Meta’s (i.e.Facebook) ad targeting is famously accurate, right?

Remember the whole Cambridge Analytica thing?

It’s been known to be a powerful election-winning tool, and brands have used it to attract new customers.

But hold the phone… New research by North Carolina State University has found that the real accuracy rate of Meta’s targeted ads leaves a lot to be desired.

Nearly a third worse

Meta ads work by taking a user’s interests and matching that to targeted advertising. However, a study by North Carolina State University has said that 33.22% of those interests were inaccurate or irrelevant.

The study also found that Meta has trouble differentiating between positive and negative mentions. For example, researchers made several negative comments about the Harry Potter series and lead actor Daniel Radcliffe but still received Harry Potter adverts.

In this instance, it appears as if Harry Potter has been registered as an interest, despite the negative nature of user comments.

An inaccuracy of this degree matters for advertisers as it could result in poorly directed ad spending and reduced campaign performance. Additionally, marketers could find their product advertised to people who have expressed a negative opinion of it – which is a pretty poor outcome.

Anupam Das, Assistant Professor of Computer Science, NC State

“This inaccuracy has both economic ramifications – since it is relevant to the effectiveness of paid ads – and privacy ramifications, since it raises the possibility of inaccurate data being shared about individuals across multiple platforms.”

Study your detailed targeting

Interest targeting lives under the Detailed Targeting roof on Meta.

Detail targeting allows marketers to dig deeper into the group of people their ads are shown to and should yield better ad outcomes.

According to Meta, detailed targeting outcomes can be based on:

• Ads people click on.
• Pages engaged with.
• Activities on Meta such as device usage and travel preferences.
• Demographics.
• Mobile device used and speed of network.

Detailed targeting can be used to refine your audience to include demographics, interests or behaviours. Marketers can also use it to target people who match several criteria at once. For example, you can target people whose interests include sport and music and TV.

Meta’s definition of a user’s interest opens up massive possibilities for advertising on the platform. Still, if they’re poorly configured, it has the potential to burn through advertisers’ revenue for little return.

At least, that’s what North Carolina State University thinks is going on.

Anupam Das, Assistant Professor of Computer Science, NC State

“Even something as simple as scrolling through a page led to Facebook determining that a user has an interest in that subject.”

As Meta determines interests based on a user’s profile information, the content they engage with, the groups they are members of, and so on, there is a large scope for misinterpretation.

Aafaq Sabir, Ph.D. student and lead author of the research, NC State

“If you posted something about how much you dislike green cheese, the algorithm Meta uses to infer your interests would likely notice that you shared something about green cheese.

But Facebook’s algorithm wouldn’t register the context of your post: that you do not like green cheese. As a result, you may start getting targeted ads for green cheese.”

Small study, big results

It is worth mentioning that the studies’ sample size was small. In its first stage, it used just 14 user accounts. In its second, just 146. For context, Meta has nearly three billion monthly active users.

However, we can still learn something from the result.

Analyzing the Impact and Accuracy of Meta Activity on Meta’s Ad-Interest Inference Process, Aafaq Sabir, Evan Lafontaine and Anupam Das, North Carolina State University

“Researchers created 14 new user accounts on Meta. Researchers controlled the demographic data and behavior of each account, and tracked the list of interests that Meta generated for each account.

This experiment allowed us to see which activities were associated with Meta inferring an interest, and the key finding here is that Meta takes an aggressive approach to interest inference – even something as simple as scrolling through a page led to Meta determining that a user has an interest in that subject.”

During the second stage of testing, researchers recruited 146 participants from various locations worldwide. Participants downloaded a browser extension that allowed researchers to collect data from each participant’s Facebook account and then asked questions about the accuracy of the interests inferred.

In short, stage two used ten times as many people, and their research returned very similar results.

Despite the small sample size, much of the report rings true. Any marketer that has ever viewed their own ad preferences will know they’re often way off, more like 30-50% of the time.

View your personal ad presences and see what percentage rings true for you.

The bigger worry for Meta and marketers alike is that its ad targeting capabilities are only likely to worsen further as the online privacy net tightens.

Apple’s iOS 14.5 update in 2021 has been disastrous for Meta et al., iOS 15, which is rolling out this year, will make things worse, and the upcoming Google Android privacy changes could be the final kiss of death.

Whatever the future brings, the era of mass, hyper-targeted social media marketing campaigns is very much drawing to a close.

Image source: Unsplash.