Google Analytics 4 (GA4) Audiences: Predictive Targeting for Ads (2026 Guide)

Old remarketing was blunt.

Someone visited your website.

You followed them around the internet.

That was the strategy.

It worked for a while.

But it was basic.

It treated a person who bounced after 5 seconds the same as a person who spent 6 minutes reading the pricing page.

It treated a blog reader the same as a checkout abandoner.

It treated a student, a buyer, a competitor and a returning customer as one group.

That is not targeting.

That is laziness with a pixel.

Google Analytics 4 gives you a better way.

GA4 audiences let you build segments based on behaviour, events, user properties and predictive metrics.

Instead of asking only:

"What did this person do?"

GA4 can help you ask:

"What is this person likely to do next?"

That is where Predictive Audiences become powerful.

A predictive audience uses a predictive metric, such as purchase probability or churn probability, to group users based on likely future behaviour.

For example:

A user has not purchased yet.

But they visited the pricing page.

They viewed several product pages.

They returned twice in 3 days.

They spent time reading delivery or returns information.

They showed the same pattern as previous buyers.

GA4 may classify them as likely to purchase in the next 7 days, if your property has enough data for predictive metrics.

That is a much stronger signal than "All Visitors".

In this "Mega-Authority" guide, we cover:

1. The Tech: How GA4 prediction works.
2. The Lists: The must-have predictive and custom audiences.
3. The Sync: Connecting GA4 to Google Ads.
4. The Strategy: Using audiences for Search, PMax and remarketing.

The goal is simple.

Stop remarketing to everyone.

Start building audiences around intent.

---

Part 1: Theory - Probabilistic Modelling

GA4 is built around events.

That matters.

In Universal Analytics, most people thought in sessions and pageviews.

In GA4, you should think in events, behaviours and users.

A user can trigger events such as:

1. Page view.
2. Scroll.
3. View product.
4. Add to cart.
5. Begin checkout.
6. Purchase.
7. Generate lead.
8. Sign up.
9. Login.
10. Watch video.
11. Download file.
12. View pricing.
13. Submit form.
14. Click phone number.
15. Start trial.

GA4 uses this behavioural data to build audiences.

Predictive metrics go further.

They use machine learning to estimate likely future behaviour.

For example:

If User A reads your blog, visits pricing, reads terms, returns twice and spends 5 minutes on the site, GA4 may identify them as high intent.

Standard remarketing may miss the nuance.

They did not add to cart.

They did not submit a form.

They did not start checkout.

But their behaviour looks like people who often convert.

Predictive audiences are designed for these cases.

They help you find users who are not obvious, but are statistically promising.

What Predictive Audiences Are Not

Predictive audiences are not magic.

They are not available to every GA4 property.

They do not work well without enough conversion volume.

They do not replace clean tracking.

They do not fix poor event setup.

They do not replace CRM quality.

They do not know your sales team’s view of lead quality.

They are useful when the underlying data is strong.

They are weak when the underlying data is weak.

Google says predictive audiences are audiences with at least one condition based on a predictive metric, such as users likely to make a purchase in the next 7 days. (Google Analytics Help)

Eligibility Matters

This is the part many guides skip.

Predictive audiences require enough data.

Google says predictive metrics need enough positive and negative examples. For purchase or churn prediction, the property needs at least 1,000 returning users who triggered the relevant predictive condition and at least 1,000 returning users who did not within a recent 7-day period over the last 28 days. Model quality also needs to be sustained over time. (Google Analytics Help)

In plain English:

Small sites may not qualify.

Low-volume lead gen sites may not qualify.

New GA4 properties may not qualify.

Poorly tracked sites may not qualify.

If you cannot access predictive audiences, do not panic.

Build custom intent audiences instead.

We will cover that below.

---

Part 2: The Predictive Audiences

You can find predictive audience templates in GA4 under:

Admin → Audiences → New Audience → Predictive

Availability depends on eligibility.

If your property does not have enough data, these options may be unavailable or limited.

