FROM THE INDUSTRY
There were other ways for ad exchanges to try to juice their results. An exchange might duplicate its bids to the same buyer just to see if the buyer’s algorithms might have some throttling or inefficiency to exploit. Some exchanges would send multiple sequential auctions but with slightly lower floor prices each time to try to “fish” for the highest possible bid from the buyer. Even without cheating, there was enormous inefficiency in the ad exchange world that had developed by the late 2010s. A typical setup for a publisher might include two header bidding wrappers (one from Amazon, one from the prebid open source project), plus Google’s product Open Bidding through the ad server, resulting in three calls to each exchange for each impression. Each exchange would then, in turn, make ad requests to all the Demand Side Platforms (DSPs) they had connected to, resulting in potentially hundreds of requests for each single impression. On top of this spider’s web, you have aggressive yield management techniques being implemented by the publishers themselves, trying to eke out additional pennies wherever possible, causing prices on the same inventory to vary enormously depending on the buying path. All of this is to say there can be culpability for Google for their actions, but we must consider the challenges of operating within this fog of war—programmatic advertising had evolved into a complex and challenging environment without ground rules or regulations and with active and innovative adversaries. Google is ultimately an engineering company. The driving ethos of the people who work there is to solve problems. They could only continue for so long acting as a leader in advertising while inefficiencies, workarounds, and complexity proliferated. By 2019 the advertising market was certainly filled with all those problems, and the AdX and Open Bidding teams were going to try to solve them all in one swoop.
Financial markets are heavily regulated. Banks have internal compliance officers with fiduciary duties to protect not just the firm but also the market. In the years since the emergence of ad exchanges, the advertising market has adopted the structure of finance markets without any meaningful oversight or controls. A stock market had organically grown, and the “banks” were put in charge of their own regulation. This gap in accountability would allow Google, but also its many independent competitors, to manipulate the market for their own gains. The actual customers, the advertising buyers and publisher sellers, would largely be afterthoughts and would have little insight into what was happening in the middle of their transactions. Five years after the transformative DoubleClick acquisition, Google found itself in an enviable position within the fast-growing and dynamic display advertising business. Google effectively had strong—or arguably dominant— technology in every corner of the digital advertising business. That meant that for many available ad auctions, Google technology, and Google customers, were on both sides of the transaction. In many cases, Google systems would set both the price the advertiser was willing to bid and the price at which the seller was willing to sell. They also controlled the ad server that determined which ad ultimately
would serve and the exchange that saw all the demand for each ad, yielding nearly perfect information on every transaction. In what would become an iconic—and quite infamous—comment in an email discussion, Jonathan Bellack laid it all out, asking, “Is there a deeper issue with us owning a platform, the exchange, and a huge network? The analogy would be if Goldman or Citibank owned the NYSE. In 2012 a New York–based quantitative team called gTrade was formed. Their mission was to develop, test, and implement changes to the display advertising systems to improve Google’s profits, particularly on the buy side, meaning the ad network and DSP. This was a proven model within Google, as the innocuously named Ads Quality Team had been regularly tweaking and influencing search ads results for some time in order to increase revenue, sometimes generating enormous returns. gTrade took on some projects that would result in improvements in the ad products themselves, such as better fraud detection. They would become better known for a series of changes to the auction mechanics that in some cases would become quite controversial. The outside advertising community would not hear about the projects with code names like DRS, Bernanke, Poirot, and Bell for many years, but they would feel the results. Google would use these projects to grow its revenue at the expense of competitors, and in some cases at the expense of its own customers. Project Poirot stands out among Google’s various auction manipulations as possibly the most reasonable. While the imperative for Poirot was formed in reaction to header bidding, the actual problem it meant to solve was real. The evidence clearly suggests that ad exchanges were running “dirty auctions,” where the second-price auction mechanism was being manipulated to raise prices for advertisers. Google’s internal data showed that reducing bids on those exchanges produced better results for advertisers with the same level of quality.
SEPTEMBER 2025 Volume 47 No.3
81
Made with FlippingBook - Online magazine maker