Google App Campaigns (still “UAC” to everyone who has run them for years) remain the highest-scale way to grow an app on Google. One campaign feeds Search, Play, YouTube, Display and AdMob. They are also still the campaign type where you control the least and your inputs matter the most.
We’ve spent millions on App campaigns as one of the first Google App Partner agencies in Europe. This guide is everything we’d tell a new client: the best practices that actually move performance, what changed recently (a lot, especially on iOS), and how to avoid getting fooled by the reporting.
The 11 best practices
1. Get your store listing right first
The algorithm reads your listing to understand who to target, and users decide on it in seconds. Put the core benefits up front, write a keyword-rich description, and have strong screenshots, a preview video and healthy ratings.
On Google Play you now get up to 50 custom store listings, and they are URL-targetable. That means you can pair campaign creative themes with matching store pages instead of sending every click to one generic listing. Use that.
2. Tracking before campaigns, no exceptions
You cannot run (or judge) App campaigns without proper tracking. Firebase is the usual route, but not the only one: a supported MMP integration works too, and Android installs track through Google Play on their own. If you use Firebase, keep the SDK current: version 11.14+ is the minimum to unlock Google’s newer iOS measurement (more on that below). Add an MMP once you’re on multiple networks. We cover the setup in more detail in our guide to tracking installs and in-app events.
If you have EEA users, Consent Mode v2 signals are mandatory. Without them, Google silently drops your conversion data.
3. Choose the right bidding strategy (tROAS works on iOS now!)
The three options are the same: installs (tCPI), in-app actions (tCPA) and in-app action value (tROAS). The 2025 news that mattered: tROAS and Maximize Conversions for in-app actions are now available on iOS globally. For years, iOS was effectively limited to installs and tCPA.
If your users’ values differ widely (they do, for almost every subscription app), value-based bidding is worth testing on both platforms now. For ad-monetized apps, there is also a dedicated Ad Revenue Optimization strategy.
4. Respect the bid-to-budget ratios
This one is ignored constantly: the budget should be around 50-100x your tCPI for install campaigns and 10-20x your tCPA for action campaigns. Starve the ratio and you get volatility, higher CPAs and an algorithm that never settles.
One important 2026 nuance. Google changed how target-based bidding works in August 2026: budget-limited campaigns are now optimized toward your set target instead of quietly beating it. App campaigns are explicitly exempt (for now). But your web2app campaigns (Search, PMax and Demand Gen driving to your funnel) are not. If you set loose targets “for safety” on those, tighten them to your 30-day actuals, or Google will happily spend up to them.
5. Structure: fewer, bigger campaigns
Run roughly two campaigns per geo group (installs and actions), with ad groups per theme, and resist the urge to fragment. Every split divides your conversion signal.
6. Optimize toward the right event
Target an event that fires 10+ times per day per campaign and sits deep enough in the funnel to mean something. Installs are volume, purchases are quality, and the right answer is usually the deepest event that still clears the volume bar. Start with installs while the algorithm learns your user shape, then graduate to action campaigns.
7. Group geos by LTV
Run separate campaigns when LTV differs materially, and group similar-LTV countries to concentrate signal. Never blend markets whose unit economics you’d never blend in a spreadsheet.
8. Localize the ads and the app
Native-language ads perform better, and localization compounds with per-market store listings (see practice 1). Service-heavy apps benefit the most.
9. Test bids gently
Keep changes under 20%, don’t touch things more than every couple of days, and give changes 5-7 days before judging. Higher bids are still what unlocks YouTube inventory, which is the growth placement. So bid testing is really inventory testing.
10. Re-engage churned users (ACe)
App Campaigns for Engagement still require deep linking and a significant install base, but the tooling got meaningfully better. Google now has dedicated deep-linking guidance and an AI agent that automates deep link implementation (“days to minutes” is the pitch, your mileage may vary). Google’s own stat: clicks landing in-app convert around 2.8x better than mobile web.
If you have the install base, re-engagement is cheaper than winning the same users twice.
11. Creatives are the whole game
The algorithm does the targeting. Your creative inventory is what you actually control.
The current asset spec: 5 headlines (30 characters), 5 descriptions (90), up to 20 images per orientation (horizontal 1.91:1, square 1:1 and vertical 4:5), and up to 20 YouTube-hosted videos in 16:9, 9:16 and 1:1 (10-60 seconds). Cover every orientation. Missing formats mean missing inventory, and if nothing gets spend, Google falls back to auto-generated store assets.
Two 2026 creative notes:
- Asset Studio replaced Ads Creative Studio. It has free generative tooling inside Google Ads (image generation and editing, auto-resizing from one master asset). Google’s AI video outpainting, which expands a landscape video to vertical, actually shipped first in App campaigns. Useful, but still no substitute for native 9:16 production.
- AI-disclosure labels are now legally required for certain AI-generated ads in some markets. The EU, India and New York are cited. Check the current policy before shipping AI-heavy batches.
