Trial Update, September 23: Google grounds its defense in games, gaslighting, and the government

It's the third, and possibly final week of US v. Google, and the first full day of Google's case-in-chief.

‍

Nitish Korula, formerly Engineering Director, Ad Serving & Ads Quality (now at Google Assistant)

Google called sell-side Ad Serving and Ads Quality engineering lead Nitish Korula to the stand. Again, Google started with superfluous questions about his education and personal background - he said twice he’s from a small town in South India, and that Google’s mission of organizing the world’s information spoke to him because the nearest bookstore was so far he would read his mom’s anesthesiology books. He is a generally likable and believable witness, seamlessly blending coached rhetoric (I almost filled my bingo card) with seemingly credible product insight. On cross, DOJ successfully highlighted the way that Google suppressed certain realities and impacts in the way they presented information in their demonstratives, and Korula was generally less evasive than most Google employees.

Google went through most of the sell-side products and features at issue in this case to explain how certain features and decisions benefitted publishers and/or advertisers. Google played their usual games: (1) they used a demonstrative that included non-public information not produced to the DOJ during discovery, (2) they obfuscated the timelines of feature roll-outs against the trends in the landscape - e.g. w/r/t adoption of RTB, and shifts to 1st Price Auction, (3) they presented the benefits of certain features like Last Look compared to not having Last Look, vs. a fair auction with all parties competing on equal footing in real-time.

With Korula’s engineering responsibility, Google took the opportunity to try to establish how difficult or costly it would have been to enable certain capabilities, like passing real-time pricing to third party ad servers, as the DOJ points out Project Yavin had enabled for eBay and LinkedIn’s in-house ad server. Google also prompted Korula to explain how several illustrations used by DOJ (e.g. First Look and Last Look) were “confusing” or “misleading” - again, an attempt to sow doubt that opens up room for Google’s preferred representation to fill the gaps. Judge Brinkema would jump in at points to ensure her own understanding is correct.

One such example is the conversation about Last Look, which Korula explains as the price sent to AdX buyers.  He says the reason is to provide buyers with as much information as possible, to help them - again jumping from protecting publishers from spammy ads just moments ago, to speaking to buyer benefits. He says Last Look  wasn’t “designed” - it was a “consequence” of min bid to win, and that it didn’t advantage AdX, because AdX isn’t bidding, buyers are.   Here, Judge Brinkema jumps in, asking to confirm the steps as follows: (1)  3rd party auction occurs, (2) highest bid goes to AdX, (3) AdX buyers can pay a penny more (if no other authorized buyers bid, Korula clarifies), and that within this process, people in Google buying systems can see what others bid (yes, but including open bidders, Korula clarifies).

Another semantics game, with Korula saying that between 2010 and 2014, DFP didn’t have real-time bids from rival exchanges, but Google “didn’t take action to prevent” this, the “mechanism existed.”  He points to Header Bidding, saying that when HB enabled RTBs from other exchanges, DFP didn’t need to make any changes. Yes, this is the same Header Bidding Google adamantly tried to kill. He also positions AdX as very exclusive. “Most” publishers don’t get access to AdX. He explains all of the workarounds publishers could use to allegedly accomplish what they wanted - e.g. static line items to use AdX without Dynamic Allocation. Google didn’t allow rival exchanges to bid into AdX, he explains, because it would be harder for Google to vet the buyers, and Google takes “high quality ads” on AdX very seriously.  This quality and safety argument surfaced throughout his testimony.

