Firebase (now folded into GA4) can hand Google Ads your purchasers, your churn-risk users, and your predicted 7-day buyers as ready-to-target audiences. Here's exactly how the link works, what breaks if you unlink it, and how to build audiences worth spending against.
Firebase audience targeting lets you build Search, Display, and Video campaigns in Google Ads around segments defined by actual in-app behavior — purchasers, churn-risk users, people who hit a specific event — rather than generic demographic or interest targeting. It works because Firebase (Google's mobile app SDK for analytics, crash reporting, and engagement) is linked directly to Google Ads, and today that data pipeline runs through Google Analytics 4, which is where Firebase's event and user-property data now lives for most properties.
If you run a mobile app and you're still targeting Google Ads campaigns with only demographic and affinity audiences, this is very likely the highest-leverage gap in your account. This guide covers the two built-in audience types, how to build custom segments, how predictive audiences extend the model with machine learning, and the operational trap that quietly breaks campaigns when nobody's watching the link between the two accounts.
A quick terminology note, because the source terminology has shifted and it trips people up: Firebase used to have its own separate analytics product (Firebase Analytics) with its own audience-building UI. That product's event and user-property data model was merged into Google Analytics 4 — so when Google Ads documentation refers to "Firebase audience segments" today, it's describing the same underlying data (app installs, in-app events, user properties like lifetime value or churn propensity) accessed and managed through the GA4 interface, with the Firebase SDK still doing the actual data collection inside the app.
Practically: you're not managing a separate "Firebase audiences" panel anymore. You're building these segments in GA4 (or, for a subset, letting the two default segments flow through automatically), and Google Ads consumes them from there.
Once your Firebase project is linked to Google Ads, two segments are available immediately with zero configuration:
These are blunt instruments, but they're useful as a floor. "All users" minus your existing customer list makes a reasonable exclusion audience for prospecting campaigns. "Purchasers" is a ready-made seed for a lookalike/similar-audience style expansion, or a base for a winback campaign once you layer in a recency filter.
Beyond the two defaults, you can build segments from "any combination of events and user properties" tracked in Firebase/GA4. Google's own example is a segment for "Android users who live in Brazil" — but the more useful segments for most apps combine a behavioral event with a value threshold or a time window. A few patterns worth building from day one:
add_to_cart or begin_checkout in the last 7 days but no purchase event — your single highest-ROI retargeting segment in almost every app category.purchase event more than 60/90 days ago, none since — a winback segment that performs very differently from a never-purchased segment and should never be lumped into the same campaign.Build these in GA4's audience builder using the event and user-property conditions your app already tracks (or start tracking, if it doesn't yet — the segment is only as good as the underlying event instrumentation). Once built, they become targetable in Google Ads the same way the default segments are.
The most useful evolution here isn't a segment you define manually — it's letting GA4's machine learning models define it for you based on purchase-probability patterns in your historical data. Google offers five predictive templates:
| Predictive audience | What it targets |
|---|---|
| Likely 7-day purchasers | Users predicted to buy within the next 7 days |
| Likely 7-day churning users | Users likely to stop engaging in the next 7 days |
| Likely 7-day churning purchasers | Past purchasers at risk of going inactive |
| Likely first-time 7-day purchasers | Users predicted to make their first purchase within 7 days |
| Predicted 28-day top spenders | Users predicted to be highest-value spenders over the next month |
These sync directly into Google Ads, so a "likely 7-day churning purchasers" audience can become the seed for a retention campaign without any manual list-building.
The catch is data volume. Predictive audiences require a sustained minimum of roughly 1,000 positive and 1,000 negative examples over a 28-day window before Google will activate the model — for ecommerce that means 1,000+ purchase events and 1,000+ non-purchase sessions in that window; for lead-gen apps, 1,000+ completed forms against 1,000+ started-but-abandoned forms. Below that threshold, predictive audiences simply won't populate. If you're not there yet, custom event-based segments (the "high-intent non-converters" pattern above) are the better near-term substitute — they don't need a trained model, just decent event tracking.
This is the single most important operational fact in this entire guide: if the Google Ads account gets unlinked from Firebase/GA4 — for any reason, including an accidental permissions change, an account restructuring, or someone "cleaning up" old integrations — every ad group actively targeting a Firebase-sourced segment stops running. Not degrades. Stops.
