Comparing predicted prices in auctions for online advertising Eric Bax a, ⁎, Anand Kuratti b , Preston Mcafee c , Julian Romero d a Yahoo! Marketplace Architecture, 3333 Empire Blvd., Burbank, CA 91504, USA b Yahoo! Software Development India Pvt. Ltd., Torrey Pines, Embassy Golflink Business Parks, Off Koramangala-Indiranagar, Intermediate Ring Road, 560 071, India c Yahoo! Research, 3333 Empire Blvd., Burbank, CA 91504, USA d Purdue University 100 S. Grant St. West Lafayette, IN 47907, USA abstract article info Article history: Received 21 December 2009 Received in revised form 9 June 2011 Accepted 9 June 2011 Available online 14 July 2011 JEL classification: C13 C44 C45 D81 D84 Keywords: Reversion Validation Bias Auction Prediction Online publishers sell opportunities to show ads. Some advertisers pay only if their ad elicits a user response. Publishers estimate response rates for ads in order to estimate expected revenues from showing the ads. Then publishers select ads that maximize estimated expected revenue. By taking a maximum among estimates, publishers inadvertently select ads based on a combination of actual expected revenue and inaccurate estimation of expected revenue. Publishers can increase actual expected revenue by selecting ads to maximize a combination of estimated expected revenue and estimation accuracy. © 2011 Elsevier B.V. All rights reserved. 1. Introduction Online publishers use auctions to sell opportunities to advertise, called ad calls, to online advertisers. There are two broad categories of online advertising auctions: search and display. In search advertising auctions the advertiser pays only if their ad elicits a click. In display advertising auctions, advertisers may select a basis for payment. Some advertisers pay when the ad is shown, others pay only when showing the ad elicits a user response such as a click or a purchase. (For details on auctions for online advertising, refer to Varian (2006, 2009), Edelman et al. (2007), and Lahaie and Pennock (2007).) When advertisers pay per click or other user response, the revenue received by the publisher for showing an ad is random. Since user response rates are not known exactly but must be estimated, there is uncertainty in addition to randomness. The estimation accuracy of response rates varies. One reason is that the amount of historical data varies. Another reason is that the response rates themselves vary, and more data is required to estimate smaller rates with the same relative accuracy. With randomness, a risk-neutral seller seeks to maximize expected revenue. Facing uncertainty, the seller may select an offer having maximum estimated expected revenue. However, this is not necessarily the best policy for maximizing actual expected revenue. The reason is that selecting a maximum estimate selects for a combination of having an over-estimate and having a large actual expected revenue. Some classes of ads are more likely to have inaccurate estimates, such as ads with lower response rates and ads for which there is less historical data. Even if the individual response rate estimates are unbiased, these classes are more likely to have the largest response rate over-estimates. So selecting a maximum estimate can favor these classes even if they offer less expected revenue than other classes. Having more buyers in the auction exacerbates the problem, because more estimates means more and more extreme over-estimates. However, having many buyers is not sufficient for selecting a maximum estimate to be a sub-optimal policy for maximizing expected revenue. Varying levels of uncertainty about revenue distributions is also required. This paper is organized as follows. Section 2 describes related work. Section 3 presents some theory on selection bias for estimated offer values. Section 4 explores correcting selection bias for online display advertising auctions. Section 5 focuses on corrections for search advertising auctions. Section 6 discusses opportunities for future work. International Journal of Industrial Organization 30 (2012) 80–88 ⁎ Corresponding author. Tel.: + 1 626 296 1946. E-mail addresses: ebax@yahoo-inc.com (E. Bax), kuratti@yahoo-inc.com (A. Kuratti), mcafee@yahoo-inc.com (P. Mcafee), jnromero@purdue.edu (J. Romero). 0167-7187/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.ijindorg.2011.06.001 Contents lists available at ScienceDirect International Journal of Industrial Organization journal homepage: www.elsevier.com/locate/ijio