PxPixel
The Predictive Engine That Could - Broadcasting & Cable

The Predictive Engine That Could

Algorithms determine what targeted viewers will be watching
Author:
Publish date:

Related: TV’s New Data Recruits Don’t Do Suits

Looking in the rear-view mirror isn’t the best way to predict what will happen next. Objects may be closer than they appear.

The same thing is true when you’re selling a precise audience to clients. Networks need a better way to know where those targeted viewers are going to be.

It’s tougher than simply looking at how many 18-49-year-olds watched a show last week; in fact, it takes some pretty sophisticated data engineering. And while algorithms aren’t crystal balls, they use more information to make a better estimate, companies say.

“We’ve actually found through our research that if you just look at historical data and you’re trying to base a future media plan just off of historical optimization, you’re only right 50% of the time,” says Bryson Gordon, senior VP of data strategy at Viacom.

Viacom’s data engineers spent more than two years developing the Vantage Predictive Engine to understand where to place ads at what time for each advertiser. “This is where our data scientists have spent a ton of time writing advanced predictive algorithms that do a very good job of predicting where a specific audience is going to show up five days from now, 10 days from now,” says Gordon.

Viacom works with clients after they buy a Vantage schedule. “We do this at the beginning of the campaign, and we optimize it again every other week,” he says.

As Turner Broadcasting developed its data products starting in 2012, it made a similar discovery about linear TV. “The fallacy of where some other media companies are leaning [is] they’re leaning toward historical indices,” says Michael Strober, executive VP of client strategy and ad innovation. “You can’t rely exclusively on historical data. That’s just one attribute you factor into your algorithm.”

Turner has applied for a patent for what it calls Competitive Audience Estimation. Even for the same show, different audiences will watch at different times, and CAE takes that into account when making predictions.

“What’s so powerful in this is depending on whatever client they deem they want to go after, the data set that they want to use—if it’s an auto company and they want to use Polk data on top of Rentrak—I can do that, and I can estimate now at 11 a.m. on How I Met Your Mother how many auto intenders will fill that particular target definition,” Strober says. “It is literally the engine that is powering our entire audience targeting platform.”

Related: TV’s New Data Recruits Don’t Do Suits

Looking in the rear-view mirror isn’t the best way to predict what will happen next. Objects may be closer than they appear.

The same thing is true when you’re selling a precise audience to clients. Networks need a better way to know where those targeted viewers are going to be.

Member Exclusive

Get Access to Our Exclusive Content

Related