Following up on naota’s post about Google’s algorithms which maximize revenue instead of achieving the optimal social welfare maximization, I will focus on the particular aspect of automatic matching. In late February of 2008, Google informed a group of advertisers that they were selected to participate in the automatic matching beta where Google uses the advertiser’s surplus budget on queries that it deems to be relevant but are currently not on the advertisers’ keyword list. For example. if your keyword bid was on “running clothing”, the query “jogging pants” would display your ad. Similarly, a keyword bid on “craft for kids” would result in your ad being displayed for “things make”. However, the question is how relevant are “craft for kids” ads when a user is searching for “things make”? Clearly, advertisers are likely to suffer because they are potentially paying even though they are failing to reach out to their targeted audience.

Unsurprisingly, Google maintains secrecy over the exact mechanics and algorithms underlying its auctioning procedure for advertisements. However, natural questions that follow are how much should advertisers pay if their advertisement is clicked solely because of automatic matching and what slot should the advertisement be placed in? This throws a new twist to the auctioning procedure. If all the bidders for “jogging pants” have received their slots, where will the automatically matched advertisers from “running clothing” be placed? If a placement downgrades the slots of those who actually bid for “jogging pants”, the original advertisers would definitely be unhappy since they were ripped off. Thus, the alternate option is to place all the automatically matched advertisements at the bottom but how much value is there? Also, how will Google determine the price to charge per click for such low positions? Charging the original price is certainly unjustified. A third option is to integrate the automatically matched results with the bidders for that keyword into one larger auction using the VCG procedure. Again, the same question remains. How will Goggle determine the prices (how much the advertiser values a slot) to enter for the automatically matched advertisers?

It appears that Google has overstepped its boundaries. In auctions, the seller does not control the value that each bidder possesses for the particular item. However, in this case, Google appears to be controlling the value that the advertisers value certain slots under certain keywords. It is very possible that the advertiser may not even agree that some of the keywords are relevant. Regardless of which of the above options Google chooses, it is essentially deciding the value that a buyer has over its advertising slots. Essentially, the analogy is that the seller in an auction decides how much each buyer should bid. Of course, the seller benefits because it can set any arbitrary value to profit. This can be considered an unfair practice which Google is using to increase its revenue. Even worse, some advertisers may not even be aware that such practice is in existence and thus, could be cheated out of a significant sum of money by literally unkonwingly participating in an auction. Of course, the optimal social welfare outcome is for each bidder to bid truthfully but how does Google even know what those values are? Instead, it has its own interests in mind to generate more revenue. Even worse, it knows how much companies are willing to spend in total since they each have a budget so it can exploit that and use as much as possible.

One possible solution to the problem is to allow advertisers to review the automatically matched keywords before deciding whether or not to bid for them. However, it does not appear that Google possess those intentions given that it intends to set the default action of each advertiser to allow for automatic matching to occur. Unlike the idea of utilizing perfect matching to maximize social welfare, Google seems to be more interested in increasing its revenue. Since this was only in beta in February, we shall see how this story unfolds.

Sources:

http://www.straightupsearch.com/archives/2008/02/googles_automat_1.html

http://www.altogetherdigital.com/20080227/google-beta-tests-automatic-matching/

http://www.channelregister.co.uk/2008/03/18/when_google_does_evil/