App Tracking Transparency Codex: the ultimate guide to ATT, IDFA deprecation, and SKAdNetwork

Published on December 29, 2020 by Eric Benjamin Seufert

Apple threw the entire mobile ecosystem into a frenzy at WWDC in June 2020 when it announced a radical new change to the way the Identifier for Advertisers (IDFA) could be accessed and used for advertising tracking. This change takes the form of the new App Tracking Transparency (ATT) framework: developers will need to collect an explicit opt-in from users before accessing their IDFA. It is believed that very few people will opt into having their IDFAs tracked, which is problematic for advertisers, as the IDFA is the primary means by which iOS advertising campaigns are measured and optimized.

The IDFA is what is known as a Mobile Advertising ID (MAID), with the equivalent identifier on Android being the Google Advertising ID (GAID). The IDFA is useful for advertising purposes in two ways:

It allows an advertising network to attribute installs to a given campaign through an attribution process that involves attempting to reconcile an IDFA to an ad click when the IDFA is seen in an app. This process is usually facilitated for app advertisers by a Mobile Measurement Partner (MMP), which reconciles IDFAs seen in an app via SDK presence against IDFAs that were recently seen on an intermediary “landing page” that was loaded from an ad click. More detail on how the mechanics of mobile attribution work via the IDFA can be found in this article;
It allows an advertising network to aggregate in-app events to user profiles and ultimately build a device graph. The Self Attributing Networks all require advertisers to place an SDK into their apps which can transmit IDFA-indexed in-app events back to those platforms. These networks, therefore, have unmatched visibility into which users spend money in apps and can use that knowledge to optimize advertising campaigns. A “device graph” is a profile of a specific device that includes historical data about the apps that have been downloaded to it and engaged with (including in-app purchases). If an advertising network — especially a very large one with almost universal reach, like Facebook or Google — knows which users monetize most frequently, and in which types of apps, they can very efficiently traffic advertising campaigns on the basis of engagement and monetization.

With the effective deprecation of the IDFA through the ATT framework, Apple removes this critical advertising measurement and optimization tool from the mobile advertising ecosystem. In its place, Apple presented SKAdNetwork 2.0 at WWDC: the second iteration of the app attribution system it launched to very little fanfare in 2018. I discussed the potential impact of SKAdNetwork in an article in 2018, but very few people at the time believed it was a harbinger of a larger effort by Apple to disrupt the mechanics of mobile advertising.

SKAdNetwork essentially replaces the functionality of MMPs in a privacy-centric way by transmitting app attribution data (and a very limited amount of in-app event data) at a campaign-centric aggregation level, versus the user-centric aggregation level that is made possible with the IDFA. In other words, advertising networks can know that a given campaign produced an install, but they won’t know which user generated that install, and they’ll learn very little about subsequent activity in the app after the install.

This article is a living document and is designed to be a comprehensive list of resources dedicated to elucidating ATT, SKAdNetwork, and the motivations behind this seismic shift. I will update this document over time as new information becomes available and new insight is published, on this website and elsewhere.