Why Facebook uses VCG rather than Google’s GSP
This article explores and compares the Vickrey-Clarke-Groves mechanism that Facebook uses to Google’s Generalized Second Price, as well as reasoning behind the choices.
According to Facebook Engineer, John Hegeman, the ideal type of advertisements make it such that users see whichever advertisements that is most relevant to them. This would result in more clicks for advertisers and keeping these advertisers as repeat customers. This ideal is hopefully reached by the fact while the Vickrey-Clarke-Groves algorithm will not generate the most immediate revenue in the short term, it is predictable and theoretically un-gameable by advertisers. Under this model, Facebook will only give an advertising spot to an advertiser, while calculating the harm done to other advertisers and keeping in mind the relevance to the user. The model used will not only compare advertisements to advertisements but also compare the advertisements to the rest of Facebook. Thus the model operates under the hopes that with more specialized advertisements, users will click on them more, leading to advertisers becoming return customers and playing for the long term.
In constrat, Google’s GSP algorithm doesn’t need to particularly keep a user’s relevance as in mind, as Google users GSP for their search engine. Google’s goal is not to keep you on Google, but rather get you to your desired destination as fast as possible. Thus for Google, GSP becomes an efficient way to make money.
This post connects back to our classes by exploring usages of VCG as well as the ways it can be used even if it does not generate the most immediate revenue. The properties that it is un-gameable with a dominant strategy are desirable factors that must be considered.