On Europe, bald men & shampoo ads

Image created by me with Midjourney

Imagine you're bald (not a wild stretch of imagination for yours truly). Now imagine you are getting ads for shampoo all the time. What a waste right? Waste of precious timeline real estate for you, waste of money for the advertiser.

Now imagine a world where you can exclude people who are bald for your shampoo ad with your campaign planning tool. Wouldn't that be great? Welcome to the wonderful world of micro-targeting.*

A world where you can target ads to people who are likely to engage with them. Micro-targeting has been one of the main reasons of Facebook's success as an advertising platform.

In short, it allows advertisers to include or exclude people on detailed criteria. In Meta's Ads manager (read: Facebook's, Messenger and Instagram's campaign planning tool), you can target on specific demographics, interests and behaviours.

Say you sell vinyl records and you specialise in punk. And when you say punk, you mean the original vintage stuff. Think Ramones, The Clash and the Sex Pistols. But you're a punk purist: you wouldn't want to be caught dead with post-punk garbage like Franz Ferdinand or the Arctic Monkeys. You can actually target exactly with those criteria, showing your ad to vinyl lovers who like punk, excluding those who care for post-punk. That is the level of detail, as you can see in the screenshot below.

Screenshot of actual micro-targeting options in Meta Ads Manager

Advertisers love micro-targeting. It allows them to show ads to people who are likely to buy their product and thus optimise their budget. Unsurprisingly, users too prefer targeted ads over shampoo ads to the bald. Win-win you might think.

Europe doesn't think so.

Europe hates micro-targeting. A recent European directive forbids micro targeting based on ‘sensitive personal data’, like ethnicity, colour of your skin, religion, sexual preferences. While that may look like a good idea from afar, it is problematic for two reasons.

Relevance, or the lack thereof.

For decades, people have complained about irrelevant ads. Think shampoo ads for bald men. We now have means to deal with that: we can show ads people might actually find useful.

Often, these ‘sensitive data‘ are explicitly shared by the user. Numerous people disclose their sexual preference in their Facebook/Instagram profile. If they chose to share it, it is questionable to think of that information as personal sensitive information that can't be used in targeting. Same goes for religion. And as a person of colour, you may not want to be targeted with ads that only show white people: the opposite is true.

Effectiveness, or the lack thereof.

While micro-targeting can be really effective when used correctly, today it gets its behind kicked by clever AI algorithms. Nowadays, it's almost always more efficient to define a broad audience and let the algorithms find the patterns and learn who looks at your ad and who clicks on it. They will then show your ad to similar people. Think of it as clever yet anonymous profiling.

Sometimes the AI may really surprises you. Maybe you'll discover those Franz Ferdinand fans will actually buy your Clash albums more than the Clash fans themselves (because the die hard fans already have those albums, duh). And if your assumptions were correct (they rarely are), the algorithm will detect the same predictable patterns from the data. It will not take long for the algorithm to learn it is not effective to show shampoo ads to bald guys.

However sensitive my receding hairline may be.


I've written this article after Belgian's leading financial newspaper published an article on how political parties use borderline illegal micro-targeting techniques for their ads (the irony right?). It's another problematic issue with the Digital Services Act which tries to address real issues with regulation where the cure is worse than the disease.

* Note that the targeting of bald men isn't technically possible in Meta ads manager. But it's a great example right? Or at least it's one I can relate to.

This article was crossposted on Linkedin

Previous
Previous

What brands can learn from Trump’s media strategy

Next
Next

Social Media decided last elections. Will AI decide the next ones?