations to Understand

Predictions aren’t guarantees.
A user with 80% purchase probability still might not buy. Use predictions to prioritise, not to assume.

The 7-day window is fixed.
You can’t predict 30-day behaviour. If your sales cycle is longer, these predictions are less directly useful.

Small sites may never qualify.
If you don’t have 1,000+ conversions per week, predictive audiences won’t work. Focus on growing traffic and conversions first.

Model accuracy varies.
Some businesses have predictable customer behaviour. Others don’t. If your users act randomly, the model won’t help much.

Building Toward Eligibility

If you don’t currently qualify, here’s how to get there:

Fix your e-commerce tracking.
Make sure purchase events fire correctly with accurate product and revenue data.

Grow your traffic and conversions.
More users means more data for the model. Focus on acquisition and conversion rate.

Wait.
GA4 needs 28 days of consistent data. If you just fixed tracking issues, give it time.

Monitor the status.
Check Admin → Predictive metrics periodically to see if you’ve crossed the threshold.

The Bottom Line

Predictive audiences are powerful when they work. Knowing who’s likely to buy or churn lets you act before the outcome is decided.

But they’re not available to everyone. The data requirements are real, and many small to medium sites won’t qualify.

If you do qualify, use them. Export to Google Ads, trigger automated campaigns, and treat high-probability users differently.

If you don’t qualify yet, focus on accurate tracking and sustainable growth. The feature will be there when your data catches up.

GA4 predictive audiences are available to all advertisers meeting the data requirements as of December 2025.