Original Definition of Facebook’s Unique VCG Auction Process
https://www.wired.com/2015/09/facebook-doesnt-make-much-money-couldon-purpose/
Recently in class we have been learning about the Vickrey-Clarke-Groves (VCG) Auction, a procedure adopted by many tech firms to match advertisers with ads slots. Based on the Second-Price Auction model, VCG interprets the slot price charged to an advertiser as the HARM that this advertiser did to other bidders by taking the slot.
Intrigued by this topic, I did further research on the real-life application of VCG, and discovered that Facebook was the first company to bring this economic concept out of academia and into business practice, according to a Wired interview with John Hegeman, the tech giant’s Chief Economist. From this, I have also quite surprisingly found out that Facebook’s version of VCG has a key nuance to what we’ve learned in lecture. And such a nuance allows for a new, practical interpretation of VCG in social media platforms.
According to Hegeman, Facebook’s VCG auction is unique, because it ranks not only ads against one another, but “ads against the rest”. In lecture, we learned that the usual VCG auction model features a list of advertisers that are being matched to slots, such that the collection of matched pairs generates socially maximum values. The advertisers are the only participants of the “game”. In Facebook’s auction model, however, both individual ads AND posts from users’ feeds are all seeking matching with their most desired slots. According to Hegeman,
“[what Facebook’s] systems are able to do is think about what is the most relevant organic piece of content [they] have and what is the most relevant ad [they] have, and [they] can balance those things against each other”; “[if] there is a really important life event, that can show up on top. If not, and there’s a really relevant ad, the ad can show up higher.”
Facebook’s take on VCG clearly adds complexity to the auction’s goals and mechanism. And it is not difficult to understand Facebook’s motive. As a social media platform, Facebook desires not only profit from selling ads (like what Google does solely), but also top-notch user experience that can facilitate sustainable brand development. This is why sometimes, even if an ad is calculated by the algorithm to have high potential to trigger a purchase, it is ranked lower than a “wish you a happy birthday” post by the users’ friend.
But how should we reinterpret VCG now? Since more than just ads are picking up slots, how can we define the players of the game? And how to define the so-called “price” and “harms”, if they still exist?
Below is my personal attempt to solve this problem. First, the advertisers are playing against the company Facebook itself. While the former wants more potential customers and in turn profit, the latter seeks user retention and engagement, which in turn also turns into profit. In a way, by adding posts into the ranking system, Facebook is advertising for itself by satisfying its users. Second, the definition of price and harm remains essentially the same as those in the original model. The only difference is that sometimes the price can be an intangible opportunity cost.
To explain this, let’s imagine a simple scenario in which there are only two players, a post and an ad, competing for two slots.
If the post is ranked higher than the ad, the post “pays” x, and x is the harm the post (i.e. Facebook) does to the advertiser by taking the top spot. Of course, as Facebook the company is the player here, they are not really paying themselves any real money. Nor are they gaining directly anything monetary. Facebook, however, is paying an opportunity cost––the opportunity of theoretically making more $ from selling that top spot to the advertiser. Such an opportunity cost could also be transformed into actual loss of money. Although Facebook’s exact auction mechanism is a blackbox, it is reasonable to assume that Facebook would give an advertiser some form of compensation (or discount when they sign the contract) for ranking their ad lower than the advertiser would anticipate. With everything being discussed, however, Facebook might happily pay such opportunity cost by ranking the post at top for brand development strategies as argued above.
In the same vein, if the ad is ranked at the top, the price y that it pays to Facebook is the harm that the advertiser does to Facebook by preventing the company to yield higher user engagement and satisfaction from showing users the post first. This is also an opportunity cost, or the possible benefit from what could have happened.
From this mini research project, I’ve learned that anything taken straight from the textbook may be flawed or incomplete in the real world. Adaptations and reinterpretations are always necessary.