Sell-Side Dynamic Revenue Share, he explains as allowing transactions to clear where they otherwise wouldn’t have by reducing rev share. He says that it increased transactions overall, where publishers wouldn’t get paid.  There is a Product Requirements Document that says the goal is to increase publisher and Google revenue, and he says that yes, AdX saw more transactions, but insists again these are ones that generally otherwise would not have occurred.  He explained the evolution, where V1 allowed only reducing rev share, V2 allowed increase and decrease, and V3 determined rev share prior to calling buyers, which he describes as “incentive compatible”. Publishers could opt out of V2 and V3, he states. AdX saw +4.17% revenue increase, and publishers saw +2.8% increase. But, as we are reminded that Korula knew well that publishers were ok trading some revenue for more control. On cross, DOJ brings up the email we saw previously where Korula says that the 5th reason publishers wanted to set variable floors is to reduce dependence on Google, and diversify revenue. Obviously, this drawback of many of Google’s unilateral product changes was omitted from Google’s direct examination. There was also another tangled web visual to “explain” pricing rules before UPR - more complexity and gaslighting to create space for Google’s preferred positioning and justifications.

The DOJ was very strong on cross-exam, unwinding a lot of the re-positioning to reorient the Court. For example, Dynamic Allocation did in fact advantage AdX, SSDRS could in fact allow AdX to win for a fraction of a cent over HB, particularly in conjunction with Last Look, and Publishers did not have the option to turn off Last Look. They went one by one down the list with ease.

Korula apparently was quite good at turning his chat history ON after the litigation hold. So much so that others would start threads without him to turn history off, as we see in chats. There were some awkward chats in there, where Korula said that losing some ad server market share would be good to demonstrate to regulators that viable alternatives exist, and Google isn’t a monopoly.  He also said AdX direct is “a concept for antitrust.” Others included discussions about AdX buyers’ informational advantage via ability to bid with info out of DFP ranking, and that competition concerns will cause Google to give up part of their strategic advantage over HB - in reference to Last Look.  Imagine what we’d have found if all of the other Googler’s chats had been preserved.

Next we get 3 government-related buy-side witnesses that add a whole lot of nothing to Google’s case.

Sarah Stefaniu, Account Manager, Buy-Side, Google

Next, we hear from an account manager that served 3 rotations with various Google Ads advertiser groups including Amazon and Government accounts before moving into a role related to DV360. She walked through the Google Ads and DV360 user interfaces, hitting most of the other buzzwords on my bingo card - e.g. safety, fraud, etc. It was a colossal waste of time and hardly relevant to the trial. They were trying to get at market definition - again, hopping to the buy side to ignore the central issue here about publisher monetization.  I did appreciate the DOJ’s cross-examination, where we learned that PMax is “important” to Google because it is a “huge” revenue opportunity. $1 Billion in incremental revenue, Google said. I’m bummed this was only used to refresh recollection, and was not entered into evidence, but if Google keeps playing with fire with these buy-side witnesses, they will eventually get burned.

Kendall Oliphant, Chief, Contract Program Office - Communications Division, US Census

Keeping with the hardly relevant Government advertiser line of witnesses, Google calls Kendall Oliphant who basically manages contracting related to US Census communications efforts, including campaigns in 2010 and 2020 managed by agency team Y&R. Another colossal waste of time, and it is clear Oliphant did not want to be here. I’m not sure what exactly Google was thinking in effectively exploiting a government witness’ deposition testimony, but alas, that is what they did. On cross, Oliphant confirms that it’s not just open web display she has never heard of. She’s also never heard of a DSP, an ad server, or an ad network. She doesn’t need to, she says. That’s what the agency is for. Did Google call the agency to testify, however? Of course not.

Marco Hardie, Industry Director (Sales) for Gov’t and Advocacy vertical, Google

Marco is Stefaniu’s former boss - another Google buy-side sales guy for government accounts. We again don’t get very far here, and Judge Brinkema eventually, thankfully cuts it off, saying that the testimony is not advancing Google’s position on key issues in this case, and is starting to sound like PR. Agreed.

Ben John, VP of Engineering at Microsoft (Read-in part II)

We finished this read-in. I left a bit early, but much of what he spoke to supports the Government’s market definitions - i.e. Search and Display are not interchangeable, etc.  What Google is driving at is demonstrating that Microsoft is a competitor, that has Bing unique demand, and data from across its properties which it combines with Xandr 3rd party data.

Tomorrow, we hear from Professor Paul Milgrom, one of Google’s Experts.

‍