Why app installs change the setup
Other Google Ads objectives leave the advertiser with keywords, placements, ad copy and a landing page to shape. App campaigns remove nearly all of that. The advertiser gives a store listing, a set of assets, a budget, a bidding goal, locations and languages; the system decides where the ad appears and how it is put together. The app install campaigns guide compares this with other networks' app formats, and the app marketing hub places paid acquisition beside store optimisation and retention work.
What remains, then, is a narrow set of choices with outsized effect. The bidding goal defines a good user for the system; if that goal is an install, it will find people who install, a group that includes many who never return. The assets determine what the system can say and show. And the measurement chain, from SDK to conversion action, decides whether the system ever learns that some installs were worth more than others. The setup below is built around those three points, plus the store listing, which every click passes through and which the campaign cannot rewrite.
Recommended setup for app installs
| Setting | Recommended for app installs | Why it differs from other objectives | Feature page |
|---|---|---|---|
| Campaign type | App campaign for installs, plus a second App campaign for an in-app action | Apps have their own automated campaign type; there is no keyword or placement build | App install campaigns |
| Bidding goal | Installs for volume; target cost per in-app action on the activation event | The conversion happens inside the app, after the store, not on a web page | Smart Bidding strategies |
| Events | Install and a small set of early in-app events, each from one source (Firebase or a partner) | Events arrive through an SDK and postbacks rather than a site tag | Deep links and app analytics |
| Assets | Several headlines and descriptions, landscape and portrait images, short videos in multiple ratios, grouped by theme | The system assembles every ad from what it is given | App install campaigns |
| Platform split | Separate campaigns for iOS and Android | Attribution and privacy rules differ by operating system | App attribution, MMPs and SKAdNetwork |
| Store listing | Screenshots, description and preview revised alongside the campaign | The listing is the only page between ad and install | App marketing hub |
Pick the in-app event that marks a real user
The second campaign needs one in-app action to pursue, and choosing it is the core of the objective. A good candidate happens soon after install, is performed by most users who stay, and is rare among those who leave: finishing onboarding, completing a first lesson, creating a first project. A purchase or subscription is the truest signal, but it often happens too rarely and too late to train bidding. The deep links and app analytics guide covers naming and recording events so the chosen one is defined consistently. The event must also be recorded often enough for bidding to learn from it: Google documents a minimum volume of conversions for in-app action bidding, and the figure can change, so check the current number in Google Ads Help before starting the second campaign. Set the first target cost per action near what the action currently costs through the install campaign, and tighten it gradually as the Smart Bidding guide describes. The CPA calculator (CPACost of each conversion.) converts the action's cost back into a cost per retained user that can be weighed against lifetime value.
Supply assets rather than ads
Because the system builds the ad, the advertiser's creative work is about variety and clarity in the raw material. Provide headlines that each state a different reason to install, descriptions that do not depend on one another, images in the ratios the help centre lists (typically landscape at about 1.91:1, square at 1:1 and portrait at 4:5; check Google Ads Help for the current sizes and asset limits), and short videos, ideally in landscape, portrait and square cuts, because video placements are a large part of where the campaign can appear. Group assets by theme where the campaign allows several groups, so that results by theme can be compared. Asset performance ratings then show which pieces the system favours; replace the weakest rather than rebuilding everything at once.
Work on the store listing as the landing page
Every App campaign click lands on a store page the advertiser controls only through the store's own console. Screenshots, the short description, ratings and the preview video decide whether a tap becomes an install, and the campaign cannot compensate for a listing that does not explain the app. The Play Console documents store listing experiments for Android and Apple offers product page optimisation on iOS; use them to improve the listing alongside the campaign. A lift in listing conversion lowers cost per installAverage ad cost of each app install. without any change in Google Ads.
Run iOS and Android as separate campaigns
The two stores measure differently. On Android, install referrer data and the SDK report installs and events in detail. On iOS, where the person has not allowed tracking, Apple's SKAdNetwork framework returns delayed, aggregated postbacks with a limited conversion value, as the attribution and SKAdNetwork guide explains. One campaign covering both would blend a rich signal with a thin one and set a single target across them. Separate campaigns allow different events, different targets and an honest read of each.
