DOJ Proposes Multi-Stage Spin-Out

Walking through the proposed Google remedies

The audience of this newsletter is likely divided into two personas:

• Please, give me more about the Google antitrust trial; vs

• OMG, stop talking about the Google antitrust trial.

Based on my conversations at the Possible event, there are more in the first group than the second, so I’ll indulge them this week.

Remedies timing

Last week the parties to the ad tech antitrust trial proposed a timeline for the remedies phase that stretched through the end of the year. Judge Brinkema, who never likes to waste time, called the parties into a hearing to discuss speeding things up, but ended up endorsing the original timeline. I’ve reproduced it below. You may note that today, May 5th, we’ll see the written proposals from the parties about remedies.

The three-stage GAM spin solution

The most interesting thing that’s come out so far is the DOJ’s proposal for making divestiture and competition in the ad serving and exchange spaces possible. We’ll learn a lot more about their proposal today once its in writing, but I’ll do some speculation based on what was said in court.

As a very, very brief overview of the court’s findings, Google has been found to have a monopoly in publisher ad serving and in open web banner ad exchanges, and of tying the two together. The DOJ’s remedies are aimed at breaking those monopolies, while Google’s argument is that any such structural remedy will be disastrous for publishers.

With the caveat that I was not able to attend the hearing and that I’m going off of those who were like @BalanceCrafting (her thread) and @jason_kint (his thread), it seems like DOJ is proposing the following:

1. Spin out AdX

2. Force AdX to bid into GAM through header bidding instead of integrating into the ad server

3. Open source GAM’s ad serving brain

4. Divest the rest of GAM a couple of years later

Let’s walk through what this step-by-step proposal might mean and where it falls short.

Step 1: Spin out AdX

On the surface this seems like the easiest part of the remedy, but actually might be the trickiest. There are so many nuances to an AdX spin-out that I’ll just bullet them out so I don’t need to write coherent prose:

• I assume Google ads (AdWords) would now be forced to bid into AdX at arms length, without any preferences or special accomodations. Would the court mandate or oversee this relationship or will they just assume that once spun into a separate company, the Ads team will work in their own best interests?

• Does “AdX” include “Open Bidding” / “Exchange Bidding”, which is integrated into GAM? If not, I assume the court would have to force a shut down of Open Bidding, or else it is just reproducing AdX through a different path. Shutting fown Open Bidding is probably good for everyone as it is a source of duplication and it never got the adoption needed to replace the header.

• AdX is a leader in Programmatic Guaranteed deals, presumably because they can use the ad server’s forecasts as an input to delivery. This needs to be rebuilt from scratch, and will likely not work as well.

• What would the spun-out AdX be worth? It transacts about $10 billion in media, but that will certainly decline without its current advantages.

• Will AdX switch cookies from ads.doubleclick.net? Will it need to rebuild all of its identity alliances and cookie syncing on a new domain?

Step 2: Move the AdX-GAM integration into the light

Ever since Scott Spencer at DoubleClick invented Dynamic Allocation, AdX has has the ability to determine the winners and losers of every auction in a non-transparent way. While the bids from other exchanges came in through the header with transparent pricing, the AdX bid, its reductions and fees, and the ultimate decision about which ad to show, have been within the Google servers. While Google claims that all of the unfair shenanigans like Bernanke, etc are behind them, the process is so complex and opaque that it seems unwise to trust, and nearly impossible to police.

The DOJ’s proposed solution has a lot of elegance to it: make AdX move to the header. This is deliciously ironic given how much the Google ads leadership despised header bidding. To quote former head of DV360, Payam Shodjai, “the problem with header bidding is it exists.” Like many disruptive innovations, the “worse” solution ultimately won.

One effect that the court might not anticipate is that by moving to the header, AdX will be on the same competitive footing as other exchanges for latency. This is a hidden advantage of AdX that the DOJ and other complainants have overlooked. The reality is that Google Ads, when bidding into AdX, is much slower than other demand sources, and that AdX gives Ads extra time to respond with a bid. This is one of the reasons it took so long for Ads to start bidding into other exchanges through the “AWBid” project. Without the latency leniency Google Ads will win less and publisher revenue will decline.

Step 3: Open source the algo

This one caught me by surprise and honestly I’m scratching me head a little at the rationale. The DOJ seems to be proposing that the algo inside GAM that is responsible for ad delivery be open sourced to allow other competitive ad servers to get to parity and allow for more competition. This is puzzling because none of the publisher or competitive witnesses at the trial claimed that GAM’s algo was a reason they couldn’t switch. Also its not clear that GAM’s algo is substantially better than competitor’s algos, or that competitors have a desire to match the algo to compete.

Here are some of the actual barriers to switching publisher ad servers, once AdX is no longer part of the problem:

• No discrepancies with buy-side GCM (f/k/a DFA)

• Shared identity and data with GCM, Ads Data Hub, Google Analytics, etc.

• Effort required to retag websites

• Forecasting tools and interruption in forecasting when switching

• Economic deals that include YouTube

• Lack of easy access to API to facilitate switching ad servers

Step 4: Divest the “rump”

This seems pretty hand-wavy. The DOJ is saying that after a couple of years of Google running GAM with an open-source algo and no AdX integration, it would somehow be easier to divest and separate what’s left. There’s no mention of whether “what’s left” would still be valuable, or what incentive Google would have to maintain and staff this cuckolded product with no growth or future. Also left unsaid is how to use these years to prepare for divestiture by moving the technology over to the public Google cloud and removing the ties with other Google products. Perhaps more details will come in the filings.

Finally, it seems obvious that this “rump” would be worth very little, existing in a declining state with core IP open sourced and a prohibition on building any transactional capabilities.

Editor’s note: If the DOJ is reading this, call me.

As a reminder, this reporting is all pretty speculative as we’re going off of oral hearings. We’ll learn more this week.

Reading list

• Meta earnings - strong, forecasts strong q2 despite tariffs (link). Zuck talking big about AI driving growth. Also released standalone AI app.

• Google will make PMAX more transparent - channels, search terms, creative assets (link)

• Comcast’s Universal Ads brings on more publishers (link)

• AdExchanger piece on take rate pressure on TTD (link)

• OpenAI enabling shopping in ChatGPT (link)

• Apple can’t charge for web purchases! (link)

• Swivel (fka Pilotdesk) Series A (link)