Google App Campaigns for Play: Setup, Bidding and Benchmarks

Google App campaigns are the default way to buy Android installs at scale, because one campaign pushes your ads across Search, Google Play, YouTube, Discover and the Display Network without you picking a placement. This guide covers the three campaign types, the creative you must supply, how tCPI and tCPA bidding behave, how Google Play, Firebase and the Install Referrer fit together, and what installs cost in published data.

Three campaign types, one machine

You supply assets, a budget and a bid target; Google assembles the ads and chooses where they appear ³. What you control is the objective, and there are three subtypes.

App campaigns for installs (ACi) target new users. The optimisation goal can be install volume, installs by people likely to take an in-app action, the in-app action itself, or conversion value ¹.

App campaigns for engagement (ACe) target people who already have the app. Audience segments such as Customer Match or Firebase lists are required or the campaign will not serve, and deep links are recommended over the home screen ^1,2.

App campaigns for pre-registration (ACpre) are Android-only and exist to collect pre-registrations before launch. The app must be set up for pre-registration in Play Console with at least one APK on a release track, you must launch within 90 days of opening pre-registration, the campaign serves only in countries where pre-registration is enabled, and it stops serving automatically in each location once the app is released there ⁴. Pre-registration is the only conversion tracked, and it is created automatically on the first pre-registration ⁴.

If you run installs alongside engagement or pre-registration, enable attribution sharing so the campaigns do not fight over the same conversions ².

Creative assets: what you must upload

Text is the only required asset type, but Google's guidance is to give every ad group at least one approved text, image and video asset and keep Ad Strength at "Good" or better ². The current spec sheet ³:

Images are JPG or PNG under 5 MB, and text is not translated, so language targeting must match the copy ³.

Bidding: tCPI, tCPA, tROAS and tCPpre

Target cost per install (tCPI) is the ACi default. When the goal is "all users" the target is optional and bid guidance is shown; when the goal is users likely to act in-app, Google advises setting the target about 20% above your baseline tCPI ¹.

Target cost per action (tCPA) optimises to a single in-app event; Google says to pick one, because several blend into a muddy target ^1,2. Once data accumulates, set the tCPA roughly 20% above the observed CPA ¹. For engagement campaigns the target must be high enough to drive more than 100 conversions ¹.

Target return on ad spend (tROAS) needs conversion values on in-app events; start from the ROAS a comparable campaign achieved over the same window ^1,2. Bid guidance for ACi tROAS is currently Android-only ¹.

Target cost per pre-registration (tCPpre) is the only strategy for ACpre, and it is the one that carries the country-availability check ^1,4.

Two official numbers matter more than any blog advice. Budget-to-bid ratios: a tCPI or pre-registration campaign needs a daily budget of at least 50 times the bid, a tCPA install campaign at least 10 times, and a tCPA engagement campaign at least 15 times ². And the learning period is counted in conversions, not days: avoid changing conversion settings, top assets, bids, goals or geo targeting before 100 conversions have registered ². Google also recommends iOS bids at about 1.5 times the Android bid, a reminder that Android volume is cheaper ¹.

Measurement: Google Play, Firebase and the Install Referrer

Google Play does the basics for free. Google Ads records Play installs as conversions with no SDK work, and with a linked Play developer account it also records in-app purchases by users who clicked an ad, using purchase value as conversion value; for subscriptions only the first payment counts ⁵. Play measures downloads and purchases only, installs from other stores are not captured, and since April 2023 Play conversions are unavailable for apps designed only for children ⁵.

Firebase or a third-party partner handles everything else. In-app events beyond purchases, and first-open installs from other stores, require Firebase or a linked analytics provider; Google recommends Firebase for non-Play installs ⁵. Auto-tagging must be on for every campaign ², and tROAS conversion windows should be 30 days or shorter, sized to capture roughly 90% of conversions ².

The Play Install Referrer carries the click ID. When an ad sends a user to the Play Store, Google writes a gclid into the Install Referrer; your app reads it through the Install Referrer API on first open and returns it as market_referrer_gclid in a conversion request ⁶. Keep stored gclid values for 180 days or until replaced, and do not confuse the referrer gclid with one taken from a deep link ⁶. Measurement partners and the Firebase SDK read the same referrer to attribute first_open on Android unconfirmed.

Benchmarks: what an Android install costs

Google publishes no App campaign cost benchmarks, so every figure below comes from vendors with their own samples; treat them as ranges, not targets.

The most rigorous recent dataset is Liftoff's 2025 casual gaming report, built on Singular data from February 2024 to February 2025 covering 2.4 billion installs and $11.9 billion of spend: casual games averaged $0.14 per install on Android against $1.41 on iOS, and casino games reached $21.03 on iOS ⁷.

For non-gaming apps the numbers are looser. Adapty's 2026 playbook puts typical Google App campaign CPIs at $1.50 to $4.50, rising to $5 to $10 or more in gaming, finance and technology, with in-app action CPAs of $5 to $20; it gives no data period or sample size unconfirmed ⁸. Admiral Media's 2026 tables list Android tier-one ranges by vertical, but the page itself admits they carry no reporting dates or reproducible method unconfirmed ⁹.

How this shows up in the charts

Google has never published how Play's top charts are computed; the available descriptions come from ASO vendors. They broadly agree that chart position reflects install velocity over a recent window, weighted by country and category, rather than lifetime installs, and that uninstalls and rating trends can drag an app back down ¹⁰. One vendor claims an uninstall rate above the category median actively suppresses rank; that is an assertion, not a documented rule unconfirmed ¹⁰.

Because App campaigns deliver installs through Play itself, they feed the same velocity signal organic installs do, which is why a tCPI campaign scaled into a 50x budget tends to show up as a chart move within days. Cheap Android installs also mean a small budget can move a category chart in a mid-size storefront, one reason Play charts churn faster than iOS charts in the AppWinners data. The flip side is retention: a campaign optimised purely for install volume buys users who may uninstall quickly, and if the vendor accounts of Play ranking are right, that can undo the gain. Bidding on an in-app action instead of the raw install scales more slowly but tends to produce installs that hold a chart position once the budget comes off.