Keep the install campaign alive
When the in-app action campaign starts performing, it is tempting to switch the install campaign off. That usually narrows the funnel: the install campaign brings in a broad population, some of whom go on to perform the action, and its reach also feeds the data the action campaign learns from. Keep both, give the action campaign the larger share of budget as it proves itself, and use the marketing budget calculator to plan the split against the overall acquisition cost (CACCost to acquire one new customer.) the business can carry.
A worked app-install setup
Parlando, an invented language-learning app
Parlando launches on Android and iOS with one App campaign bidding on installs, a £150 daily budget and a target cost per install of £1.20. After four weeks it has 3,900 installs at £1.08 each, but only 22% of new users complete a first lesson and day-seven retention is 9%.
| Setting | Before | After |
|---|---|---|
| Campaign type | One App campaign, installs | Install campaign kept; second App campaign for an in-app action |
| Bidding goal | Target cost per install £1.20 | Install campaign unchanged; action campaign target £4.50 per first lesson completed |
| Events | Install only, from Firebase | Install, first lesson completed, day-three return; one source each |
| Assets | Five headlines, two images, no video | Ten headlines in three themes, portrait and landscape video, six images |
| Platform split | Android and iOS together | Separate campaigns per operating system |
| Store listing | Screenshots of menus | Screenshots of a lesson in progress; tested in Play Console |
Six weeks later budget is £60 a day on installs and £120 on the action campaign. The action campaign's cost per install is higher at £2.40, but 61% of its users complete a first lesson, so the cost per first lesson falls from about £4.90 across the old campaign to £3.90. Day-seven retention among those users is 24%.
Install cost more than doubled in the new campaign, and the business still paid less for each person who actually learned something, which is the comparison that matters. The example does not report results by operating system, but separate campaigns allow for it: because iOS postbacks are delayed and coarse, an iOS campaign may need a looser target to keep delivering, which a single blended campaign would hide.
Measuring app installs in Google Ads
Google Ads labels the core figures plainly. Installs and Cost / install show volume and price at the top of the funnel. In-app actions and Cost / in-app action show the same for the event the second campaign pursues, and the conversion column settings decide which actions appear there. Where events carry revenue, conversion value for in-app events supports a return-based view and, eventually, value bidding. Asset performance ratings show which assets the system uses and how they compare, the nearest thing the campaign offers to an ad report. Outside Google Ads, store listing conversion from the Play Console, and its equivalent in App Store Connect, shows how many listing visitors installed.
The business measures sit beside these: retention by install cohort, cost per retained user and payback against lifetime value (LTVTotal value a customer brings over time.). These come from the app's analytics or the measurement partner, not from the ad account. Review event delivery and asset ratings weekly, cost per in-app action and cohort retention monthly, and the event choice itself each quarter. The KPI tree for app installs shows how these measures decompose into the inputs each team owns, so a KPIA metric tied directly to an objective. chosen for the campaign links to the outcome the app needs.
Where the setup goes wrong
Four problems recur:
- Changing the bidding event and the target together. This resets what the system has learned and makes the result unreadable.
- Counting the same event from both Firebase and a partner. This doubles conversions and misleads bidding. The app install campaign launch checklist links each network to one measurement partner and tests it from a device before the campaign starts.
- Running iOS event campaigns with too little volume. They receive postbacks so coarse that targets cannot be held, and stretching them to a deep event makes it worse.
- Leaving creative unchanged for months. It loses ground as the system exhausts the combinations it can make.
Beyond Google Ads, the app installs objective hub sets out the KPI set and the channels that support an install plan, from store work to email and in-store prompts.
Common mistakes
- Bidding only on installs. Install-only campaigns find people who install, including many who never open the app again.
- Choosing a rare event. A purchase that few users make gives the system too little to learn from.
- Double-counting events. Importing the same action from two sources inflates conversions and distorts bidding.
- Ignoring the listing. A weak store page raises the cost of every install the campaign buys.
- Blending iOS and Android. One campaign across both hides the difference in measurement and in cost.
