Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Analytics › Google Analytics 4

GA4 Audiences

How GA4 audiences work: conditions and sequences, membership duration, audience triggers, predictive, sharing to Google Ads, and audiences vs segments.
Guideintermediate9 min readUpdated

Audiences vs segments

The two features look alike in the builder and behave differently in every way that matters. A segment is an analysis construct inside an exploration: it is evaluated retroactively against the data within the retention window, it exists only in that exploration unless saved, and it can be defined at user, session or event scope. An audience is a property-level object: it is evaluated as events arrive, so it is populated from the date of creation onward; it is always a list of users; it appears in the reports as a dimension and a comparison; it can generate events; and it is exported automatically to linked advertising products. Segments answer "how did people like this behave?"; audiences answer "who are the people like this right now, and what shall we do about them?"

The practical workflow follows from the difference. Define and test the conditions as a segment first, where the result is immediate and reversible, check the size and the behaviour of the group, and only then create the audience — the segment builder offers to publish one directly. Creating audiences speculatively is a poor habit, because each one counts against the property's documented limit, each one begins accumulating users on creation and cannot be back-filled later, and the list of audiences in a property is one of the few things that cannot be tidied retroactively without losing what was collected. The explorations tutorial covers the segment side; this guide picks up once the group has been identified.

Conditions, sequences and scope

An audience is defined by conditions on dimensions, metrics and events, combined with and/or logic, with optional exclusion groups. The builder presents three ways to express membership. Include conditions describe what a user must have done or be: an event with parameters, a dimension value such as a country or a device, a metric above a threshold such as sessions or revenue. Sequence conditions describe an order: step one followed by step two, either directly or with other activity in between, optionally within a time window — a product view followed by an add-to-basket but never a purchase, for instance. Exclude conditions remove users who match them, either temporarily, for as long as they match, or permanently, once they have ever matched; the permanent form is what keeps buyers out of a prospecting audience.

Scope decides the window within which conditions must be satisfied together. Across all sessions means a user who did one thing in March and another in June qualifies; within the same session means both must occur in one visit; within the same event means both must be true of one event, which is how an audience of people who purchased a specific category is built, since the category and the purchase are parameters of the same event. Choosing the wrong scope is the most common definition error: an audience of users who viewed a product and reached checkout, scoped across all sessions, includes people who did the two things months apart and means something quite different from the same conditions within one session. Audiences built on custom dimensions need those dimensions registered before the audience is created; the events and parameters guide covers registration.

Membership duration and backfill

Every audience has a membership duration: the number of days a user remains in the list after qualifying, up to a documented maximum, or set to the maximum for an audience that should keep people as long as the product allows. A user who qualifies again while a member has the membership extended for another duration, so an audience of recent buyers with a short duration naturally rolls people off as they stop buying, while a long duration turns the same conditions into an all-time list. Match the duration to the decision: a basket-abandoner list for a reminder campaign wants days, a lapsed-customer list for a win-back sequence wants months, and an audience used as a comparison in reports wants whatever makes the comparison meaningful.

Backfill is where expectations go wrong. Audiences are not retroactive: the property evaluates incoming data against the conditions from the moment the audience is created, and users who matched before that moment are not added on the strength of their history. Google's documentation describes the limited circumstances in which a new audience is pre-populated, and the detail has changed over time, so the help page in the sources is the authority on what, if anything, is carried in. In practice, plan as though a new audience starts empty: create it ahead of the campaign that needs it, allow time for it to reach a usable size, and compare its growth with a retroactive segment of the same conditions to confirm the definition is doing what was intended rather than quietly matching nobody.

Audience triggers (events from audiences)

An audience trigger logs an event when a user joins the audience, and optionally again whenever their membership refreshes. The event is defined in the audience builder, carries the name you give it, and from that point behaves like any other event in the property: it appears in reports, can be used as a funnel step, and — the usual reason for creating one — can be marked as a key event. This turns a pattern of behaviour into a countable outcome without any change to the tag. An audience of users who viewed three or more product detail pages in a session, with a trigger named engaged_browser, produces an event whenever someone crosses that line; marking it as a key event under the rules in the key events guide gives a campaign that cannot yet accumulate purchases an earlier signal to report on and, through the Google Ads import, to bid towards.

The limits are the audience's own. The trigger fires on joining, so nothing is logged for users who were already members and the event is never retroactive. It fires at most once per membership unless the refresh option is on, and the refresh option generates an event every time the user re-qualifies, which can be far more often than intended. And the event's count is only as good as the audience definition, so a loosely defined audience produces a loosely defined key event; the counting method and the value need setting as for any other.

Predictive audiences and eligibility

GA4 can compute predictive metrics for a property — the likelihood that a user will purchase, the likelihood that an active user will stop engaging, and the revenue a user is expected to generate over a coming period — and offer audiences built on them, such as likely purchasers, likely churners and predicted top spenders, as templates in the audience builder. These are useful for retention and acquisition work because they describe users by expected behaviour rather than by what they have already done, which is exactly the gap that rule-based audiences leave.

