Day 7: What to do about AdX?

This is coverage of the remedies trial in US v. Google. I’ll be in Virginia writing every day on the courtroom activities.

Day 7: What to do about AdX?

The battle lines of the Google remedy trial have been drawn, and actually most of the fighting is over too. The trial is expected to wrap up Thursday and I’ll be able to move on from washing my underwear in the hotel sink back to my comfortable semi-retired existence.

Enough self pity! What was I saying about the battle lines?

It seems like the vibe of the trial is narrowing the range of likely remedies. If I was a betting man here’s where I’d put my money right now:

• YES: AdX bids into prebid

• YES: Various behavioral restrictions on working around this and diverting spend

• YES: 5-10 year monitoring

• NO: Open source auction logic because of complexity (see: Day 6: What's Up with the Open Sourced Final Auction? )

• NO: DFP contingent divestiture, mostly because the DOJ hasn’t made any case at all on this

• MAYBE: Escrow fund, hasn’t really been discussed much.

This leaves the biggest and most important decision in limbo. Should AdX be divested?

The Case for AdX Divestiture

Rajeev Goel, the CEO of Pubmatic, testified at length today about competing with Google, the various proposed remedies, and AI. The only interesting part, though, was when he talked about the limits of the proposed behavioral remedies. He said that even today he’s being blocked from getting demand by Google.

According to Goel there’s a publisher that wants to sell ads on Pubmatic through Open Bidding, and there’s some kind of “technical issue” that’s preventing any buyers using Google’s DV360 from accessing this path. This issue has supposedly been going on for eight months with no resolution while the DV360 dollars continue to flow into AdX at Pubmatic’s expense.

This type of friction is exactly what is difficult for a behavioral remedy to solve. We don’t know if Google did this intentionally or not. It’s likely not some kind of secret project they whipped up to increase AdX yield (at least we hope it isn’t). But the punchline here is that they have zero incentive or urgency to fix it as long as the money is still flowing through a different path.

(Also, as a side note, this is literally the first time “Open Bidding” has been mentioned in this entire trial, despite it being highly relevant to the remedies).

The other aspect that came up with Goel’s testimony, and that is critical to understand, is the role of AdWords demand in the competitive situation. Google’s proposed final judgement explicitly excludes any changes to AdWords as part of the remedy. As you may recall from a previous newsletter ( Day 5: Nothing happened but here are some cool charts ) over 90% of AdWords demand goes to AdX. If you don’t change that, and also don’t divest AdX, then competition would not return to the ad exchange market, though you could see how the ad server market might be loosened up a little.

The Case Against AdX Divestiture

We have heard from four different Google product and engineering leaders so far. Today started with the cross-exam of Glenn Berntson, and continued with the DOJ bringing his direct supervisor Noam Wolf. Nothing remarkable came of any of these witnesses other than the lawyers on both sides running to chambers to ask Gemini for advice on “how to effectively cross examine witnesses on the spectrum.” I kid, I kid…kind of.

The Google witnesses’ arguments all boil down to the effort of migrating being just too darn large. The combined DFP-AdX has 33 million lines of code, endless dependencies, global footprint, etc. Many of the witnesses intentionally misread the DOJ’s proposal as requiring divestiture of virtually all of Google’s infrastructure, up to and including the data centers! I think we’re all smarter than that. But also, the DOJ’s expert witness who calmly and confidently said it could be done in 18 months is also clearly full of shit.

Bottom line: yes its hard to divest AdX. No, its not impossible. And no one wants the Google datacenters.

Where’s the judge leaning?

Later in the day the judge talked about the “two elephants in the room” referring to the legally binding nature of a court order, and the threat of further civil suit. This comment, along with a similar one from last week, lends itself to an interpretation that the judge would prefer behavioral remedies.

But virtually every industry witness has stressed that behavioral is not sufficient, and I have to think that has some sway on her opinion. It’s a toss up.

Witness #18, Nirmal Jayaram

Jayaram gave what I found to be a highly humorous point of view from Google’s buy side. He argued that divesting AdX would be bad for advertisers for five reasons:

1. User data will degrade because they will be forced to use cookie matching (the horrors!)

2. Increased latency when the buy side and sell side are in different systems

3. Less innovation! You see at Google the buy side and sell side have innovated together. He gave the example of ads.txt, I would counter with the example of Bernanke.

4. Discrepancies

5. Fraud, malware, bad stuff.

Why is this list “humorous”? Well he basically explained why Google has enjoyed so many benefits by being on both sides. Its almost like he’s saying that the monopoly has benefited the business.