Day 3: Judge Asks About Breaking Up Google

Judge Brinkema asks what would happen if Google were "blown apart." Plus another ex-Googler witness, a strong expert for the government, and an industry executive weigh in.

Day 3 of the US v. Google ad tech antitrust trial began with Judge Brinkema pausing the proceedings to remind the Court of today’s date—September 11^th, and that the only suspect ever tried in court for the attacks was tried in this very courtroom. When reading stories about antitrust trials, it’s often difficult to remember that courts are places, and judges are people, with this particular contest just one of many in a wide variety of criminal and civil matters. The gravity of consecrating that day was a good reminder of the reach and import of the judicial system.

Ok, that said, before I get to the witnesses, I want to highlight the most eye-raising moment of the day, when Judge Brinkema asked what would happen if Google were “blown apart.” In this trial, internal Google assessments of spinning off its adtech business are in the exhibits, with explicit arguments about the value of integrating its lines of business. Here, for instance, is one fairly damning email.

Brinkema seems well-aware of the market dynamics, and is already considering remedies. That’s a far cry from what looks like a somewhat drawn-out remedy phase in the search trial overseen by Judge Amit Mehta.

And now, let’s go on to what happened. After the acknowledgment of September 11^th, Brad Bender, an ex-Google Product Lead, was called to testify.

Witness: Brad Bender, former Google Product Lead

Mr. Bender was initially deposed in 2020 while working for Google and was deposed again this past June.

An interesting note: A lawyer for the government asked Mr. Bender if he holds any Google stock worth more than $1 million, which he agreed he did.

Mr. Bender began his career at DoubleClick, an ad tech company acquired by Google in 2008, following which he began working for Google itself before leaving in November 2022.

After taking the Court through the branding history of Google’s ad offerings, focus turned to a litany of emails involving Mr. Bender that contained some troubling excerpts for Google.

In one, Google touted the importance of the high switching costs of its publisher ad server, DFP, a theme which we’re beginning to hear from several witnesses. Switching costs refer to the frictions or costs that publishers encounter when publishers try to move their ad inventory from one server to another.

High switching costs have the effect of increasing Google’s power in the market for publisher ad servers. The more difficult it is for publishers to switch to a competing server, the less sensitive Google is to competitive pressure. In subsequent testimony and exhibits, reference was made to “artificial pressure keeping prices high,” that “if there’s a better network or exchange, you can switch it” but “switching is a nightmare—it takes an act of God.”

Mr. Bender was technically a hostile witness to the DOJ, so his testimony also sought to correct for some of the earlier testimony less favorable to Google. Bender alleged the now infamous NYSE quote (“we’re both Goldman and the NYSE”) was merely a “metaphor.”

The DOJ showed Mr. Bender another email, in which he claimed that he “really believed” that “if we can execute on this stuff, we’ll be able to crush the other network and that’s our goal.”

Mr. Bender tried to distance himself from this, saying he disagrees with that earlier assertion, and it is not how he actually felt.

“Historically a deal would be done between a publisher ad server and an advertiser…when it comes to advertisers coming to Google’s network, we help set a different expectation—that [Google] would help marketers and advertisers appear on safe, reliable inventory,” Mr. Bender described.

Generally, Google has done a better job of arguing its value proposition to advertisers than to publishers. Recall, when publisher ad servers sought to get around Google’s dominant ad exchange through a technology called “header bidding,” Google scrambled to shut it down. In one Google email thread that included Bender as well as Jonathan Bellack of Google, the government highlights a key quote: “The problem is not so much PBM buying header bidding, the problem is header bidding exists :) “

There was then a relatively spirited back and forth between the government and Mr. Bender over a heated pretrial issue—Google’s destruction of evidence in the case.

The government questioned Mr. Bender as to whether he turned chat history on or off. He contended that he did not change it from its “default” but was elusive on specifics. Government counsel and Mr. Bender then tangled over whether he had “substantive discussions” via workplace instant messaging.

Mr. Bender said (and would repeat multiple times verbatim during the course of his testimony) that “substantive communications” would occur in “an email, [in] a document, or in a meeting.”

