Skip to main content



Shift from Theory to Reality: Facebook’s Use of the VCG Auction

https://www.wired.com/2015/09/facebook-doesnt-make-much-money-couldon-purpose/

Throughout our exploration of the Web and sponsored search markets, we’ve looked at both VCG and GSP auctions and examined bidder strategies in the two different auctions. This article expands on our discussion of the VCG auction by both describing how Facebook auctions off its ad slots to advertisers and detailing the myriad of reasons as to why Facebook would use this type of auction. While other students have also examined this topic of Facebook’s VCG auctions by describing how Facebook users’ experiences motivates how Facebook evaluates ads and sells slots to advertisers with the most relevant ads, I thought it would be interesting and beneficial to expand on this discussion by highlighting the advertisers’ real-world perspectives and analyzing external factors that may differentiate Facebook’s idealistic view from reality in today’s society.

First, while other articles and other blog posts describe how Facebook benefits from using a VCG auction to achieve the company’s goal of prioritizing content quality over short-term revenue, this article elaborates on the idea that advertisers also benefit from this style of auction. As Facebook’s John Hegeman describes, the VCG auction with its dominant strategy of truthful bidding, as opposed to the GSP auction, leads to a fundamentally healthier ad system. Since Facebook chooses ads that are most relevant to a user’s news feed and preferences, advertisers are then motivated to properly target ads and focus more on good content instead of simply trying to maximize their own revenue. Hegeman argues that advertisers that adopt this mindset will both save more money and have more confidence in the auction than advertisers who try to game the system. While these concepts cannot strictly be modeled with mathematics, this perspective on advertisers’ motivations encourages us to look beyond the theoretical situations examined in lecture examples of VCG auctions and highlights the real world example of the importance of trust and cooperation in market bidding and, due to the connection between auctions and game theory, player (or advertiser) strategies. Furthermore, this article also reminds us that Facebook’s VCG auction is still largely based on theory and certain assumptions. For example, the reality of advertisers’ behavior in today’s society poses the possibility that advertisers may not think much about gaming the auction (i.e. not bidding truthfully) or the possibility that advertisers do not really understand the specifics of the auction mechanisms. However, despite the possible existence of these realities, the article argues that sometimes, from the advertisers’ perspective, just the belief in an un-gameable slot auction system is sufficient.

Overall, this article serves two purposes as a useful supplemental resource on VCG auctions. Primarily, this examination of Facebook’s use and manipulation of the VCG auction provides us with a real-world example of how a company may run and modify such an auction based on the type of platform and audience; because Facebook cares about the relevance of ad content, Facebook calculates the VCG harm of a particular ad taking a slot by considering both the relevance of the ad as well as the typical component of how much other advertisers would pay if they obtained that slot. Additionally, this article also points out that many times, the theoretical VCG auction, market matching, and game theory ideas we examine in lecture will diverge from being strict mathematical concepts due to societal values like trust and confidence that are introduced when we place these models in realistic environments.

Relevant blog posts from previous years:

https://blogs.cornell.edu/info2040/2017/10/23/the-facebook-experience-featuring-vcg-auctions/

https://blogs.cornell.edu/info2040/2017/10/25/facebook-and-the-vickrey-clarke-groves-vcg-procedure/

Comments

Leave a Reply

Blogging Calendar

October 2018
M T W T F S S
1234567
891011121314
15161718192021
22232425262728
293031  

Archives