Is there an end in sight for ‘people-profiling’ online?

• Google’s data sucking days may be numbered
• There are signs that publishers are turning to context, rather than personal data, to add allure to their platforms

Programmatic advertising based on personal information may be peaking and ready to slide back down into the depths from whence it came (warning:  I am very partial on this topic). These are the ads that can chase you around from site to site or from video to video, even though they may be irrelevant or out of date. These are the ads that are placed automatically and can find their way to you most effectively if they know all about your likes, dislikes, habits and vital statistics.

But ‘programmatic’ actually covers much of the digital advertising you see online. Only the most objectionable form uses personal information to refine your profile. The question is, does it even work? Doubts are surfacing.

Just as one example - many people have similar anecdotes - I have a particular cluster of Japanese ads for hotels and bullet train rides that have been stalking me since 2015. I have no current plans for Japanese travel. That’s seen as a growing problem.

Let’s roll out the other side of the story

Programmatic proponents can list a whole collection of pluses, mostly for the media ‘industry’ rather than you or society at large.

They talk about granular targeting and the ‘audience of one’. This is about making best use of advertising spend to “get the right message to the right person at the right time”. Unlike  conventional ‘scattergun’ advertising, which doesn’t give the advertiser much feedback (except the sales figures), programmatic can collect data about its own performance so that campaigns can be refined.

But targeting consumers not on their data profile but on the keywords and contextual relevance of the page they are viewing can also be highly effective.

Their main assertion is that programmatic is a significant cost-saver for the advertiser, since the targeting options mean that campaigns can become more efficient with advertisers only paying for ads going to users that fit a particular audience profile. The only way to build the profile, it’s argued, is through personal information gathering.

However, most of the huge cost-savings involved in digital are not due to the programmatic targeting as understood above, but through the simple fact that the marginal cost of a digital ad is next to nothing... literally.  And even that tiny amount is getting smaller as network and cloud costs lower. The significant costs here come from content generation, ad sales and marketing. Things like paper print costs - content production, printing, mailing etc - have gone.

There still needs to be targeting, but that can be inferred by context.  In other words the user makes content choices which select the ad, rather the ad selecting the user through his/her personal profile.

This is how advertising worked, and still works, on paper products. So in the digital world if the content of an online magazine is about fly fishing, then it’s a very good bet that scattering very cheap ads (remember there is no marginal cost) through several fishing magazines will get more users hooked than will a smaller number of targeted ads.

That’s one counter-programmatic argument, but does it hold up?

It might do if programmatics based on user profiles using personal information were outlawed. If we accept for the moment that the profiles do (probably slightly) improve the targeting and therefore reduce the cost-per-ad by a certain amount, then removing personal profiling and relying on various sorts of context (title, specific content subject, time of day or time of year etc) would see ad supply and demand define a new level, perhaps involving more ads at lower cost.  

And it would overcome the privacy problem too.

So what’s happening?

There is some evidence that ‘some’ content owners may be moving in this direction.  That scourge of Trump, The New York Times, has  decided to eschew Europe’s digital ad exchanges because of the extra hassle occasioned by the introduction of Europe’s General Data Protection Regulation (GDPR). It is opting instead  for contextual and geographical targeting only for its programmatic advertising. It claims that, at this stage, it has not seen ad revenues drop as a result.

The NYTimes might be an outlier, thanks to the Trump push-back effect.  Rather than ‘failing’, as Trump would have it, it’s been booming.

But there has been other anecdotal evidence playing out over a longer time-scale, indicating that many users, at least, are now wary of information gathering and will take steps to prevent it happening to their click-streams, especially in the wake of the Facebook data scandals.

Search engine DuckDuckGo has been serving keyword-based ads (no user profiling or data collection) since 2014 and has kept on growing right through period attracting more investment as it goes.