1. Likely 7-Day Purchasers

This is the hot list.

It includes users who are likely to make a purchase in the next 7 days, based on purchase probability.

Use it for:

1. Remarketing.
2. Demand Gen.
3. YouTube.
4. PMax audience signals.
5. Search observation audiences.
6. Short-window promotional messaging.
7. Cart recovery support.
8. High-intent offer reminders.

Action: Treat this audience as high intent.

They may not need education.

They need reassurance, proof, urgency or a better offer.

Ad message examples:

Still Deciding? See Why Customers Choose Us

Free Delivery Ends Tonight

Compare Plans Before You Buy

Book Your Demo This Week

2. Likely First-Time 7-Day Purchasers

This is useful when you care about new customer acquisition.

A user may be likely to buy, but not yet a customer.

That matters if you want to separate acquisition from retention.

Use it for:

1. New customer campaigns.
2. Acquisition-focused PMax signals.
3. First-order incentive campaigns.
4. New buyer remarketing.
5. Excluding existing customers where needed.

This audience is useful for ecommerce and subscriptions where new customer growth matters.

3. Likely 7-Day Churning Users

This audience includes users likely not to visit your property in the next 7 days.

That can be useful for retention or re-engagement.

Use it carefully.

If a user is likely to churn because they are low quality, do not overspend trying to win them back.

If they are valuable users showing signs of fading, it can be worth action.

Use it for:

1. Win-back campaigns.
2. Re-engagement offers.
3. Product education.
4. Feature reminders.
5. Trial activation support.
6. Membership reminders.

Ad message examples:

Still Need Help Getting Started?

Your Trial Is Waiting

See What You Missed This Week

Complete Your Setup in 2 Minutes

4. Likely 7-Day Churning Purchasers

This is useful for businesses with repeat purchase or user retention needs.

It focuses on purchasing users who are likely not to return soon.

Use it for:

1. Repeat purchase campaigns.
2. Subscription recovery.
3. Loyalty campaigns.
4. Customer win-back.
5. Replenishment reminders.
6. Post-purchase education.

This is especially useful when the second purchase matters.

Examples:

1. Tea, coffee or supplements.
2. Beauty and skincare.
3. Pet food.
4. Subscription boxes.
5. SaaS trials and free users.
6. Membership products.

5. Predicted Top Spenders

Predicted top spenders are users predicted to generate higher revenue over a future period.

Use this carefully.

It is powerful when order values vary.

It is less useful when every customer spends about the same.

Use it for:

1. VIP campaigns.
2. Premium product campaigns.
3. High-margin products.
4. Loyalty offers.
5. Upsell campaigns.
6. PMax audience signals.
7. Search observation layers.
8. High-value customer lookalike-style signals.

Ad message examples:

Explore Our Premium Collection

Upgrade Your Plan

Private Demo for Growing Teams

Exclusive Range Now Available

---

Part 3: Execution - Creating Custom Segments

Predictive audiences are powerful.

But they are not the whole game.

You should also build custom GA4 audiences based on your own business logic.

This is especially important if you do not qualify for predictive metrics yet.

Recipe 1: The "Heavy Researcher"

This audience identifies users who show strong interest but have not converted.

Conditions:

session_duration > 180 seconds
AND page_view contains "pricing"
AND user has not triggered purchase or generate_lead

Use this for ads that answer objections.

Examples:

1. Trust.
2. Warranty.
3. Support.
4. Pricing clarity.
5. Case studies.
6. Comparison.
7. Delivery.
8. Security.
9. Implementation.
10. Reviews.

Ad example:

Still Comparing Options? See Our Customer Results

Recipe 2: The "Pricing Viewer"

Conditions:

page_location contains "/pricing"
AND generate_lead is not triggered

Use this for:

1. Demo retargeting.
2. Pricing objections.
3. Comparison ads.
4. Sales consultation ads.
5. Trial reminders.

Message:

Need Help Choosing a Plan?

Recipe 3: The "Cart Abandoner"

Conditions:

add_to_cart
AND purchase not triggered

Use this for ecommerce.

