AdAttributionKit for teams that already live with SKAdNetwork
Apple introduced AdAttributionKit at WWDC. It keeps the SKAdNetwork model and adds re-engagement and view-through support. What changes in practice for your conversion value plan.
At WWDC this week Apple introduced AdAttributionKit, a privacy-preserving ad attribution framework for iOS and iPadOS. If you have spent years building conversion value schemas for SKAdNetwork, the first question is how much of that work survives. Based on Apple's session, "Meet AdAttributionKit," the answer is most of it. The mental model is the same, and a few additions matter for how you plan.
The model stays the same
Apple describes AdAttributionKit as building on SKAdNetwork fundamentals, with crowd anonymity deciding how much data comes back: the less data a conversion set represents, the less detail is sent to ad networks and advertisers. The session states that it is fully interoperable with SKAdNetwork. Ad networks already registered for SKAdNetwork need no further enrollment, and the source identifier keeps the behavior it had in SKAdNetwork 4.0.
Postbacks look familiar too. They arrive as a signed payload plus unsigned fields, and depending on crowd anonymity they carry a fine conversion value from 0 to 63, a coarse value of low, medium or high, or neither. If your current plan relies on a fine value in the first window and a coarse fallback later, you already think in the right shape.
What is new
- Re-engagement. Impressions can opt in to re-engagement, which targets users who already have the app installed but have stopped using it. Postbacks carry a conversion type of download, redownload or re-engagement. Apple says re-engagement supports only clicks, not views, and that the first conversion value update must occur within 48 hours of the conversion.
- View-through ads. Developers can register a view-through impression, and a view must last at least two seconds. Postbacks record the interaction as click or view.
- A developer mode. A switch in the iOS developer settings removes time randomization and shortens conversion windows, so you can test all three windows without waiting days.
- Per-type updates. A conversion value update can target installs, re-engagements or both, which lets you run a different schema for each.
What this means for your conversion value schema
We would not rewrite anything on the strength of an announcement. What is worth doing now:
- Keep your SKAdNetwork schema. Because the two frameworks interoperate, there is no reason to throw it away. Treat any change as additive.
- Sketch a separate re-engagement schema. The events that signal a returning user are different from those that signal a first-session user. A lapsed subscriber coming back to renew is not the same signal as a new user finishing onboarding.
- List the networks you buy from and ask each about its plans. Whether a network supports the new features depends on that network. Ask your measurement partner the same question, since they decide how re-engagement postbacks appear in your reports.
- Do not expect user-level answers. The privacy thresholds still apply. Campaigns with low volume will return less detail, as they do now.
The open questions
Apple's session does not cover everything. It does not describe overlapping conversion windows in detail, and it leaves out how attribution works when more than one marketplace is involved beyond showing a marketplace identifier field in the postback. Release timing and the minimum operating system versions are details to confirm in Apple's documentation, which you should read before you commit a roadmap.
It is also worth noting what was not announced here: a change to the privacy thresholds, or any promise of more detail. Anyone selling this as a route back to user-level attribution is overselling it.
Testing is where the new framework earns its place. Apple's developer mode removes time randomization and shortens the conversion windows, so you can send a test install through all three windows in a short session and check that each conversion value update arrives as intended. Because the first conversion value update for a re-engagement must occur within 48 hours of the conversion, build your event logic around that limit now. Test the mapping on a staging build first, then confirm with your measurement partner that the postback shows up in their reporting the way you expect.
Practical takeaway
Watch the session, keep your current schema, and draft a re-engagement plan on paper. Then ask each network and your attribution vendor for their support timeline. If you run win-back campaigns today and cannot measure them well, that is where the new framework is most likely to help.