First Price Auctions Seek Transparency, but Drive Ad Prices Up
Due to the rise of excessive online advertisements, many of the practices used to optimize these advertisements are becoming increasingly controversial and regulated. This article presents the ongoing issue of bringing more transparency to the algorithmic purchase of online advertisement slots, or programmatic bidding. According to this article, the process of automated ad auctions was historically conducted primarily with a second price auction. Several vendors, however, have started to shift to first price auctions in hopes of providing more clarity to the procedure. As described in the article, an Ad agency, Hearts & Science, did a short 3 week study using both first and second price auctions to purchase “inventory” on a variety of sites. According to the article, this test revealed that Cost per Thousand Impressions (CPMs) were 59% higher in a first price auction than in a second price auction. The article also presented a test with a demand side platform (DSP) that uses bid shading, which I found interesting because we have not covered it in class. Bid shading is essential a technique that splits the difference between the first price winning bid and the second price winning bid. Again, however, CPMs were 54% higher than in a second price auction. Finally, the article specifies that while the prime motive behind switching to first price auctions has been better transparency, advertisers are still unable to see bids of other advertisers, even with the first price auctions.
This is directly related to work we’ve done in the course because it discusses both auctions and advertisement bidding. One specific topic I was reminded of while analyzing this article was the Generalized 2nd Price Auction that we discussed in class on 10/19. While advertisers are asked to announce their value per click in this type of auction, these values don’t necessarily have to be true. So while there may be a push for transparency in advertisement auctions like the ones described in the study above, there can never be a guarantee that whatever announced values are true. In this example, while the change from second price auctions to first price auctions could be good for the site selling the ad slot, it clearly puts the advertiser at a disadvantage and might lead them to believe that the switch to first price auctions forces them to pay more than the slot is worth. Additionally, while this article didn’t directly mention a Nash Equilibrium, I believe that this could be another aspect of the dilemma that could be explored to move closer to a better solution. Given the issue of transparency and what we’ve learned in class about the potential bidding options, it appears that the concerns discussed in this article fall in a bit of a gray area.