The impending deprecation of the IDFA has already severely impacted the mobile advertising ecosystem, and reverberations will continue to be felt through its implementation. But the pain of IDFA deprecation is not evenly distributed. Some corners of mobile advertising — across both advertisers, publishers, and ad tech companies — will experience more damage than others, as when an earthquake completely decimates certain neighborhoods within a city yet leaves others unscathed.
One misconception that is important to disabuse in discussing this topic is that mobile app advertising was absolutely deterministic in the waning years of the IDFA. It wasn’t. As I wrote in June, ahead of Apple’s announcement at WWDC that the IDFA will be deprecated, the combination of SAN primacy in attribution as well as limit-ad-tracking rates in some countries approaching 50% meant that attribution on mobile was already probabilistic to some degree. This topic was also discussed at length in a recent episode of the MDM podcast.
Another common assertion that should be stress-tested is the existence of various workarounds that will fully capture the precision of the IDFA. The latest of these is the supposition that the combination of a user’s IDFV and IP address will allow ad tech platforms to profile and target users as capably as can be done with the IDFA. I walk through the logic of why this is improbable in the below Twitter thread, but I think a good heuristic to use in evaluating these theories is:
If a plausible workaround to IDFA targeting was possible, why did Facebook claim that IDFA deprecation would severely debilitate Facebook Audience Network (FAN), to the extent that FAN might need to be shuttered for iOS?
The economics of these probabilistic solutions are unworkable. If any given probabilistic profile has a 50% chance of being who the advertiser thinks it is, then bid prices must come down by 50% to compensate for that. There is no hand-waving away the brutal economics of mobile acquisition: the vast, overwhelming majority of smartphone owners won’t monetize in any given mobile app.
Putting aside any specifics around the logistics of implementing a fingerprinting solution that serves as a sufficient substitute to the IDFA — and ignoring the fact that Apple has the total and unassailable agency to smother any such approach via its app approval process — the simple fact that Facebook admitted that FAN might not be viable without the IDFA serves as a credible indictment of fanciful schemes around precision fingerprinting.
It’s also important to keep in mind that Apple’s parallel privacy push in the browser has also advanced with iOS 14, with ITP being turned on by default for all browsers. Apple doesn’t appear to be pursuing privacy protections in a way that exists as a deliberately porous obstruction to user profiling.
These two points aside, it’s helpful to consider which sectors across mobile are hurt worst by the IDFA deprecation, and which escape mostly intact. But to do so, it’s important to first articulate what exactly the IDFA allows for in mobile advertising: understanding what the IDFA facilitates helps to create an outline of what will be missing when it isn’t available.
The IDFA allows for three things:
- Install attribution: crediting an ad campaign with an install via IDFA reconciliation;
- Events attribution: aggregating events generated by a user across apps into a unified profile via the IDFA for the purpose of ad targeting;
- User targeting: targeting a specific user, usually on the basis of a profile built from aggregated in-app events, via their IDFA through a programmatic channel.
SKAdNetwork allows for the install attribution use case, although only at the campaign level, and not in real time. But the second and third use cases cannot be done in a definitive way in the absence of the IDFA. So it stands to reason that the companies that benefit most from these use cases are hurt most in IDFA deprecation.
For this exercise, I categorize the universe of companies in the mobile app advertising ecosystem as Ad Tech Companies, Content Publishers / Advertisers, and Service Providers, and I evaluate the damage that IDFA deprecation will inflict on subcategories of companies within each.
Some commentary on each case below.
Content Publishers and Advertisers
(Note: A full overview of each of the acronym-designated business verticals below is beyond the scope of this article, but Quantmar is a great resource for becoming familiar with them)
Demand Side Platforms & Supply Side Platforms: These companies mostly assist advertisers and publishers through the targeting / yield management functionality that is delivered through their IDFA-indexed device graphs and bid information streams. The deprecation of the IDFA significantly handicaps the ability of DSPs to drive efficient traffic acquisition, especially with re-targeting. DSPs and SSPs can transition their infrastructure to contextual analysis, but it’s hard to argue against the idea that some portion of the market for this segment of the ecosystem simply dissipates with the deprecation of the IDFA. Additionally, the lack of real-time insight into install conversions, and the lack of granularity at the level of the impression, disproportionately harms DSPs. And SSPs are deprived of their user-centric flow of bid information with the deprecation of the IDFA.
Mobile Measurement Partners: I don’t see how it can be argued that the core value proposition of install and event attribution on iOS doesn’t expire absent the IDFA. Advertisers mostly want a unified attribution solution, and some are even planning to not trigger ATT in iOS 14 as a result. Managing two separate attribution mechanisms — one for users that opt in, another for those that don’t — creates onerous overhead that is potentially not justified from a value standpoint. Some of these companies will pivot, and they’ll continue to serve Android advertisers, but this category is severely impacted by IDFA deprecation.
