Trial Update, September 25: Google calls four live witnesses that either have been paid or still are paid by Google

It’s day 13 of US v. Google. Google continued to call the best witnesses their money can buy, and it still did not get them very far.

Judith Chevalier, Professor at Yale

Today started with take rate expert, Judith Chevalier, who has worked for Google as an expert on a number of matters. The intended purpose of her testimony was to poke holes in the analysis conducted by DOJ expert, Timothy Simcoe regarding overcharge, and the “event study” on UPR. This backfired, and DOJ was able to not only unwind much of her analysis, but also land some punches on cross that are detrimental to Google’s case.

She tried to show that Simcoe’s overcharge analysis was flawed and exaggerated. In part, she did so by using a “full stack average” depicting the average take rate by combining the average take of buying tools combined with the average take of ad exchanges, to demonstrate this number is higher than Google’s “full stack” average. This didn’t account for the higher take rate for Exchange Bidding. She also calculated different permutations of a full stack average by adding different combinations of DSP/Ad Network take rates with the sell-side tool take rates where she identified transactions flowing through those supply paths. The result was a chart that demonstrated clearly that Google Ads is the only buying tool that only connected with AdX, which the DOJ pointed out on cross-examination.

She also spoke to alleged flaws in Simcoe’s Event Study looking at the impact of UPR, which he expressed through impact on transacted impressions. Her perspective was also tethered to the assertion that any number of the bundle of 2019 changes could have resulted in the changes Simcoe observes, stating that because the move to a unified first price auction lowered transaction costs, it would result in more transactions. She spoke to charts demonstrating that the impact on revenue was minimal, and the DOJ clarified on cross that this would be the result of a decrease in CPM because of the lower floor prices multiplied by an increase in transacted impressions. DOJ has Chevalier confirm that Revenue = Price x Quantity, and while Chevalier doesn’t concede that this means using revenue understates UPR impact, she had already conceded earlier in cross that she didn’t look at the role of scale, impact of scale on quality, or the role of bidding data in the ad exchange market.

She also says that she didn’t see evidence that suggests that publisher sentiment was predominantly negative to UPR, pointing to the (questionable) Advertiser Perceptions study. She didn’t hear the audio from the meeting revealing UPR to publishers, she confirms. She did review depositions, but not trial transcripts. Given that the Court has heard at length about the overwhelmingly negative publisher sentiment to UPR, there is no way this landed well.  DOJ also gets her to concede that nobody forced publishers to set variable floors prior to UPR, so ostensibly the ones that did so, did so because they wanted to, and that she had seen evidence about the variety of legitimate reasons that publishers used variable pricing.

For the rest of the day, the Google called live witnesses as fodder for propaganda waterboarding.We heard from 1 “grow with google” SMB, and two Google employees. Including their expert this morning, that makes four live witnesses that have received or continue to receive payment from Google. We heard about how Pmax helps brands buy across channels, how great Google is for SMBs, and how robust their privacy and safety efforts were. A painful afternoon for those of us that operate in reality. I squirmed a lot. But, I can’t wait to get the transcript, because I’m sure I’ll be citing it in the future each time Google fails to do what they say they do.

Courtney Caldwell, CEO of ShearShare

I missed direct examination, but it was short. ShearShare is a small business that uses Google Ads, Firebase, and Google Analytics, and a Grow with Google participant. Naturally, Google wanted to land how much it does to help small business advertisers grow. On cross, DOJ highlights that ShearShare is part of the Grow with Google program, and as such is featured in case studies, receives money from Google, etc. Caldwell has used Adroll before, but didn’t know it was a DSP until trial prep. She says she used an ad server before, but didn’t know the name of Google’s ad server. She isn’t sure and doesn’t care if Google Ads bids on 3rd party exchanges, but would be supportive of it if it would increase ROI. A waste of time, and thankfully Google cut out the other SMB they planned to call, so clearly they thought so too.

