Day 9: Can Competition be Brought to the Market?

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

Day 9: Can Competition be Brought to the Market?

Readers gave me hard time because yesterday’s newsletter was titled “Day 7” for the second day in a row. Look people, after spending almost two weeks in the Hilton in Alexandria, I have no idea what day it is.

Today was supposed to be the end of the trial, but we’re extending into Friday for the DOJ’s final rebuttal. We expect to hear two of the highlight from the liability phase, Steph Layser and Matthew Wheatland, both representing the views of publishers. It will be interesting to contrast their points of view with the smaller enterprise views from wikiHow that I highlighted yesterday ( Day 8: Did Google Just Win? ).

Today I want to focus on the testimony and implications of expert economics witness Andras Lerner, who evaluated the remedies on behalf of Google. It all comes down to which remedies restore competition and which do not.

Witness #23: George Levitte (returning for Google this time)

George Levitte had already testified for the DOJ and was now brought back by Google. He was only on the stand for maybe 15 minutes, and it was specifically to rebut the story that Pubmatic CEO, Rajeev Goel had told about problems he was having with getting demand from Open Bidding (see Day 7: What to do about AdX? ).

The way Rajeev told the story there was an unknown bug that was causing DV360 to not buy a specific publisher’s inventory through Open Bidding, while that same publisher was monetizing fine through AdX. The implication (which I also made) is that it wasn’t just explicit preferences that ties together the exchange with the rest of the stack, but it could also be subtle indifference.

Levitte gave a lot more detail about this very issue, and like many things in ad tech it is a rat’s nest of interdependencies. I barely caught it all, but basically a) viewability measurement for in-app uses a specific technology; b) that tech wasn’t working in Google’s SDK on native ads specifically; c) also GAM didn’t work properly in this circumstance; d) also DV360 didn’t work properly in this circumstance; and e) also Pubmatic didn’t work properly in this circumstance.

Putting a bow on this, Levitte claims its all fixed, being tested, and that Rajeev is getting regular email updates on the status. It’s at times like this I’m very glad to not be running an ad tech company.

Witness #24: Heather Adkins

Heather is a VP of security engineering at Google. I found her testimony vaguely interesting but of no real relevance to the case, so I won’t spend time here on it. The one thing I’ll say is that we should all be thankful that there are Google employees like this keeping our data safe. God Bless America 🇺🇸 🦅.

Back to Dr. Lerner’s Testimony

Yesterday I gave short shrift to Dr. Lerner’s economic testimony. He talked about a lot of issues but I want to hone in on competition. Both sides’ economics experts have roughly agreed on the goals of a remedy:

1. Stop anti-competitive conduct

2. Resolve competitive harms

3. Prevent future monopolization

4. (Google’s expert only) Ensure benefits outweigh costs and risks

Dr. Lerner split hairs on #3, above. He said that the remedy should only prevent future monopolization in the same way as was found in this case, not in other ways. This hair splitting is critical since it opens giant loopholes in the remedies.

Lerner outlined what economists call the “But For” world, which is what the world of Google would look like right now if they hadn’t done the bad stuff. They would still have AdWords, still have the dominant ad server DFP, still have AdX, still have vertical integration, etc. He concluded that since those things would exist anyway, there’s no reason to spin them out. Case closed!

When asked specifically about AdWords on cross-examination, he argued that since AdWords was not found to be monopolistic in the liabilities phase, it couldn’t (and shouldn’t) be restricted in the remedies. When counsel followed up on whether it would be OK for AdWords to directly bid into DFP, he agreed it should. Which, as the careful readers out there should realize, would instantly make all the remedies in this case totally moot.

Let’s take a couple of steps back, and ask ourselves what we’re exactly trying to accomplish with the remedies.

The DFP Ad Server Monopoly

Step 1 is we need to allow publishers to switch ad servers if they want. I know I’m dumbing this down but after two weeks we’re in the fog of war its time to get clarity.

Why can’t publishers switch ad servers? Overwhelmingly we heard that AdX only bids into DFP, and AdX is the only way to get AdWords demand. There’s two steps, but its still pretty simple.

