Skip to main content

Auctioning and Advertising

The article “Bid Shading is Covering for Inefficient Programmatic Algorithms” by Jeremy Fain delves into a new aspect of advertisement auctions that could have unforeseen consequences on the industry. In order to understand this problem, we must first delve into the basics behind how ads are sold and how this has led to the emergence of bid shading. In the past, most companies have used second-price auctions to sell ads on websites and on television. This type of auction, as we have learned in class, has a dominant strategy of bidding your true value. While this auction is rather simple for the bidders, it is unfortunately becoming replaced across the programming advertisement industry by first-price auctions, according to Fain. This transition has led to many challenges among bidders, one of which is that in first-price auctions the bidders must pay their bid. This has in turn led to a sharp increase in the prices advertisements have gone for across the industry as bidders try to make sense of the new system.


The immediate impact of this change is that it has caused many customers to become more wary to auctioneers and interact less with the market. As a result, bid shading has emerged in order to help mitigate the perceived added costs of switching to a first-price auction system. Bid shading is, as Fain puts it, is “an algorithm to automatically tweak the marketers’ bid price lower based on historical data that considers exchange, the size of the ad and the site that the ad would be featured on”. While bid shading does reduce the price for buyers, the way it does so is wildly inconsistent, as current algorithms are not optimally developed in order to completely fix the problem. Another issue Fain points out is that even though the algorithm decreases the costs of buying advertisements for the bidder, it doesn’t do much in directing bidders towards buying ads that are more likely to reach their target audience and earn them increased revenue. In order to address these problems, Fain believes investment in machine learning and deep learning are likely to help further refine algorithms needed for bid shading and/or find a better alternative approach that favor matching up customers more closely with advertisements that are likely to give them returns (leaving them satisfied). As touched upon in earlier, this heavily connects our discussion of auctions and finding a dominant strategy with our coverage of advertisement and deciding how to match advertisers to clients.


Leave a Reply

Blogging Calendar

October 2019