Google’s Witness Factory has Some Credibility Issues

Google's case comes into view, despite credibility issues with over-rehearsed witnesses. How could Google's Director of Ad Safety not be aware of a $100M ad safety penalty assessed by the FTC?

In one exhibit, Google’s average impressions are so much greater than its nearest rivals, that Google is a statistical outlier.

It’s Google’s world, and we’re all just living in it.

At least that’s the perception you might have been left with walking out of court on Wednesday, in the penultimate day of the US v. Google ad tech antitrust trial.

Last Friday, we were stunned at how awkwardly Google began its case. On Monday, we wondered if they had a case to present. On Tuesday, Google had a good day in court, muddying the waters on a few key lines of attack from the government. But on Wednesday, with 24 hours left in this trial, Google—with its legal strategy emerging like a surfacing submarine—made a few things clear (at least in its view).

First: There is no such thing as “open web display ads.” Google is the sole guardian of the ad tech ecosystem from the onslaught of spam, scams, and Bad Actors™ on the internet. And Google is a vibrant engine of economic opportunity for creators and brands, lacking even the terminology to talk about the anticompetitive conduct alleged against it.

While Google had a number of successes in court, it also faces some challenges, which I will outline below. More than anything, today had the feeling of Google’s lawyers throwing the case into a mixer with water, adding a few cups of saccharine Google blue, green, orange, and red powder, and trying to have the Court drink the Google Kool-Aid.

The day began with Judith Chevalier, Professor of Finance and Economics at the Yale School of Management.

Dr. Chevalier is a Google expert, and is the company’s counter to the government’s expert witness, Professor Simcoe; much like Google expert and final trial witness Dr. Mark Israel will be Google’s response to the government’s Dr. Robin Lee.

Dr. Chevalier spent a good amount of time attacking the findings of government expert Prof. Simcoe, setting up a sort of he said/she said of high economics.

Dr. Chevalier told the court that “Professor Simcoe’s study is simply invalid in this case.”

There’s another discussion to be had, which we don’t need to have here, about where these experts come from. When you dig into it, it turns out that much like you might have a shortlist in a kitchen drawer of local plumbers and electricians to call when you need help, large corporations and corporate law firms have shortlists of economists (or academics of other disciplines) who they routinely call upon to work backwards from a conclusion favorable to their client to create justifications that they can then come attest to. It is quite a revolving door.

The testimony then shifted to changes in quality over time. Continuing to go in on Prof. Simcoe’s report—Dr. Chevalier told the Court that his model “can’t distinguish between broken impressions or shifted impressions” and asked rhetorically “If Prof. Simcoe’s study showed a market expansion – how is [Google’s conduct] anticompetitive?”

This of course gets back to the by-now-tired “whole vs. slice” dichotomy that keeps coming up in the trial, and for which it might make sense to simply point back to Prof. Simcoe’s wise observation, that even when Standard Oil had a monopoly, the oil industry was still growing. In other words, this observation of a growing industry does not actually tell us much.

As Google utilizes economists to provide evidence to support its case, I still believe an even stronger Google case should have affirmatively challenged more of the government’s case with the same vigor as it is challenging these narrower aspects. For instance, Google could have presented evidence on how rival ad exchanges could in fact succeed, instead of running away from the issue entirely. Google could have presented publishers that praise the company (which Google technically did, but in a very limited and contrived way), as well as presenting happy Google advertisers (which Google tried to do on Wednesday, but once again, in an unusual, limited, and very narrow way, which I will get to below). If Google fully committed to fighting these battles, its case could be so much stronger.

Alas they have chosen a different strategy.

We then got to the cross-examination of Dr. Chevalier.

Dr. Chevalier struggled on cross and lost her composure. Perhaps because the government did a good job finding the gaps, but also because, quite simply, Dr. Chevalier, the William S. Beinecke Professor of Finance and Economics at Yale, is not someone used to having to cooperate with something like cross examination.