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

Adam has been at Google for 18 years and is on the buy-side sales org for large customers. Google tried to show how many advertisers use both DV360 and Google Ads, land points about the competition they face on the buy side, and take another swing at the market definition piñata. Google shows how Demand Gen places feed ads that look similar to Instagram ads, and can run in mobile apps, including the WaPo mobile app.  Adam also explained how Pmax enables advertisers to buy across O&O and network inventory. He says Pmax uses AI decisioning to provide the greatest ROI and most effective and efficient returns. He is asked why he stayed at Google for so long and he says it’s because he can stand across from partners and see the impact of the work.

The (mis)infomercial ends, and DOJ sets the record straight. He works with advertisers, not publishers. Publishers cannot use DV360 to sell their inventory. He represents large customers, so the high dual-platform use is not representative of most advertisers, and particularly not the ones that comprise the AdX “unique demand.”  Google opened the measurement can of worms, so DOJ gets Stewart to concede that Google does not share data associated with logged in users with other ad tech entities.  DOJ pulls up documents comparing Google Ads and DV360 confirming that DV360 is positioned as different because, in part, it runs on 3rd party exchanges, while Google Ads is O&O and GDN.  DOJ did well, but I do hope they circle back to Pmax on rebuttal.

Alejandro Borgia, Director of Product Management, Ads Safety Team at Google

Another Googler from the Ads Privacy and Safety Organization. This time from the Ads Safety Team versus yesterday’s AdSpam witness. Borgia’s testimony was so incredibly rehearsed that it sounded robotic. This all fell away on cross-examination, as his tone went from polished and pleasant to outright snotty. He started with praise for Google’s mission, as nearly all of the Googlers have.  He explained all of Google’s efforts and investment in ads safety and privacy, and why it benefits users, advertisers, and publishers. He then gets to their three-pronged approach of policy, enforcement, and transparency.  He walks through the Ads Transparency center, as well as the My Ad Center privacy and preference interface. He talks about how publishers share demo, ID, usage, and other data in bid requests to 3rd party exchanges, where Google has no control of what is being used. The overarching theme here was that Google is only able to keep advertisers, publishers, and users safe when it has visibility on the buy and sell side.

DOJ did a great job at tearing down the facade. We see a document about how Google has 313 policies, but only 10% are actively managed, although the safety team believed that 58% should be. Of 49 inappropriate or harmful content policies, 19 were actively managed. Of 14 Scam policies, 2 were actively managed. Privacy was 1 of 28.  Legal risk was 0 out of 22.  Ecosystem cleanliness also had 0 policies actively managed.

Borgia tries to explain, and eventually is allowed to, that “actively managed” is only one type of enforcement mechanism, and refers to custom ML models for enforcement of specific policies versus use of a general purpose ML model. At the end of the day, as DOJ points out, the ads safety team believed that substantially more policies needed active management, and Google did not make this investment.  Further, DOJ asks if the amount of actively managed policies were disclosed in Transparency reports, questioning how transparent Google really is. Of course, they were not.  Google also had positioned it as though they invest in these areas simply to make the ecosystem safer. DOJ points out documents referencing increasing regulatory requirements.  DOJ asks if Google publishes the underlying data associated with transparency reports - e.g. what ads were removed, what pages were suspended from monetizing. Of course, Google does not. DOJ asks if Borgia is aware of Google’s $170M settlement with the FTC over childrens’ privacy. Borgia, who is part of the Ads Privacy and Safety organization, says he is not.

I left once they resumed deposition videos and read-ins, which started with the continuation of Kenneth Blom at Buzzfeed.

‍

Nearing the End

It sounds like expert Mark Israel is Google’s last remaining live witness. This will take up most of tomorrow, and they will then wrap any remaining deposition read-ins/videos. DOJ expects to start rebuttals on Friday. It is unclear how long these will last, and if Google will be allowed surrebuttals.

We did receive insight (tentatively) on trial timelines today, which I’ve included below.

‍