3 November 2021

The enduring, profitable myth of online advertising

By

The big internet companies – they know all about us, right? Facebook tracks us, evaluates our preferences and weaknesses and thereby knows enough to make us do whatever Mark Zuckerberg chooses, right? Google watches us search and read and like or dislike and uses its supercomputers to extrapolate a world where we are all but guaranteed to buy whatever they want to sell us, right? These companies know us better than we know ourselves…right? 

Actually, wrong. It is a curious irony that in an era where concerns over online privacy and ownership of data run ever higher, the big ‘internet platform companies’ – Amazon, Microsoft, Apple, plus Google and Facebook and a host of smaller lookalikes – don’t really have the knowledge and the capabilities that critics ascribe to them. Yet it is in their interests to make it seem as if they do have these satanic powers, because by making it seem so they can sell advertising space to the companies that crave such knowledge. 

In the world of marketing the knowledge derived from customer data is called ‘attribution’. Attribution is the foundation of the astonishing growth of ‘Big Tech’. Attribution is the reason that Facebook’s ad revenue in the second quarter of this year was $28.5 billion (while revenue for other things like subscriptions was less than $0.5 billion). Attribution is the reason that Google is the biggest advertising channel in the world (Google’s revenue in 2020 was $182.5 billion, of which over 80% came from advertising). 

Yet attribution is a myth. An enduring and very profitable myth but a myth all the same, based on a lot of techno-marketing hot air. It is the deepest of deep fakes.

To get under the skin of attribution we first need to look at advertising before Google. The old agency world of Madison Avenue hasn’t gone away: the ‘creatives’ still dream up the ads, but they have been humbled and shrunken by the growth of digital marketing. The world’s biggest legacy advertising business, WPP, has revenues in the billions, but despite trying to muscle in on the ‘data-driven’ digital marketing world their sales are only a tenth of Google’s advertising revenue.

There’s an adage from that Mad Men advertising world which is old but no worse for that. John Wanamaker was a 19th century US retailer, philanthropist and latterly politician, and in his day was celebrated as one of the great proponents of the power of advertising. Yet he is actually best remembered for making a devastating critique of the very idea of advertising. ‘Half the money I spend on advertising is wasted – the trouble is I don’t know which half’ he is reported to have said.

Wanamaker may not have used the word but he was pointing to the problem of attribution. Spending a lot of money on different sorts of advertising seems to help sales. But which bit of advertising does which bit of the work has always been very, very difficult to establish. 

Now fast-forward to the arrival of Facebook, Google and the rest. The digital companies arrived with an offer that advertisers could not refuse: they promised to nail down advertising attribution once and for all. Digital marketing tracked through online page views and clicks and internet buying claims to open up the motivational machinery of advertising and make a direct attributional link between ad spend and sales. And that is why the Big Tech companies are so big (and so rich).

A new branch of the marketing industry has now sprung up around attribution, and as you might guess it has its own doctrines and its own impenetrable language. It’s a good rule that wherever jargon proliferates there is usually a vacuum being filled, and so it proves with advertising attribution. The marketing-speak serves only to obscure the fact that in practice attribution that connects advertising with identifiable individual actions is impossible. 

Let us for a moment leave aside the fact that since the implementation of the EU General Data Protection Regulation in 2016 (and similar US legislation) it has become increasingly difficult for companies to collect personally identifiable information and even more difficult for them to pass it on for marketing purposes. But even in a free-for-all unregulated data environment the information that comes from browser or app clicks on ads and shopping carts is not all it seems.

Marketers call such data streams ‘deterministic’, because they are an objectively visible record and supposedly linked to real people with commercially exploitable histories. In the world of marketing intelligence deterministic data are supposed to be the premium hard stuff, and everything else such as offline advertising on TV or in hard copy publications or on billboards is known as ‘probabilistic’ (i.e. less valuable). 

Yet deterministic data are not really linked to people. The data are linked to browsers, or mobile devices, or apps, or digital robots, which can be masked to appear something that they are not. In any case half or more of all online ad clicks are thought to be either mistakes (damn, I didn’t mean to make that ad pop up), or fraud (robots designed to increase revenue for pay-for-click providers). Even where ‘deterministic’ online actions are deliberate they are only weakly linked to individuals, individuals who are only notional entities in customer data terms.

In fact, there is no real value difference between deterministic and probabilistic marketing data. It all lies in the realm of probability, and trying to calculate the probability that some advertising content had its desired effect in terms of pushing an individual towards a certain action is about as profitable as trying to count all the leaves fallen this autumn. 

And you don’t have to take that as just my opinion: a peer-reviewed academic study in The Journal of Economics concluded that calculations of advertising impact on sales cannot be made with any meaningful level of confidence, and that selection bias due to the nature of product advertising would probably invalidate the results anyway.

That pretty well disposes of digital attribution as a way of saving Wanamaker’s lost 50% of ad spend. Don’t expect that to inhibit the Big Tech ad giants though: machine learning and cognitive computing will no doubt provide new ways for them to persist in quantifying hot air and charging handsomely for it.  

Click here to subscribe to our daily briefing – the best pieces from CapX and across the web.

CapX depends on the generosity of its readers. If you value what we do, please consider making a donation.

Donations via the website are temporarily unavailable

Richard Walker is a journalist and communications adviser to financial companies.

Columns are the author's own opinion and do not necessarily reflect the views of CapX.