I jotted in my notes that she seemed like someone who went to MIT. I had never heard of Dr. Chevalier prior to this trial, knew nothing of her credentials, and so only when writing up my notes for this article and doing my research, I later found that she did in fact earn her PhD from MIT in 1993. There’s just this “I’m not saying you wouldn’t understand, but let me explain” Cambridge panache to people like this. Being from Boston, I could spot it anywhere in the world.

On cross, Dr. Chevalier denied that she was opining on Google’s market power, any geographic market, market definition, or whether Google’s conducts were anticompetitive or procompetitive. With a disclaimer like that, it sort of leaves open the question of what exactly it is that she is opining on.

She told the Court that she didn’t evaluate quality variables that were affected by scale. Recall testimony we’ve heard about how Google’s massive scale and the ensuing network effects directly impact quality, with respect to both Google’s inclination to innovate and improve quality, and the ability of rivals’ to improve their own quality.

At one point, when asked about whether AdX operates globally (remember she previously said on cross she did not opine on geographic markets) Dr. Chevalier seemed to shout “Yes” after a sequence of questions on cross that all had “yes” as the answer, and then asked to hear the next question again—seemingly frazzled, losing her train of thought, or both.

Professors at Yale don’t seem to like lawyers from Yale telling them to answer with either yes or no.

This was some of the best economics expert impeachment we’ve heard. It was clear, concise, understandable, on point, and efficient.

Dr. Chevalier then admitted one of her exhibits did in fact show AdX overcharging, and admitted this is a harm.

This cross examination appeared to reveal that her analysis was custom built to work backwards from a pro-Google position, and she at times had trouble explaining gaps, and answering with yes or no. The government highlighted that on Figure 8 of her report, a lot of the alleged competitors had orders of magnitude fewer impressions than Google did.

Moreover, the government then highlighted Figure 1 (above), which showed that Google’s average impressions are in fact so much greater than their next nearest rivals, that Google appears as a statistical outlier when trying to graph this data.

Cross examination really negated much of whatever was contributed by the witness on direct.

Dr. Chevalier then admitted on cross that UPR (uniform pricing rules) prevented publishers from setting variable price floors, which in turn lead to lower price floors, and on average, publishers would set AdX to have higher price floors. The theme here being that UPR drove revenue for publishers down.

The next witness was Courtney Caldwell, CEO of ShearShare, an advertiser witness called by Google.

This direct examination was basically just a Google ad. There’s no other way to put it. Ms. Caldwell apparently received (at least) a $100,000 grant from Google, her and her business are actually featured in Google’s marketing materials (in other words, Google holds out Ms. Caldwell as a sort of made-for-30-second-TV-commercial-higlighting-Google-Ads representation of its customers). If I were an ad agency assigned to make an ad for Google’s ad tools, I’d probably also pick someone like Ms. Caldwell. She hit all the buzzwords in her testimony that you might hear on Shark Tank. Her direct examination sounded search engine optimized for “how to run ads with Google as a small business.” It just felt like some of these witnesses were coming from some kind of Google witness factory.

It was here that we really began to descend into Google’s world. To embrace Google’s arguments, you have to embrace its vision. Google’s case, quite simply, feels like the inside of an ad Google would make for a Google Pixel phone, Google’s ad tools, or any of its other products. As I mentioned earlier this week with respect to the revolving door of lawyers and officials around Google, and the views that dominate in these circles, Google’s framings of the issues at play fully reflect these views. Google’s case has been sanitized of any notion of competition issues (remember, rival ad exchanges are largely and remarkably absent from Google’s case-in-chief, as if they simply do not exist).

Google employees keep referring to themselves and others on the stand as “Googlers” (One of the most grating terms that keeps coming up in this trial) and the whole thing has taken on the aura of the numerous works of fiction that depict this sort of Alice in Wonderland fall from reality and into some kind of bowdlerized Google fantasy land, where evidence, even the language used to describe competition issues, has been expurgated in the way a parent might childproof a home.

