Efficient Ranking in Sponsored Search

Abstract

Abstract. In the standard model of sponsored search auctions, an ad is ranked according to the product of its bid and its estimated click-through rate (known as the quality score), where the estimates are taken as exact. This paper re-examines the form of the efficient ranking rule when uncertainty in click-through rates is taken into account. We provide a sufficient condition under which applying an exponent-strictly less than one-to the quality score improves expected efficiency. The condition holds for a large class of distributions known as natural exponential families, and for the lognormal distribution. An empirical analysis of Yahoo's sponsored search logs reveals that exponent settings substantially smaller than one can be efficient for both high and low volume keywords, implying substantial deviations from the traditional ranking rule.

Revenue optimization in the generalized second-price auction

“…They also performed substantial simulation experiments, demonstrating the effectiveness of squashing as a means of sacrificing efficiency for revenue. Recently-and departing from the model we consider in our own work-it has been shown that squashing can improve efficiency when quality scores are noisy [Lahaie and McAfee 2011].…”

Ranking and Tradeoffs in Sponsored Search Auctions

“…Thus, they argue their method to be preferential. Further strengthening this justification, Lahaie and McAfee [2011] showed that, when there is uncertainty in the estimates of advertisers' relevances, introducing a squashing exponent can actually increase welfare by reducing the weight placed on these uncertain estimates. While we have not explored this issue for our proposed ranking algorithm, it certainly has the same effect, which makes it plausible that their conclusion would apply to our algorithm as well.…”

Loss functions for predicted click‐through rates in auctions for online advertising

“…While minimizing the empirical loss function is equivalent to minimizing MSE when the highest competing eCPM bid is drawn from a uniform distribution, this is not the case for other distributions. Empirically, the uniform distribution is a poor representation of the distribution of competing eCPM bids, as Lahaie and McAfee (2011), Ostrovsky and Schwarz (2016), and Sun, Zhou, and Deng (2014) have noted that these distributions are better modeled by a log-normal distribution in sponsored search auctions on Yahoo! and Baidu.…”

“…Empirically the uniform distribution is a poor representation of the distribution of competing eCPM bids, as Lahaie andMcAfee (2011), Ostrovsky andSchwarz (2016), and Sun et al (2014) have noted that these distributions are better modeled by a log-normal distribution in sponsored search auctions on Yahoo! and Baidu.…”