App campaigns are the most radical automation in the Google Ads catalogue — more so than Performance Max. In PMax you at least choose audience signals and write recognisable ads; here you hand over raw material and a goal, and the system does the rest. Understanding what it does inside is what separates operating them from suffering them.
The five surfaces
With a single set of assets, your ads can appear on:
- Google Search: queries related to your category (you don't choose keywords; the system infers them from your app and category).
- Google Play: search and browsing inside the store — the surface with the highest intent to install there is.
- YouTube: video territory, where much of the volume lives (which is why video is not optional).
- The Display Network: inside other apps and websites.
- Discover: the discovery feed.
Each surface asks for a different format — hence the insistence on having assets of every type and aspect ratio (lesson 6).
The three campaign types
| Type | What for | Requirement and nuance |
|---|---|---|
| Installs (ACi) | New downloads and, optionally, optimising for in-app actions | The starting point for almost everyone |
| Engagement (ACe) | Bringing back users who ALREADY have the app, towards specific actions | Needs a meaningful installed base — the public reference is around 50,000 installs — and deep links configured |
| Pre-registration (ACpre) | Building anticipation before launching on Google Play | Android only, and only before publishing |
Mixing goals in one campaign is the most repeated mistake: if you ask for installs and actions at once, the system half-optimises for both. One goal per campaign.
What you actually control
You have five real levers, and none of them is a keyword:
- The conversion goal (install, in-app action, value) — what you tell it "winning" means.
- The target bid and the budget — how much and at what price (lesson 5).
- The creative assets — text, images, video and HTML5 for games: the real targeting (lesson 6).
- Geography and language — the market.
- The store listing, indirectly: it isn't in Google Ads, but it decides how many of your clicks become installs (lesson 7).
And what you don't control: specific placements, queries, granular devices, dayparting, or which combination of assets gets assembled in each impression. Arguing with that is wasted time; the work is in the five levers above.
Why the system needs so much room
The logic is the same you studied in the Smart Bidding module: a probabilistic engine crossing signals to predict who will install — and, if you ask it to, who will purchase afterwards. For that prediction to be good it needs volume of events and variety of assets to test combinations with. A short budget and three creatives is asking it to shoot blindfolded.
💡 Ninja trick: think of an app campaign as a Performance Max campaign whose landing page is run by another company (the store) and whose conversions happen somewhere Google cannot look (inside your app). Everything odd you will see is explained by those two facts.
⚠️ Pitfall: creating the campaign "because it's quick" and leaving measurement for later. In apps, measurement is not an extra: it is the product the algorithm consumes. The next lesson is the most important in the module for that reason.
What you should remember
- One asset set, five surfaces, different formats.
- Three types: installs, engagement (installed base + deep links) and pre-registration (Android, pre-launch).
- One goal per campaign: mixing is the most common way to ruin one.
- Your levers: goal, bid/budget, creatives, geography and — indirectly — the store listing.