Moreover, these witnesses are so palpably in the bag for Google, if not merely extensions of Google itself, it hurts the credibility of the testimony. It feels like the campaign ads you see see at this point in an election cycle, where during a commercial break an actor who too perfectly depicts the tropes of a certain kind of voter blames one side for ruining their small business, only for the next ad to have someone praising the same politician for saving their small business.

The government brought in publishers who tore Google to shreds. Google brought in a publisher that hit all the company lines discussing how Google is great. The problem is the government’s witnesses seemed credible, were from large organizations with real life stories of how they are harmed by the company, and Google is not fighting that pound for pound—even openly having an advertiser witness like Ms. Caldwell who it has also given six figures of grant money to, and who appears in its marketing materials.

Some other key pieces of this testimony include Google continuing a trend of showing screenshots of ads placed through Google’s ad tools, and the same as campaigns displayed in walled garden environments (e.g. Instagram, etc.) as well as other places, highlighting how the ads, visually, look nearly identical.

This is obviously aimed at market definition—trying to argue that not only is there not a market for open web display ads, and if there is, it’s a thing of the past—but that the analysis is irrelevant because no matter what you call it, these ads display and function identically on many platforms, and thus can all be considered one and the same. Simply put, this is an end run by Google’s lawyers around the plumbing of the ad tech ecosystem in which it commits allegedly anticompetitive conducts, and a way to avoid the government’s open web display ad market definition, and case more broadly.

On cross, the government focused on the plumbing problem—that the witness could not buy the depicted Instagram ad through Google’s tools, so while visually similar, they are very different, and the market problem remains.

Google then called Adam Stewart, VP of Consumer Goods, Government, and Large Customer Sales at Google.

This direct examination was done by Justina Sessions. Ms. Sessions has had an interesting role in this trial. Unlike the triumvirate of Karen Dunn, Bill Isaacson, and Jeannie Rhee of Paul Weiss in Washington, Ms. Sessions is a partner at Freshfields out of their Silicon Valley office. She is almost always in court, sitting either at counsel tables or in the gallery, but has not had as vocal a role as the other attorneys.

In fact, today was interesting in that both sides went to their bench players. Isaacson, government lawyers Jeff Vernon and Aaron Teitelbaum, and a number of other blue chip lawyers were not in court, with some of the supporting cast re-appearing. It’s likely the case that the core of the dream team for each side is getting much needed rest in advance of the trial’s high stakes ending on Thursday (with rebuttal on Friday) and the showdown to come with Google expert Mark Israel.

This testimony continued Google’s shelling of market definition—showing ads looking nearly identical across platforms to advance the notion that these are all “display ads” (despite the process for advertisers placing them on these other platforms being very different, with web display ads necessarily involving Google’s waterfalls, pipes, and the allegedly anticompetitive machinations of its conduct.

Google touted “Performance Max” a tool to help advertisers more easily format ads to be displayed across platforms—working to further dilute a distinct market of open web display ads, and paint all of these channels as fungible.

Predictably, on cross the government distinguished open web display ads from the other ads shown, working to restore its market definition.

Before I get to the final live witness—I think Google is having some success here. Whether it’s enough to throw Judge Brinekma off the scent is a separate question, for which the answer is that it is unlikely, but Google’s sustained air campaign on market definition, the passing of time naturally softening once visceral memories of early evidence we heard in the trial that looked terrible for Google, is having an effect of some sort. The defense’s case does not make substantial evidence of Google’s wrongdoing and dirty laundry go away, but what once seemed quite clear, is now getting a bit brackish, if not muddy, at least with respect to certain elements of the case.

We’ve now settled into a predictable routine on this issue between the government and Google. If Judge Brinkema defines open web display ads as the relevant market, the government will most likely prevail. While Google has made a mess, I’m not sure they’ve successfully made market definition a toss-up like they have done with things like the waterfall and the more technical aspects of auctions, which others in the gallery today described as now feeling like a wash between the government and Google.

