How crappy is digital targeting, really?

Like, one of the big promises of digital media is targeting. HYPERtargeting even.

But how accurate is the data these big ad platforms have on us?

Adlook decided to find out.

They bought a bunch of banner ads from several popular data providers. They targeted specific segments with broad socio-demographic variables like gender or education. And then instead of running ads, they showed in-ad-unit survey questions asking about those exact socio-demographic questions. (“Do you have kids?” “Do you rent or own your home?” etc.)

They got about 2,500 responses. And the results were cray cray.

F’rinstance, 40% of the folks who were tagged as homeowners were actually renters. Roughly half of everyone was tagged the wrong gender. And a whopping 76% of folks tagged as being married were not. (They didn’t ask if “it’s complicated”….)

That’s pretty craptastic. Several of these basic questions are WORSE than a coin toss.Granted, demographics might not matter if you’re selling coffee. But it might matter a ton if you’re selling mortgages or diapers. I mean, how’s P&G gonna get their Venus & Gillette ads to the right audiences if half of the gender tags are wrong?  

Caveats galore, of course. But this isn’t the first study to show how bad digital targeting seems to be. And hopefully won’t be the last. But the big data providers have very little incentive to fix things right now.

Some lessons:

🍊 When it comes to digital ad targeting, don’t trust; verify! (E.g., run experiments to check your provider’s work.)

🍊 Consider paying more for high-quality & first-party inventory.