At its WWDC event in June, Apple announced iOS 10, the latest version of its operating system for iPhones and iPads. Alongside the host of new features that Apple announced for iOS 10 at the event, some time was dedicated to privacy in the operating system: specifically, Apple’s Senior Vice President of Engineering, Craig Federighi, discussed a technique that Apple would be employing in iOS 10 called differential privacy that would help mask the activities of individual users while still preserving aggregate statistics for the purposes of predictive analytics and suggestion.
What wasn’t discussed at WWDC is another change related to user privacy that Apple is introducing in iOS 10: the “zeroing” of advertising IDs (IDFAs) for users that have opted out of ad tracking. This is a significant change, and its implications aren’t totally clear yet (iOS 10 will probably be made available some time in September, although beta versions are already available).
But first, some background. Apple introduced the “Limit Ad Tracking” setting in iOS 6 back in 2012; when a user had activates this setting now, a developer is able to access that device’s IDFA but also has access to a variable that indicates whether or not the user has opted to limit ad tracking. In its documentation on the advertising identifier, Apple admonishes developers to only use the IDFA for the following purposes if the limit ad tracking variable is set to true:
If the user has limited ad tracking, use the advertising identifier only for the following purposes: frequency capping, attribution, conversion events, estimating the number of unique users, advertising fraud detection, and debugging.
In other words, when a user has set their “limit ad tracking” setting to true, their IDFA can still be used for various purposes, just not for ad targeting. For instance, the IDFA of a user that has limited ad tracking can be used to limit the number of times advertisements are shown to that user (frequency capping), but it can’t be used to optimize the specific advertisements shown based on their history of clicking on ads.
This access to the IDFA even when ad tracking has been limited serves another important purpose: ad attribution. Mobile attribution providers use the IDFA (among many other things) to attribute app installs to their source ads (for the uninitiated: ad attribution is a fairly esoteric process that allows a developer to track which ad a user clicked on before installing an app, thus allowing them to calculate return-on-investment for specific campaigns. This article provides a fairly good, basic overview of the process).
But in iOS 10, Apple will no longer make the IDFA of users that have limited ad tracking available to developers; instead, when a user updates their limit ad tracking setting to true, their IDFA will be replaced with a string of zeroes:
The use of the “limit ad tracking” feature has actually decreased since it was first introduced, according to a study conducted by mobile ad tech company TUNE. In analyzing its own proprietary data, TUNE found that only about 17% of mobile users in the US and UK had activated the limit ad tracking setting in February 2016 (Google also offers a similar setting on Android, although they haven’t announced any policy changes regarding access to the Android advertising ID), down from 22% in August 2015. The company speculated that the drop in ad tracking limitation was due to an increase in the usage of mobile ad blocking tools (although mobile ad blocking is oriented almost entirely to the mobile web).
But if use of the limit ad tracking setting remains at 17%, the zeroing of IDFAs could present a huge problem for mobile advertisers. The inventory on some ad networks is only matchable by IDFA; this means that it’ll be very hard, or even impossible, to track “conversions” (or ad clicks that lead to installs) on those networks. And without being able to track conversions, advertisers can’t calculate ROI across the campaigns they have running on various networks.
For instance, if an advertiser is running mobile ad campaigns on Facebook, Unity Ads, and Applovin, and they can’t assign individual installs to the campaigns and networks from which they were generated, they have no ability to determine whether those campaigns are running profitably; all they’ll have access to will be overall advertising spend (across the three networks) and overall install numbers (across the three networks and also including organic installs that originated from eg. search or virality).
This is an untenable situation for advertisers: even over a low double-digit percentage of the total smartphone-owning population, IDFA zeroing could create an impossibly opaque advertising environment. Paul Müller, the CTO of mobile attribution provider Adjust, told me:
We are currently seeing around 10-15% of all users with the “limit ad tracking” flag turned on in iOS. Should all of those people update to iOS 10 and no longer provide IDFAs, it would certainly have an impact on all ad networks that exclusively use the IDFA for conversion tracking. Facebook, some of Google’s inventory, and Twitter would suddenly see their conversion rates drop and the effective CPI for that traffic increase.
One possible outcome of IDFA zeroing is the accelerated transition of advertisers into the walled gardens of ad networks that control deterministic data about users from logins. Facebook, for instance, has massive reach, multiple inventory formats — video and static across Newsfeed, Instagram, Audience network, and, soon, Facebook Live — and, most importantly, targeting options based on its own proprietary data profiles. When confronted with the reality of IDFA zeroing, it’s possible that some advertisers may just decide to pick Facebook as an exclusive channel rather than undertake the statistical gymnastics required to calculate ROI when running ad campaigns across multiple channels using an MMP.