While this is merely my opinion, this answer seemed coached. Mr. Bender has been deposed multiple times, over multiple years, and is now testifying in court, having been represented by Google and outside counsel in each of these instances. Google’s destruction of evidence and failures to save chat histories after being notified of a litigation hold are notorious—causing both judges in each of the antitrust cases against to rebuke the company’s counsel, with Judge Mehta in the search antitrust case mulling sanctions over the issue, and Judge Brinkema blasting Google in this case over the same failure to preserve chats, and considering drawing adverse inferences (that is, to assume information Google failed to provide must have depicted the company in a negative light).

This is not the first time Mr. Bender has had to synthesize an answer to this issue. His answer—that “substantive communications” would occur in “an email, in a document, or in a meeting” in essence means that evidence is either in an email (and accordingly if discovery did not turn up the email, it does not exist), likewise with a document, and in the event of a meeting, much like a telephone call, is not recorded and would not be found in any event. In other words, Mr. Bender’s answer is designed to turn the attention away from Google’s troublesome chats, and frame the situation such that any evidence that’s missing, never existed.

At this point, Karen Dunn walked into the courtroom, fresh off helping Kamala Harris prepare her statements for the presidential debate—unseen in Court since giving Google’s statement on Monday.

On Google’s cross of Bender, Google presented an internal Google report from April 2018, projecting into 2020 that Google’s display market share would be flat, and slightly down.

Next, Google put up an exhibit showing who it claims (as of Wednesday) to be its alleged competitors—an amorphous list of corporations that has evolved during the trial, and now includes Verizon, AOL, Tencent, Baidu, Comcast, Apple, Twitter, Oracle, and Adobe, among others. Of course, this is a central dispute of the trial. And while Google argues that it is in vibrant competition with those walled gardens, none of them offer access to open web display advertising. That’s almost all Google.

Witness: Professor Ramamoorthi “R.” Ravi, DOJ expert witness

Next, the government called Professor R. Ravi, of Carnegie Mellon University to testify on the research into “discrete optimization” and auction design that he consulted on at Google. We’ve already heard a lot about market definition. Professor Ravi spoke largely to Google’s illegal conduct.

What might not otherwise come through, and should be emphasized, is the extent to which Dr. Ravi truly carried himself as an expert. His voice boomed, unlike most witnesses to this point who regularly are told to speak up. His recall was tremendous, his mastery of the subject matter was outstanding, he was never flustered, and his deep knowledge of data science, mathematics, and programming was readily apparent.

On direct examination, he told the Court that upon working with Google and reviewing their programs and data, he found that Google’s ad tech was designed to advantage Google and disadvantage non-Google participants in the ad tech stack.

Dr. Ravi told the Court how when Google’s DFP was used as an ad server, Google AdX bids first, which affects publishers by generating less revenue for them since the highest bidder is coming through AdX, without the opportunity for other bids to be considered.

Demonstrative exhibits depicted waterfall bidding structures where later bids, even much higher bids, could not win.

A September 2017 email was presented, in which a Google ad executive “emphatically” wrote to reject spinning off DoubleClick/AdX because “launching AdX into a non-DFP server destroys this competitive first look advantage.”

Dr. Ravi was then questioned about “feedback loops” and network effects, establishing that since AdX wins more valuable impressions, less valuable impressions fall to other exchanges, reinforcing Google’s dominant position.

The Court was then on the cusp of what could have been fascinating testimony: Yesterday, Google made use of a 2017 News Corp. study that showed revenue could go up even when downranking Google’s services.

The government wanted to get Dr. Ravi’s assessment of this report.

However, Google lawyer Bill Isaacson objected, and after a back and forth, Judge Brinkema ruled that because Dr. Ravi had reviewed this information only after Stephanie Layser attested to it yesterday, and not in the course of his work with Google, his testimony regarding it could not come in.

Dr. Ravi discussed how after the advent of header bidding challenged Google, “First Look” became “Last Look,” where Google was able to have the last word on ad bidding, “peeking inside the envelope” as I detailed yesterday.

