What WWDC 2025 changes in AdAttributionKit and what to ask your MMP
Apple's WWDC 2025 session on AdAttributionKit describes new capabilities in iOS 26. Here is what vendors report, what is unconfirmed, and what to ask your MMP.
Apple's WWDC 2025 session "What's new in AdAttributionKit" describes new capabilities in iOS 26. The framework, which succeeds SKAdNetwork, has had little adoption so far, so the practical question is less "what is new" than "does any of this change what I do this quarter." Mostly it does not, but a few items are worth putting on your roadmap now.
What is new, according to attribution vendors
Apple's own session page was thin on detail when we read it, so the specifics below come from Singular's write-up and should be checked against Apple's documentation before you plan around them.
- Overlapping re-engagement windows. AdAttributionKit previously allowed one re-engagement conversion window at a time. Singular reports that multiple overlapping windows are now possible, with a conversion tag that keeps each effort separate. That helps if you run several re-engagement campaigns or work with several partners.
- Configurable attribution windows. The defaults were a 30-day click window and a 1-day view window. Singular says you can now set lengths per ad network, per interaction type, or globally, and ignore some interaction types, such as view-through from one network. It notes this applies to installs, not re-engagements.
- Attribution cooldown. A cooldown per conversion type to stop closely timed conversions from competing for credit.
- Country codes in postbacks. An optional country code derived from the App Store, subject to Apple's crowd anonymity thresholds, so geo data appears only when volume is high enough.
- A development postbacks tool. A setting that lets you generate test postbacks without a live ad flow, which should shorten integration testing.
One discrepancy to flag: Singular lists these as arriving in iOS 18.4 and later, while Apple's session page frames the new capabilities as part of iOS 26. Which feature ships in which version affects your timeline, so ask your vendor to state it feature by feature.
Why adoption matters more than the feature list
Singular's own caveat is that SKAdNetwork 4 has not reached 50 percent adoption and most large advertisers still run on SKAdNetwork 3, while AdAttributionKit, released in March 2024, has seen little use. Large ad networks rely on their own modeled methods. The takeaway is uncomfortable but useful: improving the framework does not help you if your networks and your MMP do not support it. Apple said in April that Apple's ad platform would register with AdAttributionKit, which is one step toward the framework mattering for a channel you probably buy.
What to do now
- Ask your MMP for a dated roadmap. Specifically: which of these features are supported or planned, on which iOS versions, and how will the new data appear in your dashboards?
- Audit your re-engagement setup. If you run more than one re-engagement program at a time, overlapping windows could remove a long-standing compromise. Write down how you currently split credit so you can compare later.
- Decide your windows before you can set them. Configurable windows are only an improvement if you have a view. A subscription app with a free trial might argue for a longer click window than a utility app with an impulse purchase. Test any change, because longer windows claim more installs, not necessarily more incremental ones.
- Do not expect geo targeting from country codes. Crowd anonymity limits the data, so treat country as a sanity check on reporting, not a new optimization input.
- Do not rebuild now. iOS 26 was only announced. Most of your users are on older versions, so these features cannot be your main source of truth for some time.
What the cooldown and windows could mean for your numbers
Attribution windows decide who gets credit when several networks touch the same user. Shorter windows and cooldowns tend to reduce claimed installs, longer ones increase them, and neither changes how many people actually installed. If your networks start using different settings, their reported numbers will stop being comparable with each other and with your history. Ask for each network's settings in writing, and keep a changelog of when they change, so a drop in one channel's installs is not mistaken for a performance problem.
The point of view
Teams often respond to a WWDC announcement by rebuilding their measurement. We would do the reverse. Keep your current setup, note what a new capability could improve, and wait until your vendor ships support and a meaningful share of installs come from devices that can use it. Meanwhile, put effort into the measurement that does not depend on Apple: incrementality tests and an aggregate model.
The practical takeaway: send your MMP the roadmap questions from the first step above this week, and file the answers next to your 2026 measurement plan.