Trial Update, September 26: Google’s defense would be going great, if we just pretend publishers don’t exist

We are nearing the end of US v. Google, and Google’s defense probably would have been better off not calling any witnesses.

Jessica Mok, Finance Director, Google

Day 14 of US v. Google started with Google Finance Director, Jessica Mok. Google tried to minimize the financial importance of its adtech business, showing that it operated at a loss from 2015-2017, to downplay the role of display in generating revenue, and land an out-of-context soundbite about how much they pay publishers. On cross, DOJ highlighted that Google tracks revenue separately for display, video and app inventory, and for direct vs. indirect-sold inventory supporting their market definition argument. DOJ also pointed out that Google decides whether to classify revenue as principal or agent, which informs whether Traffic Acquisition Costs (”TAC”), money paid to publishers, is included in their figures. Obviously, without TAC, booked revenue will appear lower, and margin will appear higher.

Mark Israel, President at Compass Lexicon

Google then called Dr. Mark Israel to play some more number games. He is an economics expert, and his analysis is central to Google’s case. While I expected something at least somewhat compelling, instead we got analysis that appears completely detached from reality. Google seems to want to pretend their publisher customers simply do not exist. While DOJ’s case focuses primarily on harm to publishers, there was a 30 minute stretch of time at the beginning of Israel’s testimony where his only mention of anything publisher-adjacent was to explain the origin of the “cellophane fallacy” as the theory that prices for cellophane might go up so much that people resort to wrapping their sandwiches in newspaper.

He starts by refuting the DOJ’s definition of 3 distinct markets, proposing instead that adtech is a single two-sided market. He says that DOJ excludes other places advertisers find impressions, other ways “connections” to audiences are made, and the other types of adtech that exist.  No mention of publishers.  He talks about shifting media consumption trends, highlighting growth of apps and mobile environments, as well as the growth of ad spend on social, and the rise of retail media.  Still, no mention of publishers.  He talks about how social and retail media companies act as competitive restraints on Google and serve as reasonable substitutes…again, with respect to advertisers. Google even pulls up a deck to make their point that compares Google and Amazon, and explicitly references the fact that Amazon does not offer a publisher ad server, while Google has DFP.

Finally he mentions publishers, to say that if ROI for advertisers were to fall, both advertisers and publishers might switch. For example, he says, publishers might move to a different exchange, move more of their content to apps, or adopt a subscription model. Soon after, he explains publishers can move more of their inventory from indirect to direct sales. It’s about “how hard sales people work” to sell direct. A few minutes later, he suggests that if Google’s publisher ad server isn’t good, advertisers might switch to Facebook. I literally wrote “What??” in my notes.

We then see how impressions won by other exchanges participating in Header Bidding increased since 2018 (a.k.a. in the years since Google has been under antitrust investigation.)  He then says that SPO is increasing. In his view, Facebook is an SPO offering, and everything is an ad server.

We also see how 86.3% of publishers don’t pay for DFP, which given the tying claims and evidence, hurts Google’s case more than it helps it. Again, we hear about how publishers can just sell more App inventory. He mentions that Chick-fil-a (not a publisher) offers free sandwiches for app downloads. Volume/reach considerations are nowhere to be found. Judge Brinkema points out to him that ostensibly some publishers do not have developers to build apps, and he concedes that for those publishers, there would be an incremental cost.  We then see some wild “everything is an ad server” calculations that render Google’s share of the publisher ad server market 38%.  He somehow included Reddit, Snap, Meta, and Amazon as ad servers.

Similarly strange math shows AdX share as 17% in ‘21 and ‘22. He counts anything that facilitates auction or direct sale as an exchange and cites publisher multihoming.  Advertiser ad network is next, showing  Google ads and DV360 prices decreasing,   and how most money on Google Ads comes from large advertisers, evading once again the DOJ’s assertion about all of the unique demand comprised largely of small advertisers that do not use DSPs. He uses a misleading representation of multihoming here showing 16% of advertiser SPEND through AdX, which again ignores the SMB point. He talks about the alleged procompetitive effects of Google’s conduct, and the benefits of integration, including allowing Google to keep advertisers and publishers safe. It’s all a vibrant, safe, competitive market in his eyes - if we just pretend publishers don’t exist.