Video that performs (the short version)
The ABCD model still holds. Attract (hook the pain point immediately), Brand (introduce it naturally), Connect (emotion via pacing and audio), Direct (clear action). Start fast, reach the point early, use big text and tight framing for small screens, and avoid YouTube’s dead zones (lower left, bottom corners).
Formats that keep working: app demos with voice-over, fast-paced testimonial cuts, story-driven ads and UGC. Playables (HTML5) remain install-campaign-only and carry premium CPMs, so test them with eyes open. For more on video, read how to scale App campaigns with video creatives.
iOS on Google: much better than its reputation
For years the honest advice was “iOS App campaigns work, but you’ll fly half-blind and your MMP dashboard will make Google look terrible.” That advice is expiring. What changed:
- New iOS inventory (August 2025). YouTube Shorts placements, co-branded creator partnership ads and playable end cards on AdMob. The reach story is materially bigger.
- tROAS on iOS (see practice 3).
- On-device conversion measurement (ODCM). De-identified event matching happens on the device and improves both optimization and reporting. It needs Firebase SDK 11.14+, and support expanded to the EEA, UK and Switzerland in 2026.
- Integrated Conversion Measurement (ICM). Google now feeds modeled, near-real-time attribution directly into MMP dashboards (AppsFlyer, Adjust, Singular, Branch and the rest). The chronic “Google iOS looks like it does nothing in my MMP” distortion is finally addressable. One published case showed 6x installs at 80% lower CPI after adoption. If you paused iOS spend years ago because the numbers looked bad, this is the reason to re-test.
- SKAN reporting. Google’s SKAdNetwork report now uses machine learning to fill the privacy-threshold gaps (“redistributed” conversion values). Under the hood, the ecosystem is mid-transition from SKAdNetwork to Apple’s AdAttributionKit, and Google hasn’t documented AAK support yet. Configure per your MMP’s guidance.
Some things haven’t changed. Much of the iOS data is still modeled and arrives 2-5 days late, so don’t judge anything younger than that. Keep the campaign count low to concentrate signal, and aim for roughly 40-50+ installs per day per campaign so conversion values clear Apple’s privacy thresholds. We compare the two platforms in more depth in Google App Campaigns: iOS vs. Android.
And a cheap verification trick we use often: App Store Connect’s referrer data. “App referrer” shows your YouTube traffic and “web referrer = google” shows Search traffic, straight from Apple. It’s a rough but honest second opinion on where installs actually come from.
Web2app: Google finally connected the two halves
The biggest structural news of 2025-26 isn’t inside App campaigns at all. Google’s Web to App Connect now spans Search, Shopping, PMax, Demand Gen, YouTube and Hotel campaigns. And for the first time, web campaigns can attribute app installs and first in-app conversions. “Unified conversions” bundle web and app events into one view.
For subscription apps this pairs with the web-funnel playbook (quiz, then web checkout, then app install last). You can now run that funnel on Google’s web campaign types without losing the app half of the story. Demand Gen in particular quietly became an app-relevant channel: deep linking, tCPC bidding, product feeds and AI video.
Trust, but verify: incrementality got cheap
Attribution on App campaigns will always be partly modeled, so calibrate it. The big change: Google cut the minimum incrementality experiment budget from around $100k to $5,000, with results inside the Ads UI. Verify App campaign eligibility for your account, since the rollout led with video and Demand Gen.
For bigger questions, Meridian (Google’s open-source MMM) is free. And the scrappy method still works: baseline installs per geo, make one significant budget change, watch the trend. The point isn’t methodology purity. The point is to never make scale decisions off a single dashboard.
Asset evaluation: read the ratings correctly
Google grades assets Learning / Low / Good / Best relative to your own similar assets. The decision logic we use:
- Chasing profitability? Cut low-ROAS assets and ad groups.
- Chasing scale? Keep high-spend assets even at softer ROAS.
- The jackpot: high spend + “Best” rating + acceptable CPA or ROAS. Make more like this one.
Asset ratings are directional, not gospel. Pair them with spend distribution and downstream ROAS before killing anything.
FAQ
Are Google App Campaigns still worth it in 2026?
Yes. For most apps they’re the second pillar next to Meta, and on Android often the more efficient one. iOS deserves a re-test if you wrote it off in the SKAN era.
What budget do I need to start?
Enough to respect the ratios: 50-100x your target CPI, or 10-20x your target CPA, per campaign per day. Below that, the algorithm starves.
tCPA or tROAS?
tCPA once your action event has volume. tROAS (now on both platforms) when user values vary widely, which for subscription apps is almost always.
Why does my MMP show Google underperforming on iOS?
Historically: modeled conversions that Google saw but your MMP didn’t. Integrated Conversion Measurement (2025) largely fixes this. Make sure your MMP has it enabled before judging iOS spend.
How long before I judge a campaign?
Give learning a week. And never judge iOS data younger than 2-5 days, because it’s still arriving.
We’ve been running App campaigns since the “UAC” name was official. If you want a second opinion on yours, we do a small number of free growth audits: markus@ppc.ee.