App Tracking Transparency shipped in 2021. By 2026, the dust has settled: ~65% of iOS users decline tracking across apps. Cross-app attribution is essentially dead for most consumer apps. This is how we actually run analytics on mobile apps we ship.
What first-party analytics can still measure
- In-app events · every screen view, tap, scroll, feature use
- Session duration + frequency
- Crash-free session rate
- Conversion from activation → paid (within your app)
- Retention curves
- Anonymous device properties (OS version, device class, locale)
What it cannot measure
- Which ad drove the install (for users who declined tracking)
- Cross-app behaviour (do my users also use X?)
- Personal identifiers like email or IDFA without user consent
The stack we ship
- PostHog or Mixpanel SDK on-device for event capture
- Authenticated user ID joins server-side · user-consent implied by login
- Cohort-based A/B testing · not individual-level random assignment
- Apple Search Ads + SKAdNetwork for post-install attribution (limited, but legitimate)
- Privacy manifest + ATT prompt handled gracefully in onboarding
How we make product decisions
Product decisions on cohorts + retention curves · not individual tracking. 'Users who completed onboarding within 48h retain 3× better in week 4' is a decision signal. 'User XYZ did N things in sequence' mostly isn't, post-ATT.
If your product manager still demands user-level funnel reports, ship those only for logged-in + consented users. Design the product so key events happen after login. Post-ATT analytics is a design choice, not just a measurement choice.

By
Dezső Mező
Founder, DField Solutions
I've shipped production products from fintech to creator-tooling · for startups and enterprises, from Budapest to San Francisco.
Keep reading
RELATED PROJECTS