Day 9: The big finish

The Monopoly Report is going daily during the trial. Our intrepid founder is in a windowless room without a phone or laptop to get you the good stuff.

Day 9: The big finish

Today the DOJ substantially finished its case and we were expecting them to bring out the big guns. While Professor Lee from Harvard gave a solid analysis of the alleged monopolies, the star witness Jonathan Bellack turned out to just be a big old teddy bear who loves publishers and was forced to do bad things by the bad people.

Note: We are taking a couple of days off and will return with daily updates next week.

Here’s what happened.

Chalice is the founding sponsor of The Monopoly Report

Stuff we learned

  • AdX without Google Ads is really small.

  • AdX’s take rate without alleged monopoly abuses would be 16.2%.

  • There’s something economists use called a ”Hypothetical Monopoly Test”.

Plaintiff’s witness 28: Professor Timothy Simcoe, Boston University

This expert witness testimony started yesterday and today was the cross. In essence, he did an econometric analysis of the AdX take rate and estimated that instead of the current 19.8%, the rate without Google’s current alleged monopolistic behaviors would be 16.4%. This is a current analysis, so only includes the ties between Ads and AdX, the ties between AdX and GAM, and Unified Pricing Rules, and does not include first look, last look, etc.

Simcoe used two methodologies:

  • A comparables approach, where he analyzed data from seven other exchanges that provided raw log files; and

  • A events study, that he looked at raw data from Google for one year prior to the UPR/first-price launch and two years after to see increased wins by Google.

Google’s cross examination poked some holes in the methodology, but most of it comes down to the impossibility of separating the effects of the move to first price, the removal of last look, and the imposition of UPR, which were done simultaneously and which likely had conflicting effects.

Regarding the argument that the market was expanding across the board, here’s why that shouldn’t be a problem:

“Standard Oil was a monopoly, but the oil market still expanded”

—Professor Simcoe on stand

Plaintiff’s witness 29: Jonathan Bellack, formerly Google

There was a lot of anticipation for Jonathan given how often his name appeared in other witness testimony, and his key role as head of the sell-side at Google. Unlike most (really all) of the Google witnesses, Jonathan seemed relaxed on the stand, and appeared to be mostly answering the questions honestly.

The evidence showed that while Jonathan was often right in the middle of contentious debates, what he actually wrote was usually benign. For example, in the debates about header bidding, he expressed skepticism about the value of going totally open:

“I don’t think offering open RTB to everyone would end the options to play games with ad selection.”

—Jonathan Bellack, internal Google email

Another email from 2016 points out how small AdX is when you take out Google Ads demand, and thus how it can be at a disadvantage versus competitors with “lax policies”:

“GDN is half of AdX, and non-GDN is $700 million versus $2-2.5 billion on other exchanges. In other words ~75-80% of in-play demand is on other exchanges who offer lax policies.”

—Jonathan Bellack, internal Google email

The most interesting thing we saw was regarding Open Bidding. When asked about his goals for combatting/competing with header bidding, Jonathan made arguments to the effect that he wanted to create the best product possible for his publisher customers. But then an email was shown, sent to some of his team members, giving them guidance on how to present a proposal for “Jedi++.” Jedi was the code name for Open Bidding. And if you know anything about engineering, adding “++” at the end of something makes it better. Anyway, Jonathan wrote, seemingly in a recap of previous guidance, that the “goal of Jedi is to be ‘slightly better’ than Header Bidding.” When asked why that was the guidance (instead of creating the best product possible) he said that “slightly better” was the “guidance from leadership.” Who in leadership? What did you see Bellack!? We may never know.

Google must have though the testimony went well as they declined to cross.

Plaintiff’s witness 35: Professor Robin Lee, Harvard

We are skipping ahead to the interesting witnesses, we’ll cover #30-#34 below.

This is the good stuff. Professor Lee is the first expert to weigh in on whether we’ve got monopolies or not. He evaluated the three markets: Publisher ad servers, ad exchanges, and advertiser ad networks. For each he evaluated:

  • Whether it was a valid market definition;

  • Whether Google had market power in these markets; and

  • Whether they used their power to hurt competitors and consumers.

Professor Lee went into a lot of theoretical detail on his analyses, which I will not be repeating here. I’ll focus on the interesting bits.

Publisher display ad server

Definitely a monopoly, in the Professor’s opinion, based on some of these facts:

  • Building your own ad server isn’t realistic and is very costly.

