Part 30 - Google Replaces Its Last Look Preference from Dynamic Allocation

IV. GOOGLE’S SCHEME TO DOMINATE THE AD TECH STACK

D. Google Responds to the Threat of Header Bidding by Further Excluding Rivals and Reinforcing Its Dominance

2. Google Blunts Header Bidding By “Drying Out” the Competition

g) Google Replaces Its Last Look Preference from Dynamic Allocation with an Algorithmic Advantage and Degrades Data Available to Publishers

256. The shift of transactions from rival ad exchanges to Google’s ad exchange as part of Unified Pricing Rules came at the same time that Google relinquished one of the substantial preferences it had previously given its ad exchange: the “last look” advantage of dynamic allocation. Its reason for doing so was not altruistic. Maintaining last look would have given publishers a path to effectively floor AdX higher than other ad exchanges notwithstanding Unified Pricing Rules. And as with the shift to a first-price auction, Google believed that removing last look “allows bundling of other valuable changes, that can be positioned as procompetitive.” Google Engineering Director explained that Google “paired this change [dropping last look] with other benefits to Google (fair access and uniform reserve prices), rather than being forced by regulators to remove last look under disadvantageous terms.”

257. But even this concession was a mirage. Although eliminating “last look” might have resulted in a small decline in transactions on Google’s ad exchange—partially offset by the benefits of Unified Pricing Rules—Google deployed a replacement that was effective in replicating the prior advantage. Relying on its massive trove of data from its monopoly publisher ad server and dominant ad exchange, Google developed an algorithmic model to predict the bids of rivals for each impression. In this way, Google could still predictively “peek” at its rivals’ bids before submitting its own. Specifically, Google’s unique access via dynamic allocation to years of data on multiple bids per auction for trillions of auctions allowed Google to estimate the distribution of predicted prices necessary to replicate “last look.” Google’s technical documents explain that for accurate predictions “[w]e need to predict the full distribution of competition, not just the point estimate.” Only Google had the data to do so by virtue of its dominant position across the ad tech stack. Moreover, the effects of years of a last look advantage did not disappear overnight. Google had already obtained the benefits of preferential access and continued to realize the enduring flywheel effects of scaling its ad exchange while eclipsing rivals.

258. Google delayed giving up last look for months while it fine-tuned the “Smart Bidding” algorithm that would replace it. When Smart Bidding launched, it fully offset the 30% drop in Google Ads’ revenue that Google expected from the loss of last look (without Smart Bidding) and turned an expected 10% drop in DV360 revenue into a revenue increase of 3%.

259. Beginning in late 2019, Google made some of the data used in Smart Bidding available to rival ad exchanges, but important limitations applied. Google only shares data with rival ad exchanges that bid on the impression via Open Bidding (not header bidding); data is limited to the winning price for an auction (where the rival ad exchange lost) or the secondhighest bid (where the rival ad exchange won); and data is shared in a format that many ad exchanges could not use. This is insufficient for rival ad exchanges to replicate the broader data trove to which Google has access and against which it was able to train its bidding algorithms for years. For their part, publishers have no ability to block Google from continuing to have access to this pricing data or to allow rival ad exchanges access to similar data outside the parameters dictated by Google.

260. At the same time that it rolled out Smart Bidding, Google degraded the data it made available to publishers that previously allowed them to monitor how Google’s ad exchange was competing against rival ad exchanges. Prior to late 2019, Google made available to publishers a “data transfer file” that allowed eligible publishers to see on an impression-byimpression basis the individual bids from competition among ad exchanges and certain advertising demand sources for the publishers’ inventory. Publishers could then respond to changes in the nature of competition by tweaking the way in which they made their inventory available to their ad tech partners. Commenting on the earlier version of the data transfer files, Google acknowledged their value was to create a “more transparent auction marketplace” and “enabl[e] publishers to find opportunities for incremental revenue.”

261. Following the shift to Unified Pricing Rules and the introduction of Smart Bidding, Google altered the files to prevent publishers from linking the bids from Google’s advertising products to those from rivals using header bidding for the same impression. This has made it more difficult for publishers to make informed choices about how and where to make inventory available and to monitor Google’s bidding behavior for potential anticompetitive conduct.