After lunch, direct examination resumed. We are shown a chart where Israel modified Lee’s analysis of estimated revenue loss to publishers if they were to lose access to Google Ads advertisers. Lee’s estimate was -14% revenue. Israel’s highest estimate was -2.6%, as he (1) looked only at US, (2) took away only the bottom 99.9% of advertisers vs. all, and (3) added in direct-sold inventory.  Not sure how these changes were supposed to be legitimate, but it also excludes the knock-on effects like what we heard from other witnesses including Daily Mail about lower bids when AdX is absent from the auction. He then tells us AdX to DFP isn’t tying - it’s integration, and that forcing Google to deal with rivals would have a chilling effect on innovation. They tried to bring the shoddy AdvertiserPerceptions survey back again to prove Google was monitoring publisher sentiment about UPR, and it was shot down.

DOJ’s Teitelbaum absolutely killed it on cross-examination. He started with an impeachment line of questioning. Israel has been a paid expert on 40+ matters, not counting reports. Google is a repeat client. He makes 80% of his money/spends 85% of his time on expert witness work.  Israel has never held a tenured faculty position, and in US v. American Airlines/JetBlue, the Court found that he provided testimony that is not credible, “demonstrated a misunderstanding and misapplication of antitrust concepts, [and] rendered opinions based on false assumptions.”

Once that was out on the table, DOJ used effective analogies to deconstruct Israel’s analysis. He pulls up a pie chart that Israel spoke to about how many (>60%) advertisers use both DV360 and Google Ads, a bit less than half use just Google Ads, and a small percentage only use DV360. DOJ says that the same can be said for refrigerators and dishwashers. Just because many people have both refrigerators and dishwashers, does it mean they’re reasonable substitutes? No. Judge Brinkema says, what about stoves and microwaves? Israel says that if someone’s stove broke, they might be inclined to just use their microwave. He danced around whether this implies that they’re substitutes, and seemed to lose points with the Judge.   Now on to SmartPhones and Dishwashers. Just because spending on SmartPhones has gone up YoY, while dishwasher spending has stagnated or decreased, does it mean one is a substitute for the other? No, Israel concedes. Similarly, advertisers moving spend from one channel to another does not mean they are substitutes, right? Israel says it’s different. We learn that a chart showing FitBit ad spend and Comcast adspend trends across channels had underlying data about associated campaign objectives that Israel omitted from his spend shift chart. Not a good look.

We learn that Isreal’s positioning of adtech as a single two-sided market comprised of both buyers and sellers contradicts a position Google took in US District Court for Northern District in California. Judge sustains objection but allows it to be entered into the record and be part of closing arguments down the line. DOJ also gets Israel to concede that DV360 serves a different purpose than DFP. As does AdX and FB. As does DFP with TikTok. Advertisers can’t use TikTok to buy ads on NYT. He asks Israel if that means they don’t belong in the single 2-sided product market, and Israel dodges, so DOJ asks if that means his market includes things that are not reasonably interchangeable. We come back to this later as Israel calls them “complementary” - seemingly suggesting these products are both complements and substitutes.  Israel was saying “um” before every response by this point. DOJ posits: if advertisers shifting spend from one buying tool to another “disciplines” Google on the publisher side, as Israel suggested, how did Google still implement UPR?  DOJ also shows Israel that Meta, allegedly one of Google’s competitors, wrote internally about Google as dominant.

Now on to publishers, which were largely absent from Israel’s testimony on direct. Israel had named Reddit, Amazon, and Snap as companies that “switched” to their own ad servers from DFP to demonstrate in-house ad servers as substitutes.  DOJ reminds him that NYT had one and switched back to DFP.  NYT will still have ads to sell on its site even if advertisers change buying platforms, and even if they sell mobile app inventory.  We learn Israel (sneakily) used data from eMarketer that included social, in app, and native in its display definition for a number of charts. He says if it’s an eMarketer chart, unless otherwise noted, that it includes all of those channels/formats.  Israel didn’t do a Hypothetical Monopolist Test with substitutes - just prices. DOJ probes about what the right market definition is. Israel didn’t opine on “the outer limits” - so DOJ says, “what about broadcast?” Israel doesn’t have a great answer - just that he didn’t study it.