Create separate windows:

1. Cart abandoner 1 day.
2. Cart abandoner 7 days.
3. Cart abandoner 30 days.

Do not treat them the same.

A 1-day abandoner is hot.

A 30-day abandoner may need a stronger reason to return.

Recipe 4: The "Trial Started, Not Activated"

For SaaS.

Conditions:

sign_up triggered
AND activation_event not triggered

Activation event examples:

1. Project created.
2. Team invited.
3. Integration connected.
4. Code installed.
5. First report generated.
6. First campaign launched.
7. First file uploaded.

Use ads to guide them back to value.

Message:

Finish Setup in 3 Minutes

Recipe 5: The "Lead Submitted, Not Qualified"

For B2B.

This may require CRM integration or imported events.

Conditions:

generate_lead triggered
AND qualified_lead not triggered

Use carefully.

You may not want to spend more on weak leads.

But if they need education, you can retarget with case studies or proof.

Recipe 6: The "High-Value Category Viewer"

For ecommerce or content-led sites.

Conditions:

view_item category = "premium"
AND purchase not triggered

Use this for premium range messaging.

Do not send discount-heavy ads immediately.

A premium viewer may need trust, not cheapness.

---

Part 4: The Sync Mechanism

GA4 audiences do little for Google Ads unless they are available inside Google Ads.

You need to link the accounts.

Typical setup:

1. Go to GA4 Admin.
2. Go to Product Links.
3. Select Google Ads Links.
4. Link the correct Google Ads account.
5. Enable personalised advertising where appropriate and permitted.
6. Save.
7. Wait for audiences to populate.
8. Check Google Ads Audience Manager.

Google says Analytics can backfill audiences with up to 30 days of data where available, audiences may populate within 24 to 48 hours and daily event processing takes 24 hours. (Google Business)

Important: Consent and Personalised Advertising

Audience sharing depends on settings, consent and regional requirements.

Do not enable audience use blindly.

Check:

1. Consent Mode.
2. Cookie banner behaviour.
3. Privacy policy.
4. Personalised ads settings.
5. Regional rules.
6. Google Ads linking settings.
7. Data retention settings.
8. User deletion processes.
9. Sensitive category restrictions.
10. Internal governance.

Audiences are powerful.

But they must be used responsibly.

Where To Use GA4 Audiences in Google Ads

Use audiences in:

1. Search as Observation.
2. Search as Targeting for specific RLSA strategies.
3. Display remarketing.
4. Demand Gen campaigns.
5. YouTube campaigns.
6. PMax audience signals.
7. Exclusions.
8. Customer retention campaigns.
9. Trial activation campaigns.
10. High-value upsell campaigns.

Do not use every audience everywhere.

Match the audience to the campaign job.

---

Part 5: Summary & Checklist

If you are only using "All Visitors", your remarketing is too blunt.

All visitors are not equal.

Some are buyers.

Some are researchers.

Some are existing customers.

Some are support users.

Some are students.

Some are competitors.

Some are high-value.

Some are noise.

GA4 audiences help you separate them.

Your Action Plan:

1. Link GA4 to Google Ads.
2. Check whether predictive audiences are available.
3. Create "Likely 7-day purchasers" if eligible.
4. Create custom intent audiences such as Heavy Researcher and Pricing Viewer.
5. Apply the right audiences to the right campaigns.

Predict the future where possible.

Segment the present where necessary.

Here is the deeper checklist:

1. Audit GA4 events.
2. Confirm purchase or lead events are tracking correctly.
3. Check predictive audience eligibility.
4. Create predictive audiences where available.
5. Create custom behavioural audiences where predictive is unavailable.
6. Use separate membership durations.
7. Link GA4 and Google Ads.
8. Enable personalised advertising only where appropriate.
9. Review consent setup.
10. Use audiences as Observation in Search.
11. Use audiences as signals in PMax.
12. Use hot audiences in Demand Gen and YouTube.
13. Exclude irrelevant audiences where needed.
14. Review audience performance monthly.
15. Refresh audience logic as the business changes.

