Google App Campaigns in 2026: what changed, what’s still a black box, and how to win anyway

Original article (source): Adapty - “Google App Campaigns Playbook 2026: Setup and Optimization”

• https://adapty.io/blog/google-app-campaigns-playbook-2025/

What’s new (the bits worth paying attention to)

Adapty’s main “2026 is different” callouts are about measurement + surfaces, not a brand new campaign type:

• Web → App Connect expansion: more Google surfaces can now drive into app, and (critically) web campaigns can get credited for app outcomes.
• iOS web-to-app measurement improvements (post-ATT reality): more visibility into web → install → in-app action.
• YouTube Shorts enhancements: more interactive/creator-friendly formats, plus more delivery contexts.
• AdAttributionKit (AAK): Apple’s direction of travel beyond SKAN 4.

Translation: if your tracking/event quality is shaky, “more automation” just means faster spend into noise.

The part everyone hates (and you can’t change): the black box

They frame Google App Campaigns as a black box because you don’t get the knobs you’re used to:

• No manual keywords/placements.
• Limited placement + audience breakdowns.
• Weak asset-level reporting compared to Meta-style creative analytics.

So the game becomes: feed the system better signals (events) and give it better ingredients (assets).

Setup that actually matters: pick the right optimization event

Their most practical guidance is to avoid “install-only” thinking.

• If you can, optimize to a deeper event (trial start, onboarding complete, purchase, subscribe).
• The algorithm needs enough daily volume on that event to learn (they suggest ~30–50/day for the target event).

If you can’t hit that volume yet, use a tiered approach (install → mid-funnel → purchase) rather than jumping straight to tROAS with zero signal.

Tracking: the boring foundation that decides your CPA

They’re blunt: bad tracking = bad users.

Recommended stack:

• Firebase or an MMP (AppsFlyer/Adjust/Singular) with clean event mapping.
• Events that are timely, consistent, and aligned to value.

They also argue Firebase tends to work best with Google (audience signal syncing + tighter ecosystem integration).

Structure: split by OS + goal (and don’t cannibalize yourself)

Their structure advice is conventional, but it’s conventional for a reason:

• Separate Android vs iOS campaigns.
• Separate installs vs actions (and avoid multiple campaigns in the same geo optimizing for the same event).
• Use one theme per ad group / asset group so your creative tests are interpretable.

Creative: treat assets as your targeting

Because you can’t control placements, your assets do that job indirectly.

Useful reminders from their spec section:

• Provide multiple angles (utility, lifestyle, social proof, urgency).
• Don’t ship only one format - Shorts + Discover need portrait-friendly assets.
• Show app UI early; design for sound-on where relevant.

Budget + scaling rules (to avoid resetting learning)

Their scaling heuristics are basically “don’t thrash the model”:

• Increase budget gradually (e.g., ~20% every 48–72 hours).
• Don’t swing bids wildly (keep changes bounded).
• Let performance stabilize over ~5–7 days before big moves.

The take

If your Google App Campaigns feel opaque, don’t fight the black box.

Instead:

1. Choose an event that matches value (not just installs).
2. Make tracking rock-solid (Firebase/MMP, clean event mapping).
3. Give Google better creative inputs across formats.
4. Scale slowly so you don’t keep re-entering learning.

Read the original: https://adapty.io/blog/google-app-campaigns-playbook-2025/