State of the Sandbox, July 4th edition

We were having a slow post-Cannes week when Criteo blew it all apart with the first in-depth Sandbox study. That, plus a great interview with CTV innovator tvScientific.

tvScientific is bringing performance to TV

The disruption of the traditional television business is creating new opportunities for innovation in video advertising. Foremost among the contenders is tvScientific, who is bringing the shocking concept of performance to the space.

Marketecture Vendor interviews are free for a week, then require a subscription. Read more about subscription options at marketecture.tv.

Podcast: Rob Leathern’s car is spying on him

Rob Leathern has held senior privacy and policy positions at both Google and Meta, so he was the wrong guy for a car company to mess with. This interview is insightful and a bit hysterical.

Listen to the pod now:

• Spotify

• Apple

• Anywhere you listen

State of the Sandbox

This week Criteo released a comprehensive and well-written set of results from its Chrome Sandbox testing, in which it concluded that the CMA should not allow the Sandbox to be released in its current form and offers various suggestions for moving forward.

Criteo’s findings were pretty bleak:

• Publisher revenue down an estimated 78%

• Latency up 100%

• Disproportionate market share gain by GAM

Let’s break this down and see what we know, what we don’t know, and how it is going to play politically.

Will the Sandbox reduce CPMs and cost publishers revenue?

It is an absolute certainty that the removal of third-party cookies and replacement with Topics, Poppi, and other Sandbox technologies will have a negative affect on CPMs and publisher revenue. We know this just both empirically and theoretically.

Empirically, we have the experience of Safari, which we know monetizes at a fraction of Chrome. We also have the Criteo study, and early results from Raptive.

Using just economic theory, we can consider the ad market in equilibrium. The removal of targeting is therefore a demand shock. With the same budget and goals, buyers will achieve lower ROAS on their ad spend, and will therefore reduce their aggregate demand. This happened with ATT on iOS, so we even have a near-identical case study.

Estimating how much publishers will suffer is difficult because:

1. Tests don’t really account for complex interactions between different sources of demand;

2. There are identity alternatives in the market;

3. There are very different types of publishers, and the effects will not be spread evenly;

4. There are likely to be second-order effects in the wake of a complete removal of Chrome 3P cookies.

The Criteo test results

Criteo’s write up of their methodology is very comprehensive and you should read it if you are into that kind of thing. I’m going to just break it down in plain English.

They ran a test-and-control methodology with three cells: 3P cookies, Sandbox, and just contextual.

Their bidder bid in each cell and measured “qualified site visits” as a proxy for conversions (and thus ROAS) because of a small sample size of conversions.

Their bidder stopped bidding when a threshold of diminishing ROAS returns was hit. Essentially their model says that at some point it isn’t worth bidding, and then they assume the budget moves elsewhere.

Based on this methodology they found that on publishers who have adopted the Sandbox API, the reduction in revenue was 60%. But since only 55% of tested publishers have adopted Sandbox, the gross decline across all publishers (including those without Sandbox) is 78%.

The contextual-only test group was a massacre, let’s not even go there.

Triangulating the decline in CPMs based on our very limited data points:

Real world vs testing

As we all know, the programmatic marketplace is incredibly complex and difficult to map. Thus, there’s a big uncertainty in how to use an isolated test, like the one from Criteo, versus how the world would really look on the day after the last 3P cookie was set.

Criteo’s study, for example, just looks at their bidder, without regard to the other demand sources bidding on the same inventory. The other bidders aren’t running a test, and are likely bidding higher, with more information available on each auction, causing Criteo to bid at a systematic competitive disadvantage.

In the real world, things get messy. There will be bidders like DV360, that will religiously follow privacy directives, while other bidders will happily utilize IP addresses, alternative IDs, “ID bridging,” and email hashing to re-enable 1-1 targeting, or a mix of such techniques. Some of these techniques are at scale now, while others are still ramping up — even DV360’s “PAIR” is pretty new to the market.

The net result of all of this will, IMHO, be a much lower revenue loss percentage than we see in an isolated test. How much lower, I don’t know.

As a side note, I’m not going to dive into the claim that GAM got a disproportionate amount of inventory out of the test since I don’t fully grok the flow of bids here and don’t want to say something incorrect.

Rich publisher, poor publisher

A month ago we introduced our Internet Advertising Map, a framework for evaluating the positioning of various publishers with regard to ad tech and identity:

One might overlay the potential revenue reduction on top of this framework, with those furthest to the bottom-left (least connection with consumers) taking the brunt of it while Walled Gardens get out scott free:

Keeping in mind this is a total SWAG, and of course varies by medium (CTV vs web), but at least this gives you something to think about.

Second order effects

The day after cookies disappear are all my ad tech Tweeps just going to go to Cancun and give up? Did we give up when Germany bombed Pearl Harbor? Hell no. A new equilibrium will be reached, then we’ll spring into action.

We should expect things to change post-cookie in ways that are a bit unpredictable, but which, in aggregate, could reverse some of the pain. Some possibilities:

• CPMs will be much, much higher for logged-in users, so publishers will get off their asses and make registrations happen;

• Sites without logins will die, which will fortunately include most MFA networks;

• Publisher consolidation with the strong buying the weak. There’s an arb opportunity if you have identity or a paywall and you buy a publisher that doesn’t (see the Times move into verticals);

• Buyer platforms will change ROAS to be more probabilistic, removing the bias against Safari traffic and raising the overall sea level;

• Other stuff we can’t really imagine just now.

The politics of it all

So after a really bad performance we’re going to ask the incumbent to change their plans and find a replacement. Sounds familiar.

The Criteo write-up is explicitly tailored to the CMA, which has already forced a delay in cookie deprecation. There’s a mistaken perception that the CMA is acting as a sort of fair arbiter of competition theory, when it is really an advocate for UK-based publishers. And if you look at the map, above, many UK publishers are in the bottom-left corner.

Finally, the most shocking finding in all of this is the 100% increase in latency. Google HATES latency, and if this is real and not a solvable problem, I could see the company itself moving away from this tech. I’d expect to hear from Google on this issue sooner than many of the other issues, since it cuts closer to their core.

Reading list

• TIME does licensing deal with OpenAI for 101 years (!) of archives

• TTD cuts off Yahoo video

• Sonobi and 33Across - both veteran ad tech companies with $30-35 million businesses are both for sale

• Claritas exploring a sale from within Nielsen

• Beehiiv acquires AI site builder. Beehiiv is quickly outpacing Substack, Wordpress, and everyone else to build a creators platform and ad network.

• Brian Weiser projects that digital will outpace TV for political ads this year for the first time every in a bi-annual election (subscription link)

• AppsFlyer IPO is coming