GA4 is not only a reporting tool.

It is an audience engine.

Use it like one.

---

Membership Duration — The 7-Day vs 30-Day vs 540-Day Framework

Audience membership duration is not a default setting to ignore.

It is strategy.

GA4 audience membership duration determines how long a user remains in an audience after being added. Google says the default is 30 days and the maximum is 540 days. (Google Analytics Help)

Use different windows for different intent levels.

7-Day Windows

Maximum urgency.

These users recently showed intent.

Use for:

1. Cart abandoners.
2. Pricing viewers.
3. Demo page visitors.
4. Likely purchasers.
5. Checkout starters.
6. Trial users near activation.
7. Recent product viewers.

Message should be direct.

Examples:

Complete Your Order

Book Your Demo This Week

Still Comparing Plans?

Accept higher CPAs here if quality is strong.

These users are close.

30-Day Windows

Mid-funnel nurturing.

Useful for considered purchases.

Use for:

1. B2B research cycles.
2. Product comparison journeys.
3. Category page visitors.
4. Webinar viewers.
5. Trial users.
6. Case study readers.
7. Buyers considering repeat purchase.

Message should educate and reassure.

Examples:

See How Teams Use Our Platform

Compare Features Before You Choose

Read Customer Stories

90-Day Windows

Longer consideration.

Useful for:

1. B2B enterprise.
2. Finance.
3. Education.
4. High-ticket products.
5. Property.
6. Complex purchases.

Message should be trust-led.

Examples:

See the Full Buyer Guide

Book a Consultation When Ready

Download the Implementation Checklist

540-Day Windows

Brand ecosystem.

Use for:

1. Broad remarketing pools.
2. YouTube viewers.
3. Existing customers.
4. Past purchasers.
5. Long-term nurture.
6. Seasonal retargeting.
7. Brand familiarity campaigns.

Do not bid the same way for a 540-day visitor as a 7-day pricing viewer.

Their intent is not the same.

Rule

Create separate audiences for each duration tier.

Do not lump 7-day and 540-day users together.

Their intent and messaging are fundamentally different.

Custom Dimensions — Building Audiences Beyond Behaviour

Predictive audiences are useful.

But custom dimensions and custom events can make GA4 audiences much more precise.

This is where your tracking plan matters.

SaaS Example

You fire an event:

viewed_api_docs

This tells you something important.

Users who view API docs may be developers.

They may influence the buying process.

But they may not be the economic buyer.

Audience:

Users who triggered viewed_api_docs

Strategy:

1. Exclude from "Request Demo" campaigns if they rarely book sales calls.
2. Include in "Start Free Trial" campaigns if developers prefer self-serve.
3. Show technical documentation ads.
4. Retarget with integration content.
5. Use as a signal for product-led campaigns.

This is better than treating every site visitor the same.

Ecommerce Example

You track purchase value and product category.

Audience:

Users with lifetime purchase value > £500

Use for:

1. VIP upsell campaigns.
2. Premium collection launches.
3. Loyalty campaigns.
4. Search observation layers.
5. PMax high-value customer signals.

You can also build:

Viewed premium category but did not purchase

or:

Purchased tea accessories but not premium tea

Now your ads can match real behaviour.

Not vague interest.

B2B Example

You track:

viewed_case_study
downloaded_pricing_pdf
submitted_contact_form
qualified_lead

Audience:

Viewed case study + visited pricing + no lead

This is a strong nurture audience.

Message:

See What Similar Businesses Achieved

Or:

Ready To Discuss Your Setup?

The more meaningful your events, the better your audiences.

The better your audiences, the better your campaigns.

The Audience Hierarchy

Not every audience is equally valuable.

Use this hierarchy.

Tier 1: First-Party Commercial Audiences

Highest value.

Examples:

1. Past purchasers.
2. High-LTV customers.
3. Qualified leads.
4. Trial activated users.
5. Cart abandoners.
6. Pricing viewers.
7. Demo page visitors.
8. Closed-won CRM imports.

