EDO’s Kevin Krim wants to make TV ads as easy as search and social

Ever heard of the Adult Swim effect?

That’s what Kevin Krim, president and CEO of TV measurement firm EDO, calls the late-night spike in response to pizza ads — right around 10 PM.

EDO makes insights like that possible by linking TV ad exposure to real-world behavior, down to the minute. It’s how they know exactly when people are craving a slice.

Krim recently sat down with The Current Editor-in-Chief Stephanie Paterik to explain how EDO is bringing the precision of search and social to the world of TV advertising.

Can you take me back to the beginning of EDO. What was your motivation for starting it, and how did it come together?

The inspiration for EDO was basically my background. I spent the first part of my career as an advertising product manager doing early search ads and early social ads at some of the pioneers in the field. And then middle third of my career, I moved out to TV and I built out that first generation of digital products for streaming apps on mobile and connected TVs.

And I was struck by how different the TV was from that original world. And what I realized was that as streaming became the preference for consumers and for advertisers and even for publishers, inevitably, you would want a pricing mechanism and a measurement mechanism that is more similar to what the search advertising and social advertising do so elegantly, which is optimize automatically for outcomes.

And so, we built EDO to be a fully syndicated outcome measurement platform for this convergent TV future.

You recently announced an integration with The Trade Desk. Can you tell me what sort of insight marketers can expect from your data?

When you look at the kinds of outcomes we measure, you can know what works. The reason you know that is these are predictive behaviors that we’re measuring. We’re measuring branded searches, branded website visits, app usage after somebody is exposed to an ad across convergent TV, both linear and streaming TV.

What we’re seeing is the lift above normal baseline behaviors of people engaging with these brands. Entering the funnel, showing signals of consideration and intent that are leading them eventually to purchase to loyalty with the brand. But you don’t have to wait for it.

There are large signals of data that are very powerful to use for models around [things like]: What’s the best audience segment? What’s the best media context? What are the best creative formats or even creative messaging?

Does your data influence the creative messaging?

We see a feedback loop, yes. The really modern marketers want to look at the patterns of what their creatives in the past have done, what their competitors have done well or not done well. And that’s all influencing their creative judgment. We don’t try to replace any of that. That is still the best use for human creativity and judgment.

But what we do see is that, yes, over time, they start to see patterns in what’s working for them and what’s really working across their whole category. And that influences absolutely the evolution of creative there. We’ll see creatives recut ads that are working so they can extend the life of a really good creative idea. They’ll cut it into constituent components, take a 30 down to a 15, maybe even a 10, all based on the data that they’re seeing of what’s performing.

It’s like bringing efficiency to human creativity.

Exactly. And then we see what’s truly incremental in near real time. But we also look at a full window of a week to see if there’s a tail to that behavior that might have been influenced but didn’t trigger the immediate response.

We’re seeing all of that, and we’re seeing how much engagement lift is there cut by audience for media and go down to minutes of the day. We can see that there’s an extra spike for people responding to pizza at around 10 p.m.

We used to call that the Adult Swim effect. So, it’s those kind of things you can get very granular on.

And it’s worth pointing out that one of your co-founders is Edward Norton.

That’s right.

I think that’s gotten a lot of attention. Why is Edward Norton part of an ad tech company?

Edward is one of the co-founders, and he deserves all the credit for the initial push to say that there’s an opportunity to take these emerging AI models. This was 10 years ago, but we were already seeing the power of applying these machine-learning models to very large data sets that were otherwise being ignored as exhaust flowing out of other businesses.

And what he said was, as a filmmaker, as somebody who cares a ton about the creative process, my friends and I are making fewer of the kind of films we want to make because of the hugely expensive efforts around marketing those films.

If we can make that more effective, make it more efficient on an impact-per-dollar basis, then we’re going to be greenlit. We’re going to get more projects greenlit that we want to make.

He’s been a very engaged founder and chairman, all the way through the 10 years.

That’s so cool. And I love that idea of data being something that can really fuel and empower creativity, humanity and art.

The data doesn’t replace human creativity. It can enhance it. And he’s a big believer in that, talks about that all the time.

What is coming down the pipe that gets you most excited right now?

What we care a lot about is really intriguing is verticalized AI or just simply vertical AI. Our industry is pretty specific.

Notoriously!

What we’re finding is that if you specially build AI models and more importantly, really train them on industry-specific datasets to solve industry-specific problems, you’re going to have a better result than if you take the generic approach.

So, what we’ve done is taken 10 years of linear TV history, four and a half years of streaming TV history across the whole landscape. And we’ve trained these models on literally trillions of impressions. We’ve seen 104 trillion impressions, and it’s always growing.

And so, these models, as we’ve learned from ChatGPT and others, good quality, accurate data in at large quantities means much better results out. And so that is something that only AI models can do.

We couldn’t process that volume of data and have it be accurate and fast if we threw bodies at that problem. First of all, no one would want to do that job. Some of these things are incredibly tedious. And we would be inevitably making mistakes and we’d be slower than these machines can be.

So, we think verticalized AI is a really great and powerful application of these AI models.

The Current is owned and operated by The Trade Desk Inc.