An Algorithm for Auctions
One of the most intellectually engaging aspects of search engines is the way they auction advertising space. Often, search engines are challenged to come up with effective ways to make advertisers bid truthfully (or higher than truthfully), in order to make maximum profits. One proposed solution is the Vickrey-Clarke-Groves method. In this solution, advertisers bid on each slot. The highest bidder for each slot wins, and then pays the “damages” he has done to each other advertiser – the relative amount they would gain if he had not bid. An advantage of VCG is that it leads to truthful bidding. Another solution, used by some of the large search engines, is the generalized second price auction. In this auction, advertisers bid. The highest bid wins, and advertisers pay the price of the second highest remaining bid for their slots. Revenue generated from a generalized second price auction can be higher or lower than in VCG.
As the New York Times reports (https://www.nytimes.com/2017/11/02/magazine/how-facebooks-oracular-algorithm-determines-the-fates-of-start-ups.html?mtrref=www.google.com), Facebook uses an entirely different type of auction to sell their ads. Facebook uses an algorithm which takes in the advertiser’s budget, data about the company, and data on users to determine how valuable an ad space is to an advertiser and how likely a user is to click on the ad. Facebook’s stated objective is to maximize everyone’s utility: show users ads they will like and have advertisers pay for high yield space. The algorithm determines the amount of money spent on ads per minute, optimizes both the advertisers’ clicks and Facebook’s revenue.