Use these first.

Tier 2: Predictive Audiences

Strong when eligible.

Examples:

1. Likely 7-day purchasers.
2. Predicted top spenders.
3. Likely churning purchasers.
4. Likely first-time purchasers.

Use as high-intent layers and signals.

Tier 3: Behavioural Audiences

Useful when custom built.

Examples:

1. Heavy researchers.
2. Product category viewers.
3. Case study readers.
4. Blog readers by topic.
5. Video watchers.
6. Form starters.

Use to guide nurture.

Tier 4: Broad Audiences

Useful but weaker.

Examples:

1. All visitors.
2. All users.
3. All YouTube viewers.
4. All blog visitors.

Use for low-budget awareness or broad retargeting.

Do not rely on them as your main strategy.

How To Use GA4 Audiences by Campaign Type

Search Campaigns

Use audiences mostly as Observation.

This lets you see how different audiences perform without restricting search volume.

Examples:

1. Pricing viewers.
2. Cart abandoners.
3. Past purchasers.
4. Likely purchasers.
5. High-value customers.
6. Heavy researchers.

Then analyse:

1. CPA.
2. Conversion rate.
3. ROAS.
4. Lead quality.
5. Search terms.
6. Device behaviour.

Use Targeting mode only when you deliberately want an RLSA-style campaign.

Example:

Broad keyword:

crm software

Audience targeting:

pricing viewers last 30 days

This can work because you restrict a broad expensive keyword to warm users.

PMax

Use GA4 audiences as audience signals.

Remember:

Signals are not hard targeting.

They guide the system.

Good PMax signal stack:

1. High-value customer list.
2. Likely purchasers.
3. Predicted top spenders.
4. Search themes from converting Search terms.
5. Competitor URLs.
6. Category visitors.

Do not add every audience into every asset group.

Match signal to creative and product set.

Demand Gen and YouTube

These channels are excellent for nurture.

Use:

1. Likely purchasers.
2. Heavy researchers.
3. Pricing viewers.
4. Cart abandoners.
5. Trial not activated.
6. Churning users.
7. High-value customers.

Message should match stage.

Do not show generic brand videos to hot users.

Do not show aggressive discount ads to cold users with no intent.

Exclusions

Audiences are also useful for exclusions.

Exclude:

1. Existing customers from acquisition campaigns.
2. Job seekers where identifiable.
3. Support users from sales campaigns.
4. Recent purchasers from immediate acquisition ads.
5. Low-quality users if clear.
6. Employees and internal traffic where possible.
7. Free users from enterprise demo campaigns if they are poor fit.

Exclusions protect spend.

They are as important as targeting.

Common Mistakes

1. Only Using All Visitors

This is too broad.

Segment by intent.

2. Ignoring Eligibility

Predictive audiences need enough data.

If unavailable, build custom audiences instead.

3. Using One Membership Duration

A 7-day user and a 540-day user are not equal.

Split them.

4. No Consent Review

Audience activation needs proper consent and privacy handling.

Do not ignore this.

5. Sending Bad Events

Bad GA4 events create bad audiences.

Fix tracking first.

6. Treating PMax Signals Like Targeting

Audience signals guide PMax.

They do not restrict it.

7. No Creative Match

Each audience needs a suitable message.

Cart abandoners need different copy from blog readers.

8. No CRM Feedback

For lead gen, GA4 audiences alone may not know lead quality.

Import CRM stages where possible.

Final Rule

GA4 audiences are not just remarketing lists.

They are intent layers.

They help you separate:

1. Visitors from buyers.
2. Researchers from ready prospects.
3. Free users from activated users.
4. Low-value customers from high-value customers.
5. Cold audiences from hot audiences.
6. Churn risks from loyal users.

Predictive audiences make this stronger where enough data exists.

Custom audiences fill the gap where predictive data is unavailable.

The best advertisers use both.

They do not chase everyone.

They focus on the people most likely to act.

That is how GA4 audiences should work.