Google’s Multi-Billion Dollar Buy-Side Problem

I’m taking a break from non-stop coverage of Google’s antitrust problems to… just kidding, I have a brand new angle on Google’s antitrust problems. 

While the DOJ, EU, State AGs, and civil plaintiffs have all made more or less similar arguments about Google’s monopolistic power in the market for sell-side ad serving and exchanges, there has been notably little activity on the buy-side of the business. 

That’s because the buy-side, while enormous, does not enjoy monopolistic power or market share. DV360 is the largest DSP, but certainly has less than 50% market share. Google Ads (aka, AdWords) competes with Meta, Amazon, TikTok, and many others for advertiser spend. Antitrust enforcement cannot get traction against Google’s buy-side despite some obviously anti-competitive behavior, like the tie between DV360 and YouTube.

Civil cases, however, are not reliant on monopolistic power. Among all the evidence that was discovered during the DOJ and other cases were examples of behavior from Google that seemed to be against the interests of AdWords customers. This could open the company up to significant liability.

For years, Google was operating what have been dubbed “third-price auctions” when AdWords was bidding into AdX. I honestly couldn’t believe this was true when I first heard about it, but the evidence is clear and the documents are damning.

Take yourself back to the early days of Google’s acquisition of DoubleClick and the nascent Google AdX. In those days, it was not entirely clear that AdX would become the dominant force in publisher indirect revenue monetization. There were also many companies providing “yield management” services, including PubMatic, The Rubicon Project, and Admeld. I write about this period of uncertainty in my book. During this time, the team believed it was imperative that AdX monetized well, or else publishers might choose a yield manager as their primary pipe for indirect inventory. If the yield manager generated higher CPMs, publishers might turn off AdX’s dynamic allocation, and instead put the AdX tag into the yield manager, instead of the opposite.

The way to avoid this fate was to pump up the demand inside AdX so it monetized better. But there was a problem: Even if they increased bids from AdWords to AdX, there was still a lack of competition on most auctions, so the higher bids would get second-priced to a low clearing price no matter what. Google at some point estimated that AdWords would second-price itself 80% of the time! Google’s quant team, gTrade, came up with an interesting solution to this problem: They would send the top two bids from AdWords to AdX on every auction, so the first bid would only get reduced to one cent above the second bid. This effectively meant that the winner was paying significantly more than they would have otherwise, for no good economic reason.

To my knowledge, this behavior of artificially inflating second prices continued from early in the history of GDN (as display ads in AdWords was known) through the end of the second-price auction in 2018, almost 10 years.

I have very little original information as to the current state of buy-side litigation. I have this tweet:

I have seen various pieces of disclosure in the cases (some not in public evidence), and there are very damning statements by Google employees about this behavior that lay it out quite clearly. There’s no lack of evidence.

I also have people in the know whispering that it’s going to be a big problem. So take this all with a grain of salt. What I can do is some basic extrapolation on how big of a problem this might be.

The first question: How much did GDN spend on AdX over this time period? We have this data! (See Monopoly Report: Day 5: Nothing happened but here are some cool charts.)

Eye-balling the chart’s light blue line, we have about $500 million/month in spend from AdWords into AdX display, in the period immediately after the move to first-price auctions. We know that well over 90% of AdWords spend goes to AdX, so other third-party exchanges are a rounding error. Let’s call it a high of $6 billion a year in ad spend, ramping up from a low base in 2008, or 10 years total. Sounds like about $30 billion in spend affected by third-price auction shenanigans ($6B/year, 10 years straight-line ramp). 

The next question would be how much of this $30 billion in spend was overpriced versus a theoretically fair second-price auction. I haven’t found any document in evidence that states this exactly, but we can leave that for the courts. There’s also the issue that civil suits under the Sherman Act are subject to treble damages. That adds up.

Watch this space.