What Is App Tracking Transparency (ATT)?

App Tracking Transparency (ATT) definition: Apple’s permission framework that requires apps to ask users before tracking them across other companies’ apps and websites — typically to access the IDFA for ad targeting and measurement.

If a user taps Allow, tracking is permitted; if not, the IDFA returns zeros and you must not track.

Within the context of mobile apps it’s the privacy gate where your growth stack meets user consent. Since iOS 14.5, users see a system prompt — “Allow ‘App’ to track…?” — and can choose Allow or Ask App Not to Track.

That single choice reshaped iOS acquisition, attribution, and monetization.

What does ATT look like in apps?

System prompt: The standardized iOS dialog appears when you request tracking; you supply a short, human “why” string (NSUserTrackingUsageDescription). Apple’s rules prohibit workarounds (e.g., fingerprinting).

What counts as tracking: Accessing the IDFA or combining your app’s data with data from other companies for ads/measurement. Consent governs both.

That’s the surface. The impact shows up in opt-in rates — and those rates have moved.

Why does ATT matter?

• Opt-in isn’t one number anymore. Recent panels show ~35% average ATT opt-in (Q2 2025, Adjust) with large variance by vertical and country; AppsFlyer’s global panel reported ~50% by early 2024 (e.g., France/China ~53%). Your prompt strategy now materially affects growth math. (Source adjust.com)
• iOS ad spend recovered. Four years in, iOS non-organic installs and spend have grown again as consent stabilized and SKAN matured — evidence marketers found a viable playbook under ATT. AppsFlyer

If consent is the lever, you need to know how ATT works under the hood — and what replaces user-level attribution when users say no.

How does ATT work?

The process is simple and clear. Here is how Apple Developer describes it:

1. Ask for consent using Apple’s framework. You can educate first with your own screen, but the system prompt decides.
2. If allowed, you may track within policy (IDFA available). If denied, treat the device as opted-out — no tracking.
3. Measure with SKAdNetwork for opted-out traffic: aggregate, delayed postbacks instead of user-level attribution. (Think “campaign-level signal, privacy-preserving.”)

ATT is consent; SKAdNetwork is attribution. Keep them distinct.

What is the difference between ATT vs. SKAdNetwork?

• ATT = permission to use cross-app identifiers (like IDFA) at the user level.
• SKAdNetwork = aggregated attribution for installs/events without user-level IDs (with ongoing versions like SKAN 4 that add multiple postbacks and coarse values).

With the plumbing sorted, two practical questions remain: What’s “good” opt-in — and how do you earn it without being pushy?

What Is a “Good” ATT Opt-In Rate?

There’s no universal target. Context rules (vertical, region, audience, timing).

Directionally, expect ~35% as a blended average, with some datasets near ~50% and big swings by country. Benchmark against your category and top markets — then set realistic lift goals per placement. (Source adjust.com)

Opt-in lives and dies on timing and clarity. Here’s what’s working — compliantly.

Practical Ways to Lift ATT Opt-In (Compliant & User-First)

Here is a list of quick solutions on how to improve ATT from Appfollow team:

• Ask after a “first win.” Show value, then request consent. Apps that delay the prompt until users experience a benefit tend to earn more yeses (reflected in rising opt-ins over time).
• Explain the “why” (pre-prompt). A short, friendly screen before the Apple dialog — “Help us keep the app free and relevant” — followed by the system prompt. Localize your NSUserTrackingUsageDescription so it’s specific, not vague.
• Measure by cohort. Track prompt display rate, consent rate by country/channel, and downstream ROAS/CPI deltas for opted-in vs. SKAN traffic. Industry data shows consent can be meaningfully lifted with better timing and copy — so instrument it.

FAQs

Is ATT only about the IDFA?
Mostly — but it also covers combining your app’s data with data from other companies for ads/measurement. That still requires consent.

Do users actually opt in?
Yes, and rising. Recent panels show ~35% average (Adjust, 2025) and ~50% in AppsFlyer’s dataset (Q1 2024), with major differences by vertical and country.

What replaces user-level attribution when users say no?
SKAdNetwork — Apple’s privacy-preserving, aggregated attribution framework (now in SKAN 4 with multiple postbacks and coarse values).