What is SKAdNetwork (SKAN)?

What SKAN replaced

Before App Tracking Transparency, iOS attribution worked by matching a device identifier at click time to the same identifier at install time. It was deterministic, immediate, and user-level.

ATT made that identifier available only with explicit consent, which most users decline. SKAdNetwork is Apple's replacement: the operating system itself observes the install and reports it, so no party ever holds a linked record of who clicked what.

How a SKAN postback differs from a normal one

• Apple sends it, not you. The postback originates from the device's operating system.
• It is deliberately delayed. A randomised timer separates the install from the report, so timing cannot be used to re-identify anyone.
• It carries no user identifier. Only a campaign identifier and a conversion value.
• It can be withheld or degraded. Where crowd anonymity thresholds are not met, the conversion value is coarsened or omitted — see coarse conversion value.
• It cannot be joined to your data. There is no key. This is the point of the design, not a limitation to engineer around.

What this changes operationally

Three habits from the identifier era stop working and need replacing rather than patching.

Real-time optimisation. Timers mean performance signal arrives late, so daily bid tuning on fresh conversion data is not available. Campaign structures have to be simpler and left to run longer.

Deep funnel measurement. The conversion value is a small, fixed-width signal, so you must decide in advance which few post-install outcomes are worth encoding. Everything else is unmeasurable through SKAN.

Granular campaign structures. Splitting spend across many small campaigns pushes each below the anonymity threshold, and the data disappears entirely. Consolidation is a measurement requirement, not just an efficiency preference.

SKAN, AdAttributionKit and what comes next

AdAttributionKit is Apple's successor framework, extending the same privacy model with support for alternative app marketplaces and re-engagement. The principles carry over: aggregate reporting, deliberate delay, thresholds before disclosure.

The broader lesson generalises well beyond iOS. Aggregated, delayed, deliberately imprecise measurement is the direction of travel across the industry, and the teams that adapted best treated it as a reason to invest in incrementality rather than as a reporting problem to be worked around.