‘Unclear Effects’ Responses and Corrections

I was particularly interested to see two responses to my exploration of ad tech company financials post-App Tracking Transparency. Unfortunately the first, from Ben Thompson in today’s Daily Update, is not visible to me because I am cheap, though I understand it to be a criticism that I did not account for exchange rates. That is fair and also, though I lack the context to say for sure what Thompson’s argument, maybe loosely agreeing with the point I am making more broadly.

The second criticism comes by way of a comprehensive Twitter thread from Eric Seufert. I have to say thank you to Seufert for rebutting my work and pointing to corrections I should make.

I want to start with a study I cited which showed negligible benefits for publishers running behaviourally-targeted ads compared to contextual ads — that is, ads based on the website or subject matter of the page. Seufert points out that the findings of that specific one differ wildly from others. I regret not looking deeper into this study and appreciate Seufert’s highlighting of its unique quality.

Seufert also disagrees with my methodology of assessing revenues by geography:

But I also think simply looking at geo-level OS penetration rates in sizing ATT’s impact is misguided given the skew of value between the platforms.


No public data exists for this but in general, iOS makes up a higher share of revenue in EU than in the US, despite holding a smaller absolute platform footprint. This is likely especially true for Snapchat which, again, historically prioritized iOS. I find this logic faulty.

That is fair; I can buy this argument. Incidentally, I find this situation sort of ironic. Android is developed by one of the world’s biggest ad businesses, but it is not the best platform for ads — or, more specifically, advertisers.

On Meta, I think the amount of blame to ascribe to ATT remains murky. The amount of noise created by TikTok’s rapid ascendancy and its ability to take younger users and, therefore, ad dollars away from Meta is an astonishing coup. Is ATT really the thing holding back the growth rate of platforms like Facebook and Instagram, or is it more likely that big advertising dollars are following users’ eyeballs?

Where I have landed after reading Seufert’s thread — and which I recommend you read, as well, as I think he makes several good arguments — is that ATT certainly has an effect, but it is not as pronounced as assumed by its ardent supports and its biggest detractors. It is making advertising a little more difficult but there are so many things happening that it is just one of several factors, most of which are more significant. My argument was certainly on a side of ATT being less relevant, while Seufert’s — and, presumably, Thompson’s — paint it as more impactful.

Where my disagreement with Seufert remains is in the ethics of monitoring user behaviour and showing them ads based on that surveillance:

This is a common argument, and it amounts to: “Users are simply ill-equipped to provide informed consent given the esoteric nature of ad tech.” I reject this. I think users generally understand the trade-offs between ads personalization and utility & behave accordingly.

Seufert illustrates this with a story from an article published in August 2020, after ATT was announced but well before it was released:

Advertising influences the user experience; people respond to ads through clicks but also through broader product engagement. An anecdote that I relay often from my time at Skype: we conducted a survey of users to gauge receptiveness to ads being introduced into the product, and most respondents believed that Skype already included ads. “It’s a free product, so it must make money by showing ads to users?”

This indicates to me how little we can really absorb and pay attention to. We are all like this. There are reasons there are guardrails in just about every market for safety, quality, and consumer protection. The data marketplace, at least in the United States, is shockingly unregulated for the risks inherent to the industry. We do not have the time to fully comprehend the very many ways in which our everyday behaviour is being tracked, nor how this information is shared and repurposed. We have better things to do.

I also think advertisers have better things to do than to hyper-target thousands of ad variants, seeking out an edge that might not even be there. John Gruber:

There’s an oft-cited adage attributed to the famed Philadelphia department store magnate John Wanamaker: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” That’s the conundrum surveillance-based advertising seemingly solves. It lets advertisers know which ads generate which business, with high accuracy. It seemingly turned an unpredictable art into a very predictable science. And now, these advertisers are finding, allocating ad dollars is regressing back towards an unpredictable art.

A key word here, I believe, is “seemingly”. The data advantages of ad tech not entirely fictional, but the dashboards featuring giant numbers and beautiful charts are giving the impression of greater precision than they truly offer. It reminds me a little of the panic over email open rates last year when it announced Mail Privacy Protection in iOS 15. Ads are surely more precise than an old-fashioned system like email, but the numbers they present are often misleading or bullshit. There is an entire industry of companies that purportedly fight fraud and inaccurate metrics, but they are part of the problem as they perpetuate the myth that advertising can be damn near perfected — if only there were enough data.

You probably do not remember many ads you see on the web or on your phone. The world of digital advertising is a world filled with ugly garbage. While the technology side of digital advertising has tried to bring it toward a science, perhaps it needs to be brought back to something more artful. Less behavioural data may encourage this.

There is a middle ground between completely untargeted ads and the hyper-targeted ones advocated for by the massive industry behind them. More artful ads, placed contextually, are a good start. I also think Facebook, of all companies, actually had a fine system in place from its earliest days: it simply asked people what they liked. Because it was a place for twenty-somethings to chat, there was a social incentive to list your favourite music, books, movies, and TV shows. Marketers could buy ads for self-selected audiences. That seems entirely okay to me — it allows for well-targeted ads without strip-mining every interaction people have across their devices, every purchase they make, and every sensor they encounter.

Two stories about Apple and advertising broke since I published my piece last week. The first was about Apple’s attempts to woo Facebook into building a subscription service — which, given a fuller contemporaneous context, is not that alarming — but it also apparently felt entitled to some of its ad revenue from boosted posts. The second is from Mark Gurman reporting that Apple internally tested ad placements in Maps, and speculating the company may increase its ad products across its platforms. Both of these paint an ugly look for Apple and, when paired with its ATT strategy — which, in isolation, seem well-intentioned — will likely raise some regulators’ ire.

Privacy must be public policy. I got some things wrong in my article and I appreciate Seufert’s corrections. I am left feeling that ATT’s effects are more significant than I first thought but still less meaningful than many are eager to suggest. The biggest critics of ATT seem to believe it is a cataclysmic technology, but also that its effects have been mitigated for over a year because of strong economic factors, and that currency exchange rates are masking its more recent effects. I find this argument incompatible with itself. It seems to me these effects are still more minor than broader economic activity and changes in user behaviour. I am also glad to see we seem to agree on this last argument. If privacy truly is a human right, it should not be up to Apple and Meta to bicker over which company is entitled to elicit which compromises from its users.