The easy solution, that both sides mostly agree to, is to make AdX bid in real time into prebid. This breaks the tie between AdX and DFP since you can implement prebid in any ad server. There are some important details, though.

• If Google did what Dr. Lerner suggests, and skipped AdX and bid into DFP directly, publishers still couldn’t switch ad servers.

• Google said they only want to do this for open auction, not PMP. This means that if you use PMPs in AdX you can’t switch ad servers.

• Google’s engineering leader said on the stand that he wanted this to be a publisher option, not the default.

• There are a ton of ways Google could cheat to advantage the AdX-DFP connection, and the DOJ has listed a bunch of behavioral remedies to avoid this.

Bottom line: Prebid plus a bunch of refined behavioral guidelines gets this done.

The AdX Exchange Monopoly

The exchange monopoly is a funny thing. Publishers work with many, many publishers. Dr. Lerner’s analysis showed 66% of the top 100 publishers on DFP use 10 or more exchanges! Yet in this “hyper competitive” market (as Andrew Casale called in the liability phase) AdX has something like 8 times the market share of the number two party. How did AdX get outsized share and what can be done about it (and who does it help?)?

AdX has a lot of “unfair” advantages over other exchanges by tied to the hip of DFP:

• Near exclusive demand from AdWords, and large share of demand from DV360 (see Day 5: Nothing happened but here are some cool charts )

• Integration with DFP for tags, data, workflow, contracts, etc.

• No discrepancies with DFP

• Cross-subsidies with DFP (get DFP for free, etc)

• Programmatic Guaranteed only works with AdX

• Enhanced Dynamic Allocation gives AdX the ability to win highly valuable impressions other exchanges can’t

• Efficiency advantage by not auctioning impressions that cannot be won (reserved)

• Unified Pricing Rules (“UPR”) which took away publishers’ ability to drive traffic to other exchanges

While this is a long list, the first bullet, AdWords demand, is disproportionately more important than the others. As we’ve learned AdWords + DV360 represent 66% of all open web display impressions on the entire internet. It’s almost remarkable that AdX has so little market share, they should have driven everyone out of business by now.

In Google’s remedy, and in Dr. Lerner’s testimony, there would be no change to the bidding behavior of AdWords and DV360, except it would now go through prebid. Dr. Lerner proclaimed on the stand that this, plus the removal of UPR, would immediately bring competition to the exchange space and result in lower pricing. This is just wrong. While removing UPR might shift some spend, the majority of AdX’s revenue and market share comes from AdWords demand, so if you don’t change that..it won’t actually change.

In the DOJ’s remedies, on the other hand, they address this with what they call “Preferential Routing” or the anti-SPO:

There are two problems with what the DOJ is asking.

First, we all know that making decisions about which exchanges to buy from is very important, and reflects many factors around quality, cost, etc. Basically this remedy is a bit unrealistic and will actually hamstring the Google team.

Second, it will almost certainly make publisher revenue decline. There’s more than a monopolistic reason why Google buys from its own supply. They also get data and latency advantages when buying from AdX that are necessary to their bidding. I’ve spoken to current and former Google employees about this issue, and they really can’t bid as effectively on other exchanges. Your reaction might be, “cry me a river, Google needs to act like all the other demand sources.” I get that. But if spend goes down by 10%, by my calculation that’s $600 million from AdWords and $360 million from DV360 that is siphoned directly from publisher’s revenue, and isn’t replaced by anything.

If you spin out AdX, you eliminate all of the integration advantages bulleted above and you make the demand spread out naturally (since AdX is a different company now). But, you likely put the AdWords demand into even more peril as barriers of bidding into a second company reduce ROI for advertisers.

Bottom line (assuming NO AdX spin):

• AdX still has a ton of meaningful advantages by being part of DFP

• If the court doesn’t do something about AdWords demand this is all a waste of time

• DOJ’s proposal for AdWords demand will have unintended consequences

• Take rates might go down

• Revenue might go down

Bottom line (assuming YES AdX spin):

• Real competition in the market for ad exchanges

• Take rates very likely to go down

• Possibility for significant disruption in revenue from Google’s demand side

Looking forward

Tomorrow is the last day of the trial. 😢