AdAttributionKit re-engagement gets sharper after WWDC

What changed at WWDC

Apple used WWDC last week to confirm a set of refinements to AdAttributionKit's re-engagement measurement, tucked in among the Liquid Glass redesign and the iOS 26 naming switch. Developers briefed on the changes describe a tighter conversion value schema for re-engagement events and, reportedly, improvements to how postbacks for lapsed-user campaigns are timed against a user's return session. Apple made no changes to App Tracking Transparency.

AdAttributionKit re-engagement measurement has always been the weaker half of the framework. When Apple introduced AdAttributionKit at WWDC 2024 as SKAdNetwork's successor, the design effort went into first-install attribution: crowd anonymity thresholds, conversion value windows, source identifiers. Re-engagement was added on top, with a coarser signal and less flexibility, because Apple treats a returning user as a smaller privacy risk than a new one and built the plumbing accordingly. UA teams running winback and re-engagement campaigns have lived with that gap for a year.

Why the gap mattered

The practical effect was a measurement blind spot exactly where hybrid-casual and puzzle publishers spend some of their heaviest budget. Re-engagement campaigns aimed at lapsed players, often running through Meta and Google's app campaigns alongside network partners, reported back through a postback structure that could not distinguish a high-value returning spender from a player who opened the app once and left. Bidding algorithms optimising against that signal had less to work with than they did for new-install campaigns, and measurement teams ended up leaning on probabilistic modelling to fill the gap, with the attribution risk that implies.

As we argued in MMP Selection Criteria in the AdAttributionKit Era, the framework's early limitations pushed a lot of that modelling work onto MMPs rather than Apple's own postbacks. Reports suggest the WWDC changes narrow, rather than close, that gap: the conversion value schema for re-engagement is understood to now carry more bits of usable signal, and the postback timing changes should reduce the lag between a re-engagement event and the report a UA team can act on.

How AdAttributionKit re-engagement compares to the first-install rollout

It is worth remembering how bumpy AdAttributionKit's first-install rollout was over the past year before assuming the re-engagement refinements will land smoothly. Adoption of the framework's crowd anonymity thresholds and conversion value windows for new installs took most MMPs several months to implement fully, and early postback data during that window was inconsistent enough that several measurement teams ran parallel SKAdNetwork and AdAttributionKit reporting for longer than Apple's own migration timeline suggested would be necessary. There is no reason to expect the re-engagement update to move faster through the same MMP integration cycle, even though the underlying change is narrower in scope. UA teams that budgeted a quick two-week validation window for the new schema based on how simple the WWDC announcement sounded should plan for something closer to the six-to-eight-week integration lag the first-install rollout actually took.

There is also a coordination problem worth flagging early. Because re-engagement postbacks interact with a returning user's existing install record rather than creating a fresh one, any misalignment between an MMP's old and new schema during the transition period risks double-counting or, worse, silently dropping re-engagement conversions that fall exactly on the cutover date. Measurement teams should agree a hard cutover date with every MMP in writing, rather than allowing a gradual rollout where some campaigns report under the old schema and others under the new one for an extended period, since that scenario makes any blended reporting during the transition effectively unusable.

The postback strategy checklist

Nothing in AdAttributionKit's re-engagement update is optional to test. Before shifting budget on the strength of a WWDC keynote line, run through this before iOS 26 ships in the autumn:

• Pull your current re-engagement postback configuration from every MMP you run and confirm which fields map to Apple's revised schema versus the old one.
• Re-map your conversion value buckets for re-engagement events specifically. Do not assume the mapping you use for first-install conversion values carries over.
• Get your MMP's engineering or solutions team to confirm, in writing, when their platform will support the updated schema in production rather than in beta.
• Reset your baseline. Any month-on-month comparison of re-engagement campaign performance that spans the schema change is comparing two different measurement systems.
• Brief your creative and campaign teams on the change before the reporting numbers move, so a shift in reported re-engagement rate does not get misread as a creative or bidding problem.
• Hold a fixed set of re-engagement campaigns on the old configuration for a two-week overlap window if your MMP allows it, purely to sanity-check the new numbers against the old ones.

A worked example

Take a hypothetical mid-size puzzle publisher spending $200,000 a month on re-engagement campaigns across Meta, Google and one incremental network. Under the pre-WWDC schema, that publisher might see attributed re-engagement conversions covering roughly 40% of the return sessions its product analytics recorded, with the remainder invisible to postback-based reporting and reconstructed through modelled attribution. If the revised conversion value schema genuinely carries the additional signal reports describe, that coverage figure could plausibly move into the 55% to 65% range once postbacks are fully mapped, which is a large enough shift to change how a bidding algorithm allocates spend across channels. Those numbers are illustrative, not a benchmark, and any UA team should build its own version from its own postback logs before trusting them.

What this means for the next fiscal quarter

The timing matters. iOS 26 ships alongside the new AdAttributionKit behaviour in the autumn, which gives UA and measurement teams roughly three months to retest re-engagement postback strategy before it becomes the only strategy available. Treat this as a live measurement change to validate, not a settled improvement to bank on: the schema details are still coming through developer channels rather than finalised documentation, and any team that waits until the September ship date to start testing will be reworking postback logic in the middle of Q4 planning season instead of ahead of it.

Budget owners should also plan for a reporting discontinuity in their own quarterly reviews. Whatever coverage improvement the new schema delivers, the numbers either side of the cutover will not be directly comparable, and a re-engagement campaign that appears to jump in efficiency the month iOS 26 ships is more likely reflecting better visibility into activity that was already happening than a genuine change in campaign performance. Flagging that distinction to finance and leadership ahead of time, rather than after a quarterly report shows an unexplained swing, is the difference between this being read as a measurement upgrade and being misread as a sudden and unrepeatable win.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Updating the playable ad spec for re-engagement campaigns
• MMP Selection Criteria in the AdAttributionKit Era