Predictive Audiences

Analytics

Also: Predictive Segments · ML Audiences

What it doesPredicts who will buy or churn

Powered byMachine learning on your GA4 data

NeedsSufficient conversion volume to activate

Exports toGoogle Ads and Display campaigns

Quick definition

Predictive audiences are segments built by Google Analytics 4 (GA4) using machine learning to identify users likely to take a specific action, such as making a purchase or churning within a set time window. GA4 analyses behavioural signals from your site or app and groups users by predicted probability rather than by past behaviour alone.

How it varies across Australia

Activation rates for predictive audiences in Google Analytics 4 vary sharply depending on conversion volume. Many Australian mid-market sites sit below the minimum threshold needed to unlock the feature, which means the tool is available in theory but dormant in practice.

See data and tracking maturity across Australian industries →

The three built-in predictive metrics

Purchase probability

Likelihood a user will complete a purchase event within the next seven days.

Churn probability

Likelihood an active user will not visit again within the next seven days.

Predicted revenue

Expected purchase revenue from a user in the next 28 days.

What it actually means

Most audience targeting is retrospective. You build a list of people who bought before, or who visited a specific page, and you target them again. That works, but it is inherently backward-looking. Predictive audiences point in the opposite direction.

GA4 uses machine learning to score every user in your property against three signals: purchase probability, churn probability and predicted revenue. It then lets you build audiences based on those scores. 'Top 10% most likely to purchase in the next seven days' is a very different list from 'people who visited the product page last month.'

The practical result is that you can export these segments directly into Google Ads and run campaigns against people who haven't bought yet but are statistically likely to. That affects both your bidding strategy and your creative brief. A message aimed at a high purchase-probability user should be very different from a message aimed at a cold prospecting audience.

The limitation is a hard one: GA4 requires a minimum volume of purchase events before predictive metrics activate. If your site converts infrequently, the model doesn't have enough signal to make predictions worth using. This is the most common reason Australian businesses find the feature greyed out in their GA4 property. Pairing predictive audiences with a solid conversion rate strategy and clean attribution setup is what unlocks the feature's value.

Predictive audiences shift targeting from who bought last time to who is about to buy next time. The difference matters more than it sounds.

How it shows up

Predictive audiences appear in the Audiences section of GA4 under suggested audiences once your property meets the minimum eligibility requirements. When active, you can create segments like 'likely seven-day purchasers' or 'likely seven-day churning users' and publish them directly to linked Google Ads accounts.

In Google Ads, they appear as standard remarketing lists and can be used in observation or targeting mode on Search, Display and Performance Max campaigns. The quality of the predictions is reflected indirectly in campaign performance rather than shown as a confidence score.

The Australian context

Australian ecommerce and lead-generation sites often sit close to or just below the conversion volume thresholds GA4 requires for predictive audiences. The threshold is not published precisely by Google, but commonly reported minimums suggest a property needs several hundred purchase events per week before predictions become reliable. Many Australian mid-market businesses hit this threshold seasonally, which means predictive features activate during peak periods and go dormant in the off-season.

Where people get this wrong

Assuming predictive audiences are available by default.GA4 requires a minimum conversion volume before the machine learning models have enough signal to run. Many Australian sites never reach the threshold and the feature stays inactive.

Using predictive audiences without checking attribution setup first.If your conversion events are misconfigured or duplicated, the model trains on bad data. Garbage-in predictions can actively harm campaign performance by targeting the wrong users.

Treating predicted purchase probability as a guarantee.The model outputs a probability score, not a certainty. A high purchase-probability audience still needs the right message and offer. Prediction improves targeting, it does not replace creative or conversion rate work.

Common questions

How do I activate predictive audiences in GA4?

Your GA4 property needs to meet Google's minimum conversion volume thresholds, which require a meaningful number of purchase or relevant conversion events logged per week. Once eligible, GA4 displays suggested predictive audiences in the Audiences section automatically. Link your property to Google Ads to export them.

Can predictive audiences be used on channels other than Google Ads?

Natively, GA4 predictive audiences export to Google Ads only. To use similar predictions on Meta, LinkedIn or other platforms you would need to build equivalent models using your own first-party data or a third-party customer data platform.

How is a predictive audience different from a lookalike audience?

A lookalike audience finds new users who resemble your existing customers, based on platform-side signals. A predictive audience scores your existing site visitors based on their own behaviour signals in your GA4 property. One finds new people, the other prioritises among people already in your funnel.

Does using predictive audiences affect data privacy compliance?

Yes. Predictive audiences rely on user-level behavioural data processed by Google. In Australia, this intersects with the Privacy Act and consent obligations. Confirm that your consent banner covers Google's modelling and advertising use cases, and review your data processing terms with Google.

Keep exploring

About New Rebellion

New Rebellion is a marketing intelligence consultancy. We build tools, score Australian businesses on how their marketing actually performs, and publish Debrief every day. This dictionary is part of how we work in the open.

How we think →