iOS attribution with ATT and SKAN

What are ATT and the IDFA?

App Tracking Transparency is Apple's framework that requires apps to ask permission before tracking a person across other companies' apps and websites, and the IDFA is the device advertising identifier that permission unlocks. On iOS and iPadOS 14.5 and later, an app must request authorisation before it can read the IDFA.

If the request is not made, is denied or is restricted, the IDFA comes back as all zeros. The same happens in the Simulator. With no IDFA, an MMP cannot use it to join an ad click to an install.

When the prompt does not appear

Apple's documentation lists cases where the ATT prompt is skipped and the result returns at once:

• tracking is restricted on the device;
• the person has turned off "Allow Apps to Request to Track";
• a request is already pending;
• it is called from an app extension;
• the app is not active.

ATT opt-in rates are not given here, because no primary source publishes them.

What is SKAdNetwork?

SKAdNetwork (SKAN) is Apple's privacy-preserving attribution framework, in which the device itself records ad interactions and installs and then sends a delayed, aggregated postback to the ad network. Apps do not need ATT permission to use SKAN.

The flow, in short: an ad network signs its ad, the device records the interaction, the person installs and opens the app, the app updates a conversion value as the person does things, and after the windows close Apple sends postbacks. Apple's SKAdNetwork page now points developers to AdAttributionKit for app ad campaigns, but SKAN 4 rules still describe how much of the ecosystem works.

SKAN 4 postbacks and conversion windows

SKAN 4 can send up to three postbacks for one winning install, one for each conversion window. This needs iOS 16.1 or later and ads signed with SKAN 4.

Other rules worth knowing:

• An app can lock a window early with lockWindow, which ends updates for that window and starts the delay sooner.
• With SKAN 3 and earlier, each higher value reset a rolling 24-hour timer, then a further random delay followed.
• Apple says the minimum time from impression to postback is 24 to 48 hours, and a testing profile cuts this to 5 to 10 minutes.
• Install windows: 30 days after a StoreKit-rendered ad or attributable web ad, 24 hours after a view-through ad, and the app must be launched within 60 days of install.
• Besides the winning postback, up to five non-winning postbacks can be sent. They carry no conversion value and only the first postback is sent for them.

Conversion values, fine and coarse

A conversion value is a number the app sets to describe what the person did after installing, which Apple passes back in the postback if privacy thresholds allow.

• Fine value: a 6-bit integer from 0 to 63. Values outside that range are rejected.
• Coarse value: low, medium or high. The app or network decides what each means.
• Fine values are only used in the first window. The second and third postbacks carry coarse values only.
• The older update method without coarse values works only in the first window and does not support multiple postbacks.

A worked example

A language learning app maps its fine values in window 1: 0 for an install with no activity, 10 for finishing the first lesson, 20 for starting a trial and 40 and above for buying a subscription, banded by price. For windows 2 and 3 it uses coarse values: low for any lesson, medium for a streak of three days, high for a renewal.

A person installs, finishes a lesson on day 0 and buys on day 1. The app sets the fine value to 40 and locks window 1. About one to two days later, the network receives the first postback with value 40, if the crowd anonymity tier allows a fine value.

Crowd anonymity tiers

Apple assigns each install a tier based on how many similar installs there are, considering the publisher app or domain, the advertised app, the country and the source identifier. Smaller crowds get less detail.

This is why a postback can arrive with no conversion value at all. Other reasons include the app never setting a value, the postback being a non-winner, and an out-of-range fine value. Consolidating campaigns to raise volume per source is general practice for reaching higher tiers.

AdAttributionKit

AdAttributionKit (AAK) is Apple's newer attribution framework that builds on SKAN, adds re-engagement and works with alternative app marketplaces as well as the App Store. It is available from iOS 17.4 for installs, with re-engagement added in iOS 18.

• AAK uses the same three conversion windows and the same random delays as SKAN 4.
• SKAN and AAK impressions are ranked together and only one wins. Clicks beat views, then the most recent wins, and up to six impressions are considered.
• Postbacks state a conversion-type of download, redownload or reengagement.
• Attribution windows can be configured in Info.plist: 1 day view and 30 days click by default, configurable from 1 to 7 days for views and 1 to 30 days for clicks.
• Country code is a bonus on top of the four tiers, present only with enough volume.
• Like SKAN, AAK does not need ATT permission.

See attribution windows for how these defaults compare with other platforms, and self-attributing networks for how Meta reports iOS campaigns through its own Aggregated Event Measurement.