Tech firms use audiovisual signals to target TV ads

For the purposes of targeting TV ads, cross-identification often involves linking set-top box data to users’ devices through the use of software that recognizes specific TV content from just a few seconds of audiovisual data. Although the tech is in its early stages, some companies are using Shazam-like features and webcams to make ads more addressable.

“We’re not an ad network or looking to be the de facto measurement,” said Liz Hughes, chief commercial officer of data collection company Axwave. “We’re just providing data and insights into actual behavior.”

Identifying users
One common way to link users to their devices is for targeting companies to form partnerships with app developers and publishers who then share their data. After the targeting company has a known user linked to one device, they then can pair the user to their other devices using inferred information such as IP addresses and browser settings. Marketers can then leverage this data to make their TV buys more precise, said Kate O’Loughlin, svp of media at Tapad.

Another option is to use a panel. Similar to how Nielsen gets its TV ratings, Axwave gets its targeting data by tracking a panel of people. Axwave’s panel is currently 2,000 people, but it should expand to 20,000 people in 2017, said Hughes.

Using audio to target
Axwave tracks its panelists by turning on the microphone on their phone and sucking in ambient audio signals. By constantly running recognition technology, Axwave uses sound alone to determine the show and station that a person is watching at any given time.

Alphonso is another company that targets users through audio signals. An obvious use for this data is to leverage it for serving addressable TV ads, which allow marketers to target commercials to individual households through cable and satellite TV distributors. But that application is limited by TV’s older infrastructure because only 42 million households of the 118.4 million total TV households in the U.S. are addressable, and cable and satellite providers have access to only two minutes of addressable commercial space per hour.

Ashish Chordia, Alphonso CEO, said that just as mobile data can inform TV targeting, the TV data can also be used to improve targeting on the Web. For example, if AB Inbev wanted to advertise to “The Voice” viewers but the TV inventory was already entirely sold, AB Inbev could tap into Alphonso’s cross-device identification data to determine exactly who is watching “The Voice.” So when those viewers then open up an app on their phone or tablet, they’ll be targeted with a Budweiser ad.

Other vendors are using video, rather than audio, to link people to their TVs. By setting a camera on top of people’s TVs, TVision is able to scan facial features to determine which person in the household is watching TV. Microsoft and Google have acquired firms with similar capabilities.

Early stages
Chordia said that it’s still difficult to target down to the individual level using audio alone, even though Alphonso is capable of targeting households.

One potential way to get more granular data on which person in the house is watching TV would be to monitor which person’s device was closer to the TV during the broadcast or which device was used during the broadcast to visit websites that are indicative of the show’s demographic. For example, a football viewer is likely to visit ESPN.com during a broadcast. But these solutions are only theoretical as of now, Chordia said while stressing the infancy of TV targeting.

As Hughes put it, “We’re still kind of in the evangelical stage of talking to advertisers and broadcasters about the value of this.”