Attribution tracking on an alternate timeline

In Betting on Democracy: Notes from the Fight for “Article 88b”, Aleksandre Zardiashvili writes:

Whatever the outcome of the Google Adtech case, Google is already creating other opportunities to gain even more access to data, increase its power, and evade the rules. The next opportunity could be advertising measurement and attribution. If privacy signals succeed, advertising will need to be measured without tracking. This is why browser vendors and others are working hard to standardise a device-level attribution API called Attribution Level 1, developed with the editors being from Google, Meta, and Mozilla. In my view, there is nothing inherently wrong with on-device processing for advertising use cases such as attribution, but, of course, the involvement of Google and Meta raises many questions and invites criticism. Don Marti calls the group standardising the API an “attribution cartel”, and competition lawyer Thomas Höppner compares the situation to a fox guarding the henhouse.

While these concerns are real, the situation is not straightforward, and, to be honest, has given me a lot of headaches lately. As Martin Thomson (Mozilla) explains in a conversation with Alan Chapell, the API offers some privacy and efficiency benefits, and I believe something of this sort can support contextual advertising to make it more appealing to advertisers. The problem is not on-device attribution itself, but who controls the devices and for whose benefit the attribution API works. In no way does it make sense for Google to be the largest online advertising publisher and to measure its own performance against its own competitors. The largest online advertising publisher (Google Search, YouTube) cannot own device software (Android, Chrome) or the advertiser ad server (CM360). This is something for the competition authorities to solve, and the sooner we put this in front of them, the better.

Read the whole thing.

Internet optimism, antitrust department

What if Lex is right, and the problems with in-browser attribution tracking turn out to be just that a powerful tool is being handed to some existing oligopoly companies?

And what if, on some future optimistic timeline, the Big Tech companies are broken up, effectively regulated, or otherwise brought low, and legit sites are free to reap the benefits of a somehow cartel-free fair attribution system?

On that timeline, a bunch of good things happen.

• Google goes legit

• Meta gets shut down (either as the result of a$1.4 Trillion lawsuit, as much-delayed consequences of the genocide in Myanmar, or because of some other crime)

• Apple reverses its course of Tim Cook era enshittification, and stops turning its A-list hardware into delivery devices forsketchy App Store ads.

Also on that timeline, the attribution tracking gets handed over to some neutral, well-resourced entity or entities—attribution tracking is totally independent of any party with a stake in pushing search, social, and app store ads, and totally under the control of someone who will fairly report thehalo effect of legit ad-supported content as it pops up.

So, in that bright future, would all be well for honest attribution?

Of course, it doesn’t hurt to try. The global benefits of bulk-erasing Meta and all their works, of restoring Google to “create more value than you capture” mode, and of resetting Apple to work for the hardware customers instead of against them would be huge in a bunch of other ways, so of course I’d be all for trying the experiment. But I’d bet against it working.

The attribution cartel problem is more than a problem of empowering an existing cartel by giving them attribution tracking. Adding attribution tracking to some future, squeaky-clean, advertising business would create a new cartel, because of theBruner paradox. On the optimistic timeline, the opportunity to do what Martin Thomson calls “sniping credit” is still there. With the Attribution proposal in place, every player that could touch user data has the temptation and incentivization to do nasty surveillance in order to “snipe” or “front-run” and claim attribution. Some company you have never heard of that operates a demand-side platform (DSP) using data from a network of “smart” air fryers and tire pressure sensors would be in a better position to claim ad effectiveness than even the best legit site.

And that hypothetical unknown company has incentives to consolidate with some other companies that have different surveillance data, because machine learning and lobbying work better at scale. So the air fryer and tire pressure “sniper” merges with someone else who has smart TVs. Learning management systems.E-toilets. Pretty soon there’s a new oligarch in the business.

The problem is not just that the Attribution proposal works in favor of the existing Big Tech cartel. It does. The underlying problem is that the proposal is so cartel-friendly that it would bring forth a new cartel even if started in a situation without one. (Unless you could somehow ban all surveillance other than the attribution tracking itself, with totally effective enforcement.)

So what do we do instead, smartass?

I don’t know. In advertising, some kinds of data practices tend to support the creation of ad-supported short stories, news, bus benches, and other resources. Other kinds of data practices, not so much.

The future advertising measurement methods that turn out to be both acceptable to people who get advertised to, and rewarding to people who make the ad-supported content, will have to be the result of market design grounded in advertising history and user research.

Starting with cool math is a great way to get a publishable paper (which the Attribution proposal is, and it could probably be applied in other fields) but not the way to get where we need to go for ads.

Bonus links

OpenAI may have made a fatal misstep in copyright fight with news orgs by Ashley Belanger. Among the most shocking revelations, OpenAI allegedly pretended from the earliest stages of the case that it did not have the technical ability to search large anonymized samples of ChatGPT logs when it had actually already conducted such searches prior to the start of litigation, NYT alleged.

Ways to think about token pricing by Benedict Evans. (But is the LLM API market going to be affected by advertising and propaganda subsidies in the way that other high end vs. commodity IT markets were not?)

Dentsu strikes Meta deal to build plumbing for mass influencer activation by Sam Bradley (Another layer that needs to be addressed: even if honest attribution reports do get generated, will there be someone at the ad agency who can read them, or will future agencies be hollowed out andchickenized?)

Now, More than Ever, Madison Avenue Needs Unified C-Suite Leadership for Strategic (not Tactical) Transformations by Michael Farmer. The marketing and advertising initiatives of the past 15 years have failed to deliver anything other than lower costs. 40 out of the top 60 advertisers in the world have grown at only half the rate of nominal GDP growth for 15 years!

Google, AI, Oligarchy and the End of the ‘Open Web’ by Josh Marshall. Google is making a decisive move away from the open internet. They are building their own closed information garden and that’s a decisive shift away from the model that undergirded all of the company’s history down until the last couple years. Because Google is so big and has such a dominant role in the architecture of the internet, that’s a decisive shift for the future of the internet as well.