Will Facebook Audience Network take on the mobile ad networks?

Back in August 2020, a few months after Apple revealed the App Tracking Transparency (ATT) privacy framework at its WWDC developer conference, Facebook (now Meta) published a blog post that outlined its strategy for navigating the post-IDFA advertising environment on iOS. I provide an overview of what that strategy was, at the time, in Facebook has abandoned the IDFA and may kill FAN for iOS. What’s next for mobile measurement?. From that article:

Facebook is in a position to define the new standard, and with its blog post, it did just that: it won’t rely on rough workarounds like sampling from opt-in traffic, but rather, it will build a new measurement infrastructure around probabilistic attribution that is not dependent on platform-specific user-level identifiers being made available to it…The reality is that the IDFA has always been living on borrowed time, and its deprecation was wholly foreseeable. Facebook, especially, should have expected the “June surprise” that was announced at WWDC; it has been locked in a cold war with Apple since at least 2017, as I detailed in The coming war between Apple and Facebook, published in January 2017.

Meta’s August 2020 blog post announced one policy change and hinted at another: the blog post proclaimed that the company would no longer collect the IDFA in its apps, and thus it would not be obligated to expose the ATT opt-in prompt to users. And the company hinted that the deprecation of the IDFA might cause its Audience Network (previously referred to as FAN, and now, presumably, MAN?) to become so ineffective as to be non-viable.

Meta ultimately abandoned this strategy, as I detail in Understanding Facebook’s updated iOS14 advertising guidance from December of the same year: the company would expose the ATT opt-in prompt to its users. My understanding is that Apple informed Meta that, given the volume of personal user data the company commands, exposing the opt-in prompt was obligatory. The company was required to provide users with the option of opting out of tracking.

This chronology of events is important because Meta, to my knowledge, never publicly clarified its position on Audience Network after its strategy reversal, except for in an email to advertisers in July 2021. In that email, the company proclaimed that the Audience Network would transition to contextual targeting.

An Audience Network is effectively a DSP that allows ad platforms to utilize their aggregated, user-level data to fill third-party impressions on properties that they don’t own. I explain the mechanics of an audience network in Can Snap make an Audience Network work?. Meta asserted in August 2020 that its Audience Network might become non-viable on iOS absent the IDFA because the data used to fill non-owned impressions is collected from advertiser conversions generated outside of Meta’s properties, and that flow of data — what I call the events stream — is broken with ATT. I describe this flow of data, and the concept of the hub-and-spoke model of digital advertising, in this piece.

(Aside: the fact that Meta claimed its Audience Network would be rendered unworkable through IDFA deprecation lends support to the notion that Meta’s onsite engagement data, such as likes and comments, is not useful for targeting advertising.)

Contextual targeting utilizes the broader context of a property to fill impressions served within it. User-level data isn’t the primary input, or may not be an input at all, for contextual targeting: ads are paired with content on the basis of presumed or measured relevance. In October 2020, in a post titled IDFA deprecation: winners and losers, I hypothesized that the mobile “broker ad networks” — the ad networks that match mobile advertisers with publishers and primarily, or entirely, operate within the mobile gaming category — would be only minimally impacted by IDFA deprecation through ATT because they mostly utilize contextual data, and not user-level conversion data, for ads targeting.

What ultimately happened following the rollout of ATT was that these broker ad networks benefited from Apple’s reticence to police “probabilistic install attribution using device parameters” (PIAUDP), sometimes referred to as fingerprinting. The threat to broker ad networks with ATT was that install attribution — which is the principal optimization metric that broker ad networks deliver to advertisers, as opposed to value-based optimization targets like purchase events — would be disrupted because Apple’s measurement framework for iOS, SKAdNetwork, is dysfunctional. But PIAUDP allows for installs to be attributed to campaigns in real-time, and with a very high degree of accuracy when a large number of device parameters are incorporated into the fingerprinting model. So rather than imposing even a minimal cost on their businesses, ATT is a boon to the mobile broker ad networks, given that it dramatically undermines the targeting efficacy of ad networks serving owned-and-operated inventory. I posited last month that Apple might police PIAUDP in iOS 16 with an SDK runtime that could mirror what Google recently announced.

And Meta is entering this market with Audience Network: it is transitioning its targeting methodology away from user-level conversion data, of which it has been starved by ATT, to contextual signals. For gaming advertising, contextual targeting can be very effective, and it has spawned three multi-billion dollar public companies in Applovin, Unity, and ironSource. Meta’s Audience Network is estimated by Jounce Media to have generated roughly $3BN in revenue in 2021 (my understanding is that the Audience Network’s gross revenue in 2020 was a multiple of that). For 2021, revenue across the public broker networks breaks down as such:

  • Applovin’s Business Software Platform, of which its ad network is the largest component, generated revenues of $674MM;
  • Unity’s Operate Solutions, of which its ad network is the largest component, generated revenues of $709.1MM;
  • ironSource generated revenues of $553MM;

For reference, Meta generated $114BN in advertising revenue in 2021.

Note that the companies listed above are all unique, and the quoted revenue metrics are not directly comparable (or comprised totally of advertising revenue). But broadly, these revenue numbers (which exclude those of private companies like Vungle) help to guide an estimate of the size of the market for contextual advertising in mobile games. A few characteristics of this market:

  • Advertising spend across mobile gaming broker ad networks is extremely fluid. For the most part, publishers work with all of these networks, and therefore advertisers do too. Advertiser spend on a particular network might react to performance in any given week in ways that are unpredictable. Publisher exclusivity was a phenomenon that created real differentiation in the latter half of the last decade, but consolidation and in-app bidding has mostly put an end to that;
  • Broker ad networks command very high take rates, and no broker networks are incentivized to try to compete on that basis. If any network decreased its take rate broadly as a policy, then the rest of the networks would follow suit and the market would be caught in a race to the bottom. Broker ad networks enjoy much higher margins than DSPs;
  • The broker ad network market is not winner-takes-all. Advertisers and publishers enjoy the optionality of working with multiple partners, even though the products are mostly undifferentiated (of course, each ad network would dispute this!).

Meta entering this space with Audience Network creates a formidable challenger, but also potentially helps to increase the size of the contextual advertising market for mobile games. Again: demand follows supply, and the Audience Network’s SDK footprint is already sizable. My belief is that, prior to ATT, the primary purpose of Meta’s Audience Network was to generate conversions for advertisers that would benefit targeting efficiency within its owned-and-operated products. And relative to Meta’s overall advertising revenue, even a scaled contextual broker ad network wouldn’t meaningfully impact its business. But it could change the landscape of that part of the market substantially.

Photo by Kenny Eliason on Unsplash