The last testimony of the day came from Alejandro Borgia, Director of Product Management on the Ads Safety Team at Google. This testimony sounded like a deposition read in. It was so obviously rehearsed, that it was in fact rehearsed too well. Mr. Borgia began answers at times before the questions were finished being asked, and the answers were of such empty industry jargon and boilerplate, that they sounded written by ChatGPT.

With that said, Mr. Borgia regaled the court with expansive but endemically shallow and detail-light accounts of how Google’s Ads Safety Team works to keep spam and scams out of the ad tech ecosystem, and how important this is to advertisers. The implication here is apparently that Google’s role as a hegemon in the ad industry is necessary to ward off the Bad Actors™, a term that would be used so much in this testimony, that if it still had any meaning given its chronic overuse in American political and legal life today, it was lost pretty quickly.

Mr. Borgia mentioned that Google’s “scale” is necessary for these important efforts, and used intentionally vague terms like “environment” when referring to markets or parts of the adtech ecosystem, a continued dilution of market definition on the part of Google. Judge Brinkema, always aware of what’s going on—asked Mr. Borgia what he meant by “environment,” asking him if it is referring to an app or a website.

On cross, this witness, given the very canned nature of his direct examination, was surprisingly quick on his feet. Cross examination was focused on showing internal documents that revealed Google’s policy enforcement efforts were far more limited than the picture Google paints—with only 2 spam related policies out of 148 total Google policies having active management at one particular time.

When Judge Brinkema asked the witness what he meant when he kept referring to “other means” of protecting against spam that Google used, he pointed to artificial intelligence and large language models that Google relies on to catch bad content, giving the example of a picture of a kid on a motorcycle with no helmet—something an AI could detect as dangerous and take down.

The government then asked Mr. Borgia if he knew that Google paid the FTC a $100m penalty for collecting data on kids to use for ad targeting—he said he did not, leading the government lawyer, in a mic drop moment, to ask him again, how, as the ad safety director at Google, he could not be aware of this.

This witness was robotic, and evinced all the dystopian corporatist qualities Google emits in its view of the world, and of this case. I think that these “security” arguments, while valuable in providing a pro-competitive basis for Google’s problematic conducts which can help allay antitrust problems, are not being explicitly connected up on that basis by Google’s lawyers, at least not sufficiently. Google would be well-served by completing the argument it is trying to lay here.

Finally, for observers of issues concerning Big Tech, online content, and regulation in recent years, there were a number of common threads during this security and safety testimony that have become awfully familiar. I was reminded of Murthy v. Missouri, a Supreme Court case that arose in the last couple years, when a number of states sued the federal government over what was revealed in the Twitter Files—specifically that different compartments of the government, largely in secret, directed Big Tech platforms what speech to allow or disallow, much of it related to online content concerning COVID during the pandemic. With the plaintiffs alleging unconstitutional coercion in violation of the First Amendment, in essence, censorship, the Biden Administration lost in the district court, and lost again at the appeals court, before getting relief earlier this year at the Supreme Court, seemingly against the evidence.

Many observers in that case believe the Supreme Court, on a wink and a nod, went along with the federal government with respect to fears that placing limits on the centralized power of Big Tech and the ability of the government and these firms to “cooperate” with respect to issues of “safety” or “security,” could have damaging unintended consequences.

These arguments at times sounded indistinguishable from what Google was arguing today, and what Big Tech, and officials who find themselves in this camp like to posit—that disintermediation of Big Tech firms (in this case, on antitrust grounds) presents too great a danger of unintended consequences by allowing spam (or misinformation, or Russian interference, or whatever the Bad Actor™ du jour happens to be) to not view this gatekeeping role as a more compelling interest than whatever anticompetitive harms may exist.

In short, “security” has become a favorite defense argument, but it’s unclear how much traction it will get given the weight of evidence that still remains stacked against Google.

Well that’s all for today! I’ll be back to sum up the final day of this trial! Thank you for following along for the last few weeks!