Another analogy: the Costco Rotisserie Chicken. DOJ makes the point that Costco takes a loss on the chickens to draw people into their stores. Israel concedes that yes sometimes companies will take a loss on one thing for a broader strategic benefit.  DOJ draws the comparison to DFP, where Google gets value from scale and data. He also gets Israel to confirm that procompetitive benefits of integration can’t be attributed specifically to the conducts at issue (e.g. First/Last Look, UPR, SSDRS, Poirot, Tying).

He asks if Google balancing the interests of advertisers and publishers (a cited benefit of integration) would mean that Google isn’t solely acting in advertiser interests. Israel says not solely, but benefitting advertisers will benefit publishers. So DOJ asks if buy-side only products have incentive to act in publisher interest, or just their buyer customers? Israel concedes that largely their incentive would be to focus on advertisers, highlighting the conflict of interest in the model.

We see a deck Google presented to DOJ touting that interoperability is key to its success, contradicting Israel’s points about the detrimental effects of requiring Google to deal. We see that Google had 1500+ product launches each year from ‘16 to ‘22 - each of which, Israel concedes, would have required some degree of technical work. It’s just their publisher customers’ requests that are too much of a burden, I guess.

We see charts of how crude oil production (Standard Oil), PC shipments (Microsoft), and Long Distance Calls (AT&T) all trended upwards at a time when those markets were monopolized. Digital ad spend shows the same trend, so does this on its own show that the market is healthy? Nope.

We see an email exchange with Neal Mohan and Scott Spencer about Larry Page’s cautionary green light to proceed with a business case to open up the network a bit - they needed to propose baseline controls about which exchanges to let in, and which competitors to restrict. Israel says this doesn’t change his opinion.

DOJ shows the slide from Israel’s testimony of all of the puffery by adtech companies about unique demand, and asks if there’s a difference between marketing collateral and internal business documents. Israel confirms. He’s shown LaSala’s email about how he is convinced Google can only maintain AdX 20% rev share because of the GDN unique demand. Does this change Israel’s opinion? No.   Israel says he didn’t know about Communicate with Care and the direction not to define markets or calculate market share, but this doesn’t change his opinion either.

We get a hypothetical about a small local newspaper switching from AdX, to illustrate the point that the alleged 2.6% revenue loss isn’t really 2.6%. It was an average, that doesn’t account for what would happen to individual publishers - specifically small publishers, and Israel is assuming most of this demand can be replaced. Israel says that specific publishers might want to stick to what they already use.  On redirect, Isaacson makes this worse by asking something along the lines of “how many small independent publishers are left?” “Fewer,” Israel says.
‍

Read-Ins

We move on to read-ins that also didn’t add much to, or actually hurt, Google’s case.

James Glogovsky, VP of Revenue Ops at NYTimes

Google wanted to land the point that NYTimes stopped working with IndexExchange and moved everything to AdX because they weren’t finding incremental value in other exchange partners. We also hear about all of NYTimes’ different ad formats. Unfortunately for Google, Glogovsky also says if it were up to him, publishers would be able to set variable floors, as it was useful, for example to weed out low quality ads. He also says that Header Bidding wrappers were useful, despite increased latency.

Omri Farber, Product Manager at Meta

I really don’t know what Google was thinking on this one.  Sure, they got some (cumulative) stuff in on market definition, and Farber said that Exchange Bidding was better than Header Bidding in some ways and that Audience Network a 2-sided market place. But he also said publishers stopped trusting Google, and said it would be anticompetitive to use 1st Look advantage to win auctions.  And he called TTD an SSP.

We’re THIS close to the end:

Google has one more read-in tomorrow, and DOJ will start rebuttals.Monday is the latest Judge Brinkema expects the case to wrap. DOJ submitted 5 potential rebuttal witnesses to Google. They do not plan to call all of them, and Google is pushing back hard against at least two.

‍