Docs

Android Privacy Sandbox Sunset Tracking

What to track on Android now Privacy Sandbox has shut down and attribution is converging on SKAdNetwork-style conversion values.

Acquisition

Google shut down Privacy Sandbox in late 2025, taking the Attribution Reporting API, Topics, and the rest of the on-device measurement stack with it. For teams that had spent a year preparing migrations off third-party cookies and device identifiers, this was a reset rather than a relief — Android attribution is now converging toward a coarser, SKAdNetwork-style model, and most of the tooling built specifically for Privacy Sandbox needs to be retired or rebuilt.

The risk for product and growth teams is treating this as “back to normal.” It isn’t. Google Play’s install referrer API and advertising ID access continue to tighten independently of Privacy Sandbox, and several ad networks have already shipped their own conversion-value abstractions to fill the gap. Without a clear view of which attribution signal is actually live on Android right now, campaign reporting quietly degrades and nobody notices until CAC numbers stop making sense.

Data Points to Track

  • Attribution method in use: explicitly log whether a given install was attributed via Play Install Referrer, an ad network’s own conversion-value postback, or a deterministic SDK-to-SDK match
  • Advertising ID availability: whether the device exposed an advertising ID at install time, since this increasingly varies by Android version and OEM
  • Conversion-value postback payload: the raw value and timestamp returned by any SKAdNetwork-style mechanism your ad partners use on Android
  • Referrer install timestamp lag: time between install and referrer data becoming available, which affects how attribution windows should be configured
  • Fallback attribution rate: the percentage of installs that land in an “unattributed” bucket because no signal was available

Setup Steps

  1. Audit every ad network and MMP integration for references to Privacy Sandbox APIs (Attribution Reporting, Topics, FLEDGE/Protected Audience) and confirm what each vendor has shipped as a replacement.
  2. Standardise an internal attribution-method field on every install event so reporting can be sliced by signal type, not just by campaign.
  3. Re-test Play Install Referrer integration end to end, since referrer timing and reliability have shifted as Google adjusts surrounding privacy controls.
  4. Set explicit alerting on the unattributed-install rate so a silent drop in signal coverage shows up as a metric, not a mystery in next month’s CAC review.
  5. Re-baseline campaign performance for at least two full reporting cycles after any vendor migration, since absolute install and CPI numbers will move even when underlying campaign quality hasn’t changed.

Actionable Insights

The teams that come out ahead here are the ones who stop assuming Android attribution behaves like it did pre-2025 and start treating “which signal attributed this install” as a first-class dimension in their dashboards. That single field turns a confusing month-over-month install drop into an obvious explanation — a vendor changed their measurement approach — instead of a panic about campaign performance. It also gives you the evidence to push back on ad networks whose reported numbers stop reconciling with your own server-side install counts.

Expert help

Need help tracking this in your app?

Our team sets up analytics pipelines for mobile and web teams every day. Talk to us and get your first events flowing in under an hour.

Talk to an expert