Data Hiding in Plain Sight: An Ad Industry Lesson from Trump’s Win

Let’s admit it, as an industry we have a nagging tendency to chase shiny objects… the latest whiz-bang tech, apps of the moment, drones, artificial intelligence, you name it. 

The same goes for the use of data in advertising. 

But sometimes the most useful data is hiding in plain sight. You just need to know the data exists somewhere, know where to find it, and have a relationship with the right people to turn the data into action.

I’ve been struck by the number of people, both press folks and the general public, who have expressed shock about what’s emerging regarding the mix of voters who elected Trump. The reality is that it didn’t take a data rocket scientist to see this result might be coming.

Let’s take a simple look at Facebook behavioral data? 15-20 million monthly active users on Facebook had demonstrated an affinity for Trump, so we’re looking at a large behavioral data-set – not a small panel of people completing a survey or being interviewed on the phone.

Sure, the Trump supporters clearly over-index as males, but there is still a massive female portion, 43% to be exact.

And what about the Hispanic audience? 3 million of them had shown through their actual digital behavior that they supported Trump, with younger generations over-indexing – particularly males.

And all those Trump supporters are high-school educated former factory workers, right?  Wrong. Pretty obvious when a quick look at the data uncovers 8-9 million people had completed either a college or grad school degree.         

And here’s a real kicker… a significant number of liberal, or “very liberal” Americans (4.5 million to be exact) showed a strong affinity for Trump through their digital behavior. And the generation that over-indexed was liberal millennials.    

As the DNC was engaged in all their advanced data modeling and targeting, I can’t help but wonder if anyone bothered to look at data in the Facebook Audience Insights interface, readily available to anyone with a Facebook advertising account.

PS. Of course, there’s another lesson to be had around the use of behavioral data based on what people DO, not survey data based on what people SAY they do.  But’s I'll save that for another post…