  • Google was able to build features customers didn’t want, like UPR, and suffered no consequences.

  • From 2018 to 2022, of the top 100 publishers using DFP, only a single one left.

  • Using Google and competitor data, Professor Lee estimated DFP’s market share as 91% of open web display impressions worldwide, and 86.5% in the US.

It is worth noting that in yesterday’s email we presented data that DFP’s rate card had declined over time.

Ad exchanges

The Professor considered whether there were alternative ways for publishers to sell inventory other than an ad exchange. He noted that selling direct is a very different market, and that the average CPM on direct sold impressions in DFP was $7, compared to $1 for indirect. He noted that Open Path and similar SPO efforts from DSPs were very limited in scale, and other monetization options, like Facebook’s FAN have dried up.

Evidence for a monopoly in ad exchanges:

  • An internal Google doc estimated that the elasticity of AdX pricing across all publishers was <1, and for small publishers was zero. An elasticity under 1 means you can increase your profit by raising prices.

  • Based on a single day of raw data from GAM, he found that AdX wins ~50% of all auctions and that in over 60% of the auctions AdX faced zero competition.

  • Based on a simulation using raw GAM data, if AdX were to remove its demand, publisher revenue would decrease by 27.9%. No other exchange would reduce revenue by more than 0.5%

Here is the Professor’s estimates for AdX market share in 2022:

Impressions

Net revenue to exchange (fees)

Worldwide

56%

43%

US only

47%

36%

He also calculated that if you remove Verizon/Yahoo (which has since exited the market) these estimates go up by about 2%.

Advertiser ad networks

For the first time in this whole trial a witness talked a bit about search! Yes, Google Ads has millions of advertisers because of search!

On the buy-side Professor Lee made the point that the buyers in a DSP are really different from those in Google Ads. To do this he shared some fascinating data from Google:

Type of customers

Number of advertisers

Spend on display (2022)

Use Google Ads only

4.05 million

$8.2 billion

Use DV360 only

18,487

$359 million

Use both

22,700

$9.1 billion: $4 billion on Ads and remaining $5.3 billion DV360

Other items:

  • An internal Google study showed they could increase the GDN take rate from 14% to 15% and it would increase profit by 2.2%.

  • If you removed Google Ads from bidding, publisher payout would decrease 14.1%. The next highest impact was DV360 at 7.2%, and no other source was above 0.5%.

The Ad Network monopoly case seems to have been conducted largely from the sell-side point of view. The Professor asked which channels of revenue can replace publisher revenue on open web display. But from the advertisers’ point of view, there are Amazon, Meta, and a bunch of other replacements, which were not discussed. I imagine on cross a truck will be driven through this logical gap.

Plaintiff’s witness 30: Sam Cox, formerly Google (read-in)

Sam likes to think of himself as the most interesting man in ad tech, but his testimony didn’t live up to that billing. Nothing came out of this except for him defining display ads as “anything that is pictographic”, which is one way of putting it.

Plaintiff’s witness 31: Bonita Stewart, formerly Google (read-in)

Literally nothing was discussed other than her chat history.

Plaintiff’s witness 32: George Levitte, Google (read-in)

After the last couple of read-ins we were all expecting this to be uninteresting, but there were some very good bits here. George was the product manager for the buy-side of AdX, meaning tools to get demand into the exchange.

He divulged that total spend through Open Bidding is roughly $1 billion — a stat thishas not been divulged previously.

He was quoted in an email saying “Exchange [Open] Bidding is a better alternative to header bidding, more favorable to Google.”

There was a discussion of “Secret sauce”. Asked what that meant, George said there are some data that are the “secret sauce” and are reserved for Google’s use, not for buyers on AdX. For example, Google data on cross-device identification was used by Google’s buy-side but not by buyers on AdX. He said this was part of the strategy to increase DV360 at the expense of Authorized Buyers (the term Google uses for external DSPs buying on AdX).

Plaintiff’s witness 33: Woojin Kim, Google (read-in)

[Regarding expanding AWBid] “I was in favor of it”

—Woojin Kim deposition

Plaintiff’s witness 34: Michael Shaughnessy, Kargo (read-in)

Not much of interest, just that DFP has slowed innovation and that Google Ads demand cannot be replaced with Criteo or any other source.

What’s next?

The DOJ is done. Starting tomorrow Google will put on its case. They indicated they had 19 total witnesses, six of whom were advertisers of agencies. This could finish up next week potentially!

Reply

or to participate.