Eligibility is strict and documented. The property has to send the purchase event for purchase-related predictions, has to have produced enough users who did and did not take the predicted action over a recent window for the model to train, and has to keep producing them for the prediction to remain available; the help page in the sources states the thresholds. The audience builder shows whether the property is eligible, and a predictive audience that was available can become unavailable if the data thins. Treat predictive audiences as a privilege of properties with steady purchase volume, test them against rule-based audiences before trusting them with budget, and remember that the model's definition of a likely purchaser is Google's, not the business's; what it means for conversion rate (CVR) and value has to be measured, not assumed.

Sharing audiences to Google Ads

With the property linked to a Google Ads account and personalised advertising enabled in the link — the Google Ads integration tutorial covers both — every audience in the property is exported automatically and appears in the Ads account's audience manager among the account's own data segments. From there it can be applied as targeting or observation to search campaigns, as targeting to display, video and discovery-style campaigns, and as a signal to campaign types that only accept signals. The display targeting guide describes how these lists combine with the other targeting layers, and the author's remarketing expertise page shows how audiences from analytics fit a wider Remarketing programme.

Several conditions govern whether the shared list actually serves. Each network enforces a documented minimum list size and an audience below it shows as too small. Users only enter the exported list when the consent signals for advertising allow it, so a site whose banner withholds advertising consent for a large share of visitors will export a much smaller audience than the GA4 count suggests. Google signals extends the list to signed-in users across devices for those who have allowed ads personalisation. And the audience in Ads is a copy that follows the GA4 definition and duration; archiving the audience in GA4 stops it being updated. A useful habit is to map audiences to journey stages — the See-Think-Do-Care framing works well — so that each one has a named purpose in a campaign rather than existing because it could be built, and to check in the Attribution reports whether campaigns using the audience actually contributed to outcomes rather than merely reaching people who would have converted anyway.

Illustrative example — not a real result

A sportswear retailer's audience set

A sportswear retailer builds four audiences for the autumn season. Basket abandoners: users with add_to_cart not followed by purchase within the same session, membership of 7 days, permanently excluding anyone who later purchases. Running category browsers: users with at least three view_item events where the item category equals running, across all sessions, membership of 30 days, with an audience trigger named running_intent that is marked as a key event. Lapsed buyers: users whose last purchase was more than 90 days ago, membership set to the maximum. And the predictive likely purchasers template, which the property is eligible for because it records roughly 2,400 purchases a month. After two weeks the abandoner audience holds 3,150 users in GA4 and 1,980 in Google Ads, the gap explained by advertising consent being refused by about a third of visitors. The running audience's trigger has produced 5,600 running_intent key events, which the paid social team uses as an early signal while the category campaign accumulates purchases.

When an audience stays empty or cannot serve in Google Ads, the diagnosis page for a GA4 audience that is not populating tests each condition against real events.

Common mistakes

  • Creating audiences before testing them as segments. The definition is wrong, the audience has started accumulating, and the slot and the history are both spent.
  • Wrong scope. Conditions meant to describe one visit evaluated across all sessions, producing a list of unrelated behaviours.
  • Expecting history. A campaign planned around an audience created the same week.
  • No permanent exclusion for buyers. Prospecting lists that keep advertising to people who already purchased.
  • Membership duration by default. Abandoner lists that hold people for weeks, lapsed lists that drop them after days.
  • Refresh events on triggers. A trigger that logs on every re-qualification, turned into a key event, flooding the count.
  • Reading GA4 size as reach. The exported list is smaller after consent and minimum-size rules, and the gap is not a bug.
  • Audiences without a purpose. Dozens of lists with no campaign attached, counting against the limit and confusing the next person.

The GA4 hub places audiences in the wider model, and the explorations tutorial is where the groups worth turning into audiences are found.

Frequently asked questions

Why is a new audience empty or much smaller than expected?

Audiences accumulate members from the moment they are created rather than being filled from history, so a list defined today starts near zero and grows as users match the conditions. Any backfill Google applies is limited and documented on the audience help page, and it should not be relied on for planning. Give the audience time, and check it against a retroactive segment of the same conditions to confirm the definition is right.

Can a GA4 audience be used in Google Ads search campaigns?

Yes, once the accounts are linked with personalised advertising enabled, the audience appears in the Ads account's audience manager and can be applied to search campaigns as an observation or a targeting layer, as well as to display, video and other campaign types. Each network applies its own documented minimum list size before serving, so a small audience will show as too small on some and not others.

Can I edit an audience's conditions after creating it?

The name and description can be changed at any time, but the conditions of a created audience are fixed in the current interface, so a changed definition means building a replacement and archiving the old one. Treat the first version as a draft: test the conditions as a segment in an exploration, check the size, and only then create the audience that will be shared with advertising.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. [GA4] Audiences — Google Analytics Help
  2. Google Analytics Help — Google Analytics Help

Related