Auction manipulation: what is fair?

This is the latest post of my new Marketecture newsletter: The Monopoly Report. Each week we will cover everything relating to Google’s ad tech antitrust trial. Subscribe for free here.

Over the coming weeks we’re going to explain and analyze the allegations against Google from the Department of Justice’s complaint. This week we’ll look at the ways Google is accused of manipulating the auction to benefit themselves.

Context: Second price auctions

During the time period in which most of the allegations take place, both AdX and most of the rest of the digital ad industry were operating using second-price auctions. In theory, with these auctions, the winner is supposed to bid what they truly value the auction to be worth, and then have their bid reduced to $0.01 greater than the second bid. This is how search ads work.

The programmatic display market didn’t work out quite the same way as search, and as a result there were inefficiencies, and opportunities for unfair auctions. The biggest factor was that many auctions lacked bid density, that is there was only one bidder, and no second-best auction to use for reduction. Unsavory platforms might not reduce the bids as expected in order to take additional margin. Or they might artificially add bids to increase prices.

The market, including Google, has since moved to first price auctions, but that change happened subsequent to most of the allegations discussed here.

Multi-Bid

The programmatic marketplace is complex and opaque, as platforms on both sides may have bidding logic that is not obvious to their counterparties. One results is that the highest bid on a given auction may not always win. For example, an exchange may have a poorly communicated set of advertiser block lists, or quality controls they don’t disclose to buyers.

The logical way for buyers to account for some of this “noise” in the auction is to pass more than one bid to the exchange when they have more than one buyer interested. If Bid A is rejected, then Bid B might have a chance.

But there’s a problem! What if Bid A and Bid B are the only bids on an auction, and the presence of Bid B serves to “second price” Bid A and thereby increase the clearing price. Buyers don’t want bid density, sellers do.

Astute readers can see where I’m going here. What if you happen to run businesses on both the buy-side (Google Ads & DV360), and the sell-side (AdX)? Do you want more bid density to raise prices or less to lower prices? Hmm.

Cutting to the punchline, Google regularly sent two bids from Google Ads into AdX, 85% of the time according to a 2013 study cited by the DOJ. They could do this because Google Ads is priced on clicks, conversions, and other outcomes, so the end advertiser isn’t able to interrogate ultimate CPM prices or take rates. As long as advertiser outcomes were achieved at acceptable prices, the margin could be increased in Google’s favor, thus padding prices and winning volume for AdX against competitive exchanges.

Ironically, it isn’t clear that this two-bid setup was actually optimal for Google’s profits. The DOJ further quotes a Google study showing that publisher revenues would drop 30-40% if they moved to single bid, but that Google Ads profits would increase by 50%! Presumably this is because they could win a lot more auctions for advertisers at lower prices and still achieve those outcome goals.

Who are the winners and losers here?

Google Ads Advertisers: 📉 Worse ROAS for their outcomes and fewer outcomes for their budget.

AdX Publishers: 📈 Higher clearing prices.

Competition: 📉 Lost market share for demand and supply.

Google simply could not risk a change that weakened its ability to keep publishers locked into its publisher ad server and ad exchange. --—DOJ complaint, page 60

Dynamic Revenue Share

Another advantage to being on both the buy- and sell-sides is the ability to choose where on the demand-supply curve Google wanted to execute transactions. When Google Ads bids, the bid level is calculated by projecting the value of the auction, and then adding a margin. When calculating the winning bid in an AdX auction, a similar bid-reduction is executed, in order to account for the sell-side fee. Combined, these two fees could be ~30% or more.

Google would reduce the ~30% fee as needed to increase the volume of auctions it won. If the auction was going to clear at $1.00, and Google’s highest post margin bid was $0.98, they could simply increase the bid to $1.01, win the auction, and take a 3% lower margin on the transaction. (my math is wonky, this is just illustrative).

Google Ads Advertisers: 📈 More wins, higher ROAS.

AdX Publishers: 📈 Higher clearing prices.

Competition: 📉 Lost market share for demand and supply.

Bernanke

Project Bernanke was named for former US central banker of the same name because it created a “bank” for moving money around between auctions.

The idea is kind of simple. When AdX sees a competitive auction where the Google Ads demand was not going to win, they would increase the bid in order to win, and call it a credit from the bank. Then when Ads demand was going to win an auction without much competition, they would reduce the bid less than it should have been reduced, generating a debit to the bank. In Google’s reasoning, as long as the money balanced out at the publisher level.

Google Ads advertisers: 📉 They bought mis-priced inventory, in some cases getting significantly over-charged (the DOJ says up to 50% margins taken).

GAM Publishers: 😐️ Unclear. While the banking calculations tried to assure publishers were left whole, they may have lost demand from third-party sources as Google Ads sucked up more auctions than it should have.

Competitor exchanges and DSPs: 📉 The unfair auction meant that demand from other parties won auctions less than it should have, reducing market share.

Global Bernanke

For fans of the show Breaking Bad there are a lot of discussions about at which point Bryan Cranston’s character really broke to never return to his old ways. Well, if you think that Google’s actions in Bernanke, above, are sort of fair and balanced, since publishers ended up making the same amount of money, I have some bad news for you…

In 2014 Google implemented “Global Bernanke” which removed the key banking constraint that each publisher would ultimately make the same amount. Instead, Google would debit and credit across publishers, choosing winners and losers. And the losers were called “non-competitive publishers” and were chosen based on their perceived inability to switch away from GAM/AdX. In effect Google was taking money from publishers is disfavored and giving that money to more loyal customers.

Project Bell

One of Google’s key tenets is to never get dis-intermediated from its consumers. That’s why it built Chrome and Android. In the publisher ad tech world, the enemy was “first look” deals. Companies like Criteo or Amazon would work with publishers directly to buy their best inventory before it was ever subject to an auction. Google absolutely hated this.

In 2014 Google extended Global Bernanke to take into consideration whether the publishers were giving other party’s preferential access to inventory, and simply reduced bids by up to 20%.

“…only a company like Google with substantial market power across the entire ad tech stack would have the incentive or ability to implement such a program. Project Bell both insulated Google’s ad exchange from this new form of competition and preserved preferential access for buyers on Google’s ad exchange, including Google Ads.” -- DOJ complaint, page 72

I’ll leave it up to the reader to decide what is a fair auction, and what is not. Next time we’ll be diving into the epic story of header bidding.

This is the latest post of my new Marketecture newsletter: The Monopoly Report. Each week we will cover everything relating to Google’s ad tech antitrust trial. Subscribe for free here.