Apocalypse Soon: What happens when the iOS advertising ID is deprecated?

NOTE: THIS POST WAS WRITTEN IN FEBRUARY 2020 AND PROVIDED A PREDICTION FOR HOW THE IDFA MIGHT BE DEPRECATED. This post is old. For a more recent take on Apple’s decision to deprecate the IDFA, see: Apple killed the IDFA: A comprehensive guide to the future of mobile marketing.

It’s rare for me to speak with a mobile advertiser that doesn’t view the deprecation of the IDFA — Apple’s proprietary advertising identifier for iOS devices — as a certainty, if not an imminent one. While the deprecation of the IDFA has been a possibility for years, and rumors of the identifier’s demise reliably circulate before every WWDC, a few recent developments support the notion that 2020 might usher in the end of an advertising ID for iOS devices:

  1. In its most recent earnings call, Facebook’s CFO cryptically made reference to upcoming changes by mobile platform operators that will introduce friction into the process of mobile advertising measurement;
  2. Facebook has also made some other changes that tangentially could relate to a potential end of the IDFA: it announced that it will no longer operate its Audience Network on the mobile web, and that it has modified its MMP program to disallow the sharing of view-through attribution data at the user level (see a note from Adjust below). Again, while neither of these issues directly relates to mobile advertising tracking or the relevance of an IDFA, they do both support a more privacy-centric posturing by Facebook that was perhaps taken to appease Apple;
  3. Apple updated its iOS developer guidelines after last year’s WWDC developer event to bar apps in the Kids category from using third-party analytics or advertising SDKs. This was interpreted by many as a dry-run for full IDFA deprecation;
  4. Google announced that it will phase out 3rd-party cookies from the Chrome browser within two years. Although Apple already does this with its Safari browser on mobile by default, Chrome has a much larger marketshare, and so this is a bigger development. While it is popular wisdom that the deprecation of the IDFA by Apple would likely instigate Google to deprecate the Android-equivalent GAID, given Chrome’s reach, it could be that Apple decides to one-up Google on the basis of privacy by reacting with the IDFA deprecation;
  5. Apple introduced Sign in with Apple at WWDC 2019 as a means of allowing users to tie app- and website-specific authentication to their Apple IDs, meaning that Apple could ultimately be the route between devices and third-party platform IDs.

Of course, all of these development could merely be coincidental to a decision by Apple to deprecate the IDFA; Tim Cook seems zealously protective of user privacy, and the large mobile platforms generally appear to be eager to stave off invasive regulation in the wake of public furor over privacy incidents such as with Cambridge Analytica, etc.

And obviously, IDFA deprecation may not actually be on the immediate horizon, and it may not happen at all. But it’s interesting to consider what happens to the mobile advertising industry if it does, given the incredibly prominent role that the IDFA plays in even the most sophisticated advertisers’ attribution models for iOS. To that end, I present: a speculative, assumption-dependent vision of the first few months after the deprecation of the IDFA.

June 2020

Apple’s annual WWDC developer conference reveals a slew of new features coming to iOS 14 later in the year. Cryptically, Apple’s Vice President of Engineering, Craig Federighi, reveals that Apple has implemented a “simplified system profile” on iOS that limits the surface area of device information that can be collected by an app. The “simplified system profile” was first introduced to Safari in macOS Mojave in 2018 to stop devices from being fingerprinted by advertising networks. Marketing blogs and MarTech Twitter alight with speculation about the implications of this change.

Following the keynote, the first beta of iOS 14 is released to registered developers, and an important update to the documentation for the advertisingIdentifier property, which stores the user’s IDFA, is discovered (bolded text is new):


In iOS 10.0 and later, the value of advertisingIdentifier is all zeroes when the user has limited ad tracking.

In iOS 14.0 and later, limit ad tracking is enabled by default.

Tech commentators immediately seize upon this revelation and declare that default IDFA zeroing will effectuate the end of direct response mobile advertising. Assuming that the iOS 14 adoption timeline resembles that of iOS 13, it is estimated that 50% of iOS devices will have limit ad tracking (LAT) enabled within a month of iOS 14’s release in September, and nearly all iOS devices will have limit ad tracking enabled by September 2021.