Judge Brinkema interjected and questioned the witness herself at numerous instances, which, during Dr. Ravi’s extraordinarily technical testimony, revealed that the judge has a strong, thorough, and incisive understanding of the subject matter at hand. It is important to note this, as so often in our mahogany-paneled halls of power in this country, understanding the abstract nuts and bolts of various alleged high-tech excesses can often prove elusive. Brinkema seems to be following along just fine.

Dr. Ravi spent the better part of the afternoon in deep, dense technical analysis of the economics of these transactions. I noticed that across the press section of the courtroom, pens were down.

The cross examination of Dr. Ravi was an impressive intellectual stalemate. Dr. Ravi simply knew the material too well, and could not be successfully led down a path of questioning that allowed for any leverage towards making points useful to Google.

At one point, Mr. Isaacson asked Dr. Ravi if he “did any analysis of the scale Google contributed to its rivals.” After Dr. Ravi almost automatically shot back: “That’s just the flip side of the coin of the ads it did not take for itself.” Google had no further questions.

Witness: John (Jed) Dederick, Chief Revenue Officer, the Trade Desk

Next was John (Jed) Dederick, Chief Revenue Officer at the Trade Desk, a demand side platform, and the primary competitor to Google’s DV360.

Mr. Dederick’s testimony was valuable in that it reinforced anticompetitive effects elicited from other testimony, such as the fact that only through DV360 (Google) can an advertiser bid on inventory appearing on YouTube and other Google-owned properties.

On cross examination, Karen Dunn returned to the lectern for the first time since opening statements. She came in heavy on Mr. Dederick, who held his own.

At one point, Judge Brinkema asked an eye-opening question—a hypothetical of what would happen with respect to scale if she were to order Google to be broken apart as a result of this case.

Mr. Dederick said that publishers would not let inventory go unsold; in other words, the marketplace would adjust. The fact Judge Brinkema is playing out remedy scenarios for breaking up Google should she rule against it, is huge.

Mr. Dederick testified to how confused he was when Google made changes following the success of header bidding, as bid prices began to make little cognizable sense, as Google manipulated its system and algorithms to try and derive a new advantage.

The day ended with some of the most heated (but collegial) disputes over evidence between counsel we have seen yet. Ms. Dunn, and other Google lawyers, wanted to admit a May 2024 company report from the Trade Desk that opened with a statement that the industry is “at a tipping point” where less than half of all digital advertising spending is going to Google and Facebook, into evidence. It's clear why Google would want to use this—it could be utilized to depict the twilight of Google’s alleged dominance, and impeach Mr. Dederick’s testimony about Google. It’s also unclear what other revelations were contained within the report itself.

The government’s lawyer for this witness rose to his feet, formulating objections. Ms. Dunn, speaking quickly, attempted to preempt objections and offered an exhibit number, while also trying to lay a foundation with the witness. This stretch had the feeling of an NFL team trying to hurry up and snap the ball before officials had time to review a questionable play.

Judge Brinkema however called timeout before Ms. Dunn could get the play off, and upon objections that the report was produced after the deadline for producing evidence, that Mr. Dederick was not the writer of the report and that these were not his opinions, ruled it could not come in.

There was a clear sense Google had something considerable planned prepared for this report.

To close out the day, counsel sought to introduce evidence under seal. Judge Brinkema ordered everyone out to discuss whatever that was (likely trade secrets) and it is unclear when we will be able to return tomorrow.

One final thought that has been lingering in my mind: Something that’s become conspicuously absent from the trial, is the ads themselves. Of course, the trial isn’t necessarily about the content of ads, merely the economics around them. But it is surreal from inside the courtroom to see the presentation of the trial—where ads in this case may as well be stocks, or “widgets.” It becomes plain that ad tech has become so far removed from ads, that the industry is far more in the business of financial engineering than it is in advertising. Do these ads even work? Are advertisers even selling more because of them? The closest we get is the sanitized term “ad impressions,” seemingly synonymous with views, used ad nauseum.

That’s all for Day 3!

UPDATE: This is Matt Stoller. I edited this piece, and wrote the original sub-heading. I’ve changed it for obvious reasons.