Self-Attributing Networks: SANs are impacted in their ability to aggregate user events in service of targeting high-value individuals. Facebook acknowledged this when it retired the Value Optimization campaign strategy in its post-IDFA platform adjustment: without the event streams that Facebook and Google and other SANs were ingesting from across the mobile ecosystem, these platforms can’t evaluate the likelihood of a given user monetizing in a given app when it determines which advertiser should win an auction. My sense is that the SANs can ultimately replicate the targeting performance of IDFA-indexed user profiles with contextual signals, but this will take time.
Broker Ad Networks: The reason I think that Broker Ad Networks will only be minimally impacted by the deprecation of the IDFA is that these companies mostly only minimally used the IDFA, relying instead on contextual signals and a core skew towards gaming advertising.
High IAP Monetization Games: “Mid-core” games like RPG card collection and 4X build-and-battle games, as well as social casino games, rely disproportionately on the ability of SANs to aggregate event streams at the user level to deliver high-value users to acquisition campaigns. These games tend to also utilize re-targeting campaigns to a greater degree than games from other parts of the monetization spectrum. These games tend to chase niche audiences, and thus these games are very dependent on the ability of ad platforms to do hyper-segmented creative optimization and filtering based on monetization potential, both of which are made possible by the IDFA.
Moderate IAP Monetization Games: While many “Casual” games like puzzle and simulation games rely on the monetization profiles owned by ad platforms for targeting, they have larger underlying appeal, they tend to be easier to advertise without hyper-targeting, and they rely less on extreme monetization from a very small percentage of the user base — what I call the long-tail LTV distribution — for economic viability.
Ad Monetized Games: “Hypercasual” games generally don’t target specifically for traffic acquisition purposes, and so their marketing efficiency will be only trivially impacted by IDFA deprecation. But hypercasual games are heavily dependent on event profiles for ads monetization: from a contextual standpoint, traffic in these games is considered to be very low quality for advertisers, but they allow advertisers to go “hunting” for high-value users with IDFA targeting. When the ability to target specific users by IDFA is revoked, the CPM value of traffic in these apps will plummet.
IAP Monetized Utility Apps: This is a broad category. Most IAP-monetized utility apps rely heavily on lookalike audiences and events-based feedback loops for advertising, both of which are mostly made possible by the IDFA (Facebook will allow for lookalike audiences in iOS 14 traffic, but only on the basis of campaign conversions, not custom audiences, and targeting for campaigns will necessarily be limited). Utility apps tend to carry fairly well-defined contextual signal, however, and so CPMs in these apps may not be impacted materially by IDFA deprecation.
Subscription Monetized Apps: Subscription apps are an interesting case, with most having adopted web-based onboarding flows that mostly don’t rely on the IDFA for attribution. The advertising efficiency of these campaigns will not change in IDFA deprecation, although the loss of lookalike audiences will impair targeting (as it does for IAP monetized utility apps).
Data Management Platforms: Much like MMPs, the mobile use case for DMPs is almost inextricably linked with deterministic identification: joining user profiles to devices. It’s hard to imagine that DMPs will be able to deliver value in a world where users cannot be identified with a high degree of precision.
Agencies: Agencies will likely fall victim to a systemic decrease in advertising efficiency; given that most agencies don’t employ proprietary technology or have access to exclusive pools of traffic, and that almost all agencies operate on a percentage-of-spend commercial model, a reduction in global mobile app advertising efficiency will linearly impair their revenue.
Facebook Marketing Partners: FMPs build automation tools for Facebook (and other SANs). These companies are adversely impacted by structural limitations of SKAdNetwork: campaign limits (100 for SKAdNetwork, but just nine for Facebook in the early days of IDFA deprecation) and no native creative information in the SKAdNetwork postback. And, significantly, Facebook is shuttering its Marketing API for iOS 14 traffic (it remains to be seen if the API will be re-opened).
The categories I identify here are somewhat arbitrarily defined, and they are not comprehensively explored. I have almost certainly missed some interesting cases in this broad analysis. But I tried to identify the most relevant cases, or the cases that sit at extreme ends of the harm spectrum.
IDFA deprecation, in my view, was always a question of infrastructure re-tooling: the need to build an entirely new analytics and measurement system that isn’t foundationally supported by the IDFA. Every operator across mobile is feeling and will continue to feel this specific pain as their advertising infrastructure is decommissioned and replaced, but the death of advertising identifiers is only existential to some parts of the ecosystem.
Thanks are owed to David Philippson and Nebo Radovic for providing feedback on this article.
Photo by Simon Connellan on Unsplash