The mobile advertising technology ecosystem enters a state of panic, organizing war room meetings to come up with strategies for handling the monumental change coming to their industry in September, just three months away. Ad tech suppliers wonder if any additional device-level properties can be added to their models to improve the precision of fingerprinting, which, under the new “simplified system profile,” will only be reliable for a few minutes after an install. But most recognize that Apple is committed to removing or obfuscating these unique device signatures, such as when it added noise to device sensor data in iOS 12.2 as per the recommendations of this research paper. The changes implemented to iOS 14 are interpreted to be grievously dangerous by mobile ad tech vendors.

Privacy advocates very vocally champion both the mobile simplified system profile and the new default limit ad tracking setting.

July 2020

Facebook announces that it will stop sending user-level data to its attribution partners for both iOS and Android, across all campaign objectives, by September. The move is unexpected and widely interpreted as being a concession to Apple in order to continue attributing users with its own proprietary data collected from its SDK. Google, Twitter, Snap, and Pinterest all announce very similar policy changes just a few days after Facebook.

These announcement stun the mobile advertising industry: if Facebook, Google, et al no longer relay user-level data to measurement partners, then install and event-based campaign types like Facebook’s AEO and VO and Google’s UAC can only be assessed for ROAS via aggregated, campaign-level data from the platforms themselves.

Mobile advertisers become anxious: what does this mean for measurement? And how will non-Self Attributing Network (SAN) sources of traffic be evaluated? App advertisers are unable to glean much helpful information from MMP partners, Apple, or their account representatives at various networks.

August 2020

Apple announces the first set of members of its new Certified Mobile Advertising Partners program: advertising platforms that are registered to attribute installs via the SDKAdNetwork API that Apple introduced in iOS 11. The list of partners includes most of the largest and most prominent providers of mobile traffic.

At this point, advertisers begin to get a clearer picture of the future of mobile measurement and the severity of the changes being instituted. If attribution is handled directly from iTunes and doesn’t include user device identifiers, then the impending new reality of advertising dictates:

  1. ROAS and CPE campaigns will only be possible via the SANs that are able to do any form of fingerprinting via their proprietary SDK data and the revenue data they collect;
  2. Cost Per Install values for campaigns from the networks in Apple’s Ad Partners program are calculable, but building effective ROAS models for these campaigns will be difficult, if not impossible. Without being able to attribute revenue to campaigns (because all IDFAs are zeroed, and SKAdNetwork transmits no identifiable user information), the traffic sources of monetizing users are unknowable;
  3. Most of the infrastructure currently supporting mobile advertising will soon become obsolete.

Advertisers scramble to assemble data science teams to build Media Mix Models from their existing data and to validate their current ROAS models. Shifting from deterministic measurement to probabilistic measurement will take time; in the interim, their ROAS models can be used to provide credible assumptions around what various traffic is worth. Advertisers also begin attempting to re-negotiate their contracts with technology and data partners, recognizing that many of these partners won’t be needed in their new measurement paradigms.

Google announces that it will also turn on “Opt out of Ads Personalization” by default in the next Android release.

September 2020

iOS 14 is released to the general public. Almost immediately, CPMs on mobile plummet as advertisers reduce media spend across the board. Some apps see their advertising revenues drop by half; apps that rely exclusively on advertising for user acquisition experience shrinking DAU levels. Advertisers concentrate media spend across Apple Search Ads, Facebook, and Google while conservatively layering in traffic from other sources as measured through attributed installs from iTunes (via SKAdNetwork) in a media mix model. Many advertisers note that their Facebook install volumes have decreased relative to levels from before iOS 14, recognizing that fingerprinting may have previously over-attributed LAT installs to Facebook.

December 2020

By the end of the year, most advertisers have adjusted to the new ecosystem reality, although overall mobile advertising spend is about 10% lower than it had been one year prior. Since iOS 14 was launched, a major ad network acquired an attribution provider and is using its data to develop a media mix model service. Because of the added complexity of tracking and measuring the effectiveness of media spend, a large proportion of smaller advertisers are now only advertising on either Facebook or Google and Apple Search Ads; larger advertisers have built their own infrastructure to aggregate campaign-level install data and integrate it into their media mix model.

Consumer sentiment around Apple’s privacy changes to iOS 14 was overwhelmingly positive in June, but the fanfare waned over time and those changes are rarely cited or referenced going into the new year.

Photo by S Goswick on Unsplash