Building GA4 audiences for remarketing that actually convert
GA4 audiences feed Google Ads remarketing — but a badly-defined audience wastes spend on people who already bought or never cared. Here is how to build ones worth targeting.
The audiences that convert are built on behaviour that signals intent, not on demographics or blanket "all visitors" definitions. GA4 audiences update dynamically, export to Google Ads, and can be defined on any event or parameter you collect — which means the quality of your audiences is capped by the quality of your event tracking. Vague events produce vague audiences, and vague audiences waste budget.
How GA4 audiences work
An audience is a rule evaluated against your event data. Users matching the rule join; users who stop matching can leave, depending on how you define membership duration.
Three properties worth knowing before you build any:
- They are not retroactive by default, with a limited lookback. Create an audience today and it does not immediately contain everyone who qualified last quarter.
- They take time to populate and to sync to Google Ads.
- They have a minimum size before they can be used for advertising, which small sites hit constantly.
That third point shapes strategy: on a low-traffic site, three narrow audiences may each be too small to use, while one moderately broad one works.
The audiences worth building
Intent-based, in order of value
Cart abandoners. Fired add_to_cart or begin_checkout, did not fire purchase, within a short window. The highest-value remarketing audience on almost every ecommerce site, and the window matters — seven days usually outperforms thirty, because intent decays.
Checkout abandoners specifically. Reached begin_checkout without purchasing. Smaller and hotter than cart abandoners; worth separating if volume allows.
Product viewers who did not add to cart. Broader, cooler, useful for prospecting-adjacent campaigns.
Pricing page visitors on a B2B or SaaS site. Usually the single strongest intent signal available, and frequently the audience nobody has built.
Value-based
High-value purchasers. Purchase with a value above a threshold. Use as a seed for similar-audience targeting and to exclude from discount campaigns.
Repeat purchasers. More than one purchase in a period — the base for loyalty and cross-sell.
Recent purchasers, for exclusion. Just as important as targeting: excluding people who bought yesterday from acquisition campaigns stops you paying to advertise to existing customers.
Engagement-based
Deeply engaged non-converters. Multiple sessions, meaningful scroll or time, no conversion. Warm, and often responsive to a different offer rather than more of the same one.
Lapsed users. Previously active, nothing recent. Reactivation rather than acquisition.
| Audience | Signal strength | Typical window |
|---|---|---|
| Checkout abandoners | Very high | 3-7 days |
| Cart abandoners | High | 7-14 days |
| Pricing page viewers | High (B2B) | 14-30 days |
| Product viewers | Medium | 14-30 days |
| Recent purchasers (exclude) | n/a | 30-90 days |
| Lapsed users | Low | 90+ days |
Why demographic audiences are a poor default
Age and gender targeting is intuitive and weak. It describes who someone is rather than what they did, and behavioural signals outperform it consistently.
It has also become less available: with Google Signals no longer contributing demographic data, coverage has thinned considerably. What changed covers the practical impact — if you had demographic audiences, they are now smaller and less reliable than when you built them.
Behavioural audiences are unaffected, entirely yours, and better anyway.
Membership duration is a real decision
The default is often too long. An audience of cart abandoners with a 540-day membership contains people who browsed eighteen months ago and have no current intent — you are paying to advertise to them at the same rate as someone who abandoned an hour ago.
Set duration to match how long the intent plausibly persists: days for checkout abandonment, weeks for product interest, months for lapsed-customer reactivation.
Audience quality depends on event quality
An audience can only be as precise as the events underneath it.
If your add_to_cart fires on cart page views rather than on the add itself, your cart abandoner audience contains people who merely looked at a cart. If purchase double-fires, your exclusion audiences are fine but your conversion counts are not. If events lack parameters, value-based audiences are impossible.
So audience work usually starts as tracking work. DataLayer design and the items array are the prerequisites for anything sophisticated.
A build order that works
Rather than creating a dozen audiences at once and watching most of them sit below the size threshold, build in this order:
- Exclusions first. Recent purchasers, existing customers, current subscribers. These are immediately useful, cost nothing to maintain, and stop the most obvious waste — paying to acquire people you already have.
- One high-intent audience. Cart or checkout abandoners for ecommerce, pricing-page visitors for B2B. Prove the mechanism works end to end, including the Ads sync, before multiplying.
- One value-based audience. High-value purchasers, usable both as a targeting seed and as a discount exclusion.
- Then segment, only where volume supports it. Splitting one working audience into four rarely survives the minimum-size threshold on a mid-sized site.
Check size before building the campaign around it. An audience that never reaches the threshold is a plan that quietly does not run, and Google Ads will not tell you loudly.
Consent and audiences
Users who decline advertising consent should not be added to advertising audiences. A correct Consent Mode v2 implementation handles this, which means audience sizes in EEA markets will be smaller than your total traffic — expected, not broken.
If an EEA audience is at zero rather than merely smaller, that is a consent bridge failure rather than a privacy success. Verify it.
FAQ
How long does it take for a GA4 audience to populate?
Typically 24-48 hours to begin filling, and longer to reach a usable size for advertising. Audiences are not meaningfully retroactive, so create them before you need them.
Why is my GA4 audience not showing in Google Ads?
Check that the accounts are linked, that the audience has met the minimum size threshold, and that enough time has passed for the sync. Small sites frequently hit the size floor rather than a technical problem.
What is the best remarketing audience for ecommerce?
Cart and checkout abandoners within a short window — usually the highest-converting audience available. Pair it with an exclusion for recent purchasers.
Should I build demographic audiences in GA4?
Generally no. Behavioural audiences outperform them, and demographic coverage has thinned considerably since Google Signals stopped contributing that data.
Do audiences respect user consent?
They should. Users who decline advertising consent should not enter advertising audiences, which makes EEA audiences smaller. An EEA audience at exactly zero indicates a broken consent implementation rather than compliance.
Audience quality starts with event quality — check yours with the free tracking audit, which inspects tags, duplicates, and consent on any URL.
See where your tracking stands
Run the same 13-check audit referenced in this post against any URL. No signup, results in seconds.