This applies to shared segments too — if your agency or a parent company manages the GA4-to-Ads link and that link breaks, every account downstream that was consuming those shared segments loses access simultaneously, often without an obvious alert pointing at the actual cause. A campaign that was delivering fine on Monday can go to zero impressions on Tuesday with no creative or bid change, and the root cause is buried in an integrations settings page nobody checks day-to-day.
The practical defense: whoever owns paid media performance should have a recurring (monthly is reasonable) check of Admin > Product Links in GA4, independent of whoever owns the analytics implementation. Don't assume the link is stable just because it was set up correctly once. Also worth noting: Firebase-created segments stay under Firebase/GA4's management — Google Ads account holders can't edit segment logic from the Ads side, so a segment definition change always has to happen upstream in GA4, which is one more reason cross-team communication between whoever owns analytics and whoever owns paid media matters more here than in most integrations.
Firebase/GA4 segments aren't a replacement for the rest of your Google Ads targeting stack — they're a layer that should sit on top of it:
| Targeting type | What it's good for | What it misses |
|---|---|---|
| Firebase/GA4 event-based segments | Precision based on actual observed behavior in your app | Cold prospects who've never touched your app |
| Predictive audiences | Forward-looking targeting without manual segment logic | Needs 1,000+ positive/negative samples to activate |
| Standard demographic/affinity | Broad top-of-funnel reach | Weak signal — proxy attributes, not behavior |
| Customer Match (uploaded lists) | CRM/email-based targeting, cross-device | Requires clean, consented first-party data outside the app |
The strongest app-marketing accounts layer these: broad or Performance Max campaigns for cold acquisition, Firebase-sourced event segments as observation layers to understand who's actually converting, predictive "likely churning purchasers" segments feeding dedicated retention campaigns, and lapsed-purchaser segments feeding winback flows — each with its own budget and message, rather than one undifferentiated remarketing campaign trying to talk to all of them at once.
Once you've got your segments defined, use a tool like the ROAS Calculator to sanity-check whether the incremental spend on a narrow predictive segment is actually clearing your target return before you scale budget into it — a smaller, higher-intent audience can look great on CPA and still be a bad allocation of budget if the absolute volume is too thin to move revenue.
Do I need to use "Firebase" specifically, or is this just GA4 now? Functionally, it's GA4. Firebase is still the SDK collecting the data inside your app, but the audience-building interface, the event/user-property model, and the predictive audience templates all live in GA4 today. Google Ads documentation still uses "Firebase audience segments" as the term for app-behavior-based audiences, which is why the terminology can feel dated.
What are the two audience segments available by default once I link Firebase to Google Ads? "All users" (everyone who has installed your app) and "Purchasers" (everyone who has made a purchase or in-app purchase). Both are available immediately on linking, with no additional configuration.
What happens if I unlink my Google Ads account from Firebase/GA4? Any ad group actively targeting a segment based on Firebase/GA4 data stops running immediately — not gradually. This applies to shared segments across multiple linked accounts as well, so unlinking can silently break campaigns in accounts you weren't even thinking about when you made the change.
How much data do I need before predictive audiences will work? Google requires roughly 1,000 positive and 1,000 negative examples sustained over a 28-day period — for example, 1,000+ purchases and 1,000+ non-purchasing sessions for an ecommerce purchase-probability audience. Below that threshold the predictive audience won't populate, and you should use manually defined event-based segments instead.
Which campaign types can target Firebase/GA4 app segments? Search, Display, and Video campaigns can target segments based on mobile app data, per Google's documentation. Performance Max campaigns can also use these audience signals as a targeting input, though PMax treats audience signals as a starting point for its own automated targeting rather than a hard restriction.
Can I edit a Firebase-created segment from within Google Ads? No. Segments created in Firebase/GA4 remain under Firebase/GA4 management. Any change to segment logic — adding a condition, adjusting a threshold — has to be made in GA4, not in the Google Ads interface, even though the segment is targetable from Ads.
MarketerTools
Marketing Practitioners
Written by the MarketerTools team — practitioners who build and use tools for marketers every day.
Put this into practice
Use the free tools on MarketerTools to apply what you just read.
Browse all tools →