AI Leadership Insights: AI in AdTech

Amanda Razani: Hello, I’m Amanda Razani with Techstrong.ai. I’m excited to be here today with Tamara Zubatiy. She is the CEO and co-founder of Barometer. How are you doing today?

Tamara Zubatiy: Hi, Amanda. Thank you so much for having me. I am doing very well today. How are you?

Amanda Razani: Doing well. Can you talk a little bit about Barometer, and what are the services you provide?

Tamara Zubatiy: Yeah, thank you. Thanks for the opportunity. Barometer is an AI powered brand suitability and contextual targeting solution for podcast advertising. We provide a variety of services in what we call our AI powered Brand Integrity Cloud. These services include brand suitability analysis for the content of podcasts at the episode and show level, contextual targeting of the content at the episode and show level and host intelligence, which is our own kind of secret sauce for Q scores for hosts, analyzing how hosts superior in the media. Those first two offerings are in line with the industry standard frameworks, GARM for brand suitability and safety, and then the IAB content taxonomy 3.0 for contextual. So those are kind of the three main offerings and then we have three main products. So first product is our platform that’s used by advertisers for vetting and planning for publishers, for curation, and understanding their content from that brand safety and contextual targeting perspective. Then we also offer pre-bid targeting and then in and post-campaign monitoring.

Amanda Razani: Wonderful. Thank you for sharing that with our audience today. So our topic of discussion is the need for AI in AdTech and its impact on the podcast advertising landscape. So my first question is, what is AdTech? Can you explain that?

Tamara Zubatiy: Wow, that’s a great question. I don’t even know if I’m the most qualified person to answer that, but I’ll give you my stab at it. I think AdTech is the technology layer that has needed to be built, the infrastructure layer to power the exchange of advertising across the different channels. So whether that’s DSPs and the many data layers and infrastructures that they maintain or data enrichment providers like us and other measurement companies that provide the technical detail and expertise needed to deliver the data that the marketers actually need to validate that their marketing is working and doing what they want it to do.

Amanda Razani: Got it. So when I think about advertising, combining that with technology, a lot of the big topics of the day are AI. So how does AI fit into the advertising world, and particularly with podcasts?

Tamara Zubatiy: Yeah, absolutely. There’s so many places where AI is used in advertising. It’s kind of amazing. Everything from generating creative based on a brief, to optimizing how much you should bid based on historical data about what bids have been successful. It’s been used even for many, many years before the AI hype bubble, particularly through expert systems and other highly calibrated systems that these companies have been developing for a very long time. The way that it relates specifically to podcasts, I think in many ways it’s similar to how it works in other media types, but I think one of the unique properties of podcasts is the challenge of scale. So there’s so much content, there’s literally millions of podcasts and they’re releasing content daily, weekly, sometimes multiple times a day.
These are millions of hours of user generated ad data, so there’s no top person who knows exactly what everybody’s saying. So even the simple things like discovery or filtering can be a real challenge when you have such a huge scale of inventory. So I think that’s really the opportunity for AI and podcasts is being able to mine the content and metadata about the content to refine and optimize approaches for podcast advertising from advertisers, as well as to refine how selling is done, frankly, on the publisher side as well, enabling new opportunities for new types of targeting, new types of AI data informed buys.

Amanda Razani: Wonderful. So let’s talk a little bit about AI in regard to safety. I know that’s a big concern that enterprises have when they’re trying to implement AI into different processes. There are some concerns with safety. So what about brand safety? What do you have to say in regard to that?

Tamara Zubatiy: Yeah, so brand safety and suitability. So I’ll briefly explain what is brand safety and then we can talk about how AI relates to brand safety. So the concept of brand safety is not a new concept. The idea is there’s certain contexts that brands can’t align with, either can’t or don’t want to. The nuance there is very important. There’s two levels. There’s brand safety and brand suitability. For media, there’s a working group within the World Federation of Advertisers called the Global Alliance for Responsible Media. That was started partially because in 2019 a L’Oreal ad ran on an ISIS video on YouTube. So YouTube, actually RevShare, with ISIS, because that’s their creator, RevShare policy and the people representing L’Oreal in this were very unhappy; the agency that they were working with, and they saw this