TVREV

View Original

How KERV Interactive Trained Their AI To React Like A Human

In our latest TVREV In Cannes video, KERV Interactive CEO Gary Mittman explains the value of KERV’s shoppable and contextual capabilities and how they’ve trained it to react the way a human would.

GARY MITTMAN: KERV Interactive is an AI platform for videos, both ads and content, for shopability, interactivity and contextual relevance.

Contextual cues are critical, because in a world of relevance, a consumer typically, if you have a show that is whatever the content may be, and an ad slot that's out of left field, has no relevance. Nobody's paying attention. And the value of that slot is wasted. It's a wasted impression. The idea is to give the continuity to the viewer to enjoy the process.

So if they're watching a show that has outdoors and adventure and a four wheel drive truck, and the next ad slot should be continuing that relevance. And it should be something that, "Wow, I could really get that that's a value to me!", or "that's interesting to me". Or a vacation, a scene that could be in a beach where it's a party, and the next ad slot should be, go to Hawaii, go to Ibiza. So the continuity of it is really the value.

And we've seen from extensive testing that the legitimacy of that is real, that the engagement, the value, the brand lift, all the components are solid.

So when we analyze content, there are many elements to incorporate— all the cues like the human natural eye identifies. What we've trained our AI to do is to be able to react similar to a human's response. So we know happy and sad is very basic and far deeper than that is, is the real reaction of a human and how they would react to something, and being able to define that and create moments around that that are relevant to either shopability, ad placement or other forms of contextual value.

Creating the show shoppable is a couple different ways. We have had great success in doing it in partnership with people like Walmart, where we bring in their entire catalog in our platform, and we analyze a show and then match and correlate products to the objects within the shows.

We could do that with multiple catalogs. We can do that with sponsorship or other forms of monetization, both for the advertiser and the product content owner. We work very closely with the content owners because they have rules, restrictions and guidelines around what you have to do to make a show shoppable.

There are many different variables that we have to comply with in their contractual obligations. So the partnership with the content owner is critical, and then the importing in of the advertiser is the other layer.

So the monetization comes from the partnership with the advertiser, and that could be transactional or sponsorship, and the content owner is the one who has the rules and guidelines around what we can match, where what and how.