Dear FTC, repeat after me: ad platforms don’t set prices

The FTC today re-filed its antitrust lawsuit against Facebook, with its previous lawsuit thrown out in June after a federal judge deemed that the agency failed to prove that Facebook exerts monopoly power in the market for personal social networking (and indeed that the agency had not even established which metrics might be used to determine such a position).

The primary innovation of the new complaint is the intellectual acrobatics it employs in classifying Facebook into a separate category of social products than TikTok, such that Facebook is deemed to face no credible competition for users’ finite attention. From the complaint:

TikTok is a prominent example of a content broadcasting and consumption service that is not an acceptable substitute for personal social networking services. TikTok users primarily view, create, and share video content to an audience that the poster does not personally know, rather than connect and personally engage with friends and family. The purpose for which users employ TikTok, and the predominant form of interaction on the platform, is not driven by users’ desire to interact with networks of friends and family.

I don’t have a law degree, so I feel unqualified to assess the legal merits of the FTC’s market definition.

But what I can capably assess is the FTC’s arguments around Facebook’s control of advertising prices on its platform. Multiple times throughout the complaint, the FTC declares that Facebook’s monopoly control over the market for personal social networking resulted in unnaturally high “advertising prices.” This is simply incorrect, and it reveals a lack of understanding of the digital advertising ecosystem and how advertising inventory is priced. This example comes from page five of the complaint:

By monopolizing personal social networking, Facebook thereby also deprives advertisers of the benefits of competition, such as lower advertising prices and increased choice, quality, and innovation related to advertising.

Digital advertising inventory on large platforms like Facebook is sold through an auction: advertisers bid for impressions, the highest bidder wins, and, depending on the auction design used, either the second-highest (in some flavor of a second-price auction, such as the Vickrey-Clarke-Groves auction design that Facebook employs) or the highest (in a first-price auction) bid sets the price for the placement. Most modern, sophisticated ad platforms allow advertisers to bid against conversions — purchases, registrations, etc. — versus simply bidding for an impression, and the ad platforms use campaign performance to throttle delivery based on calculated click and conversion probabilities for any given user (denoted through an example in the below screenshot with P(Click | User ID=4, Ad ID = 1)).

I describe this dynamic in this QuantMar thread in which I deconstruct a presentation given by a Facebook auction economist on the mechanics of its advertising system. David Philippson and I also discuss the auction process in an episode of the MDM Podcast: How a bid becomes a DAU.

Advertisers tend to price ad platform bids based on the expected value of conversions from that platform. For instance, an eCommerce retailer selling pants might price the bid for a click for a Facebook campaign at $4.50 because, on average, clicks on that campaign ultimately result in $5 in sales. In this case, the advertiser would be capturing $0.50 of margin on average for every click. If the clicks from that Facebook campaign become worth less, on average, then the advertiser will decrease its bid. And vice versa.

The price that an advertiser bids on inventory is wholly dependent on the value of conversions that are produced by that platform. And the degree to which advertisers win auctions is dependent on the competition for that inventory. There is no reason to believe that any advertiser would be paying less for advertising inventory on the Facebook Blue app (or website) today if Facebook had not acquired Instagram or WhatsApp, or if any number of competitive products (but not TikTok!) had entered the very-specifically-defined market in which the FTC believes Facebook operates.

The price that advertisers bid for ad inventory is determined by predicted revenue and campaign ROI (which is referred to as Return on Ad Spend, or ROAS, in the advertising context). And the price that advertisers pay is determined by competition for inventory from other advertisers. But what advertisers optimize for is yield: the amount of ad spend that they recover from reaching users. Price is mostly irrelevant as a standalone metric: most advertisers only consider price in the context of ROAS. The targeting that an ad platform brings to bear on behalf of advertisers impacts the relevance of the users it reaches, which impacts the value that advertisers realize from their campaigns.

If Facebook CPMs — the cost of advertising inventory — suddenly cratered, it wouldn’t be a cause for celebration: it would mean advertisers are bidding less because they are seeing less value from Facebook inventory and their models and machinery have priced that inventory at a lower level via their bids. Interestingly, this is what is happened on Facebook in the wake of ATT: the price of impressions dropped because targeting became more blunted and the cost of conversions spiked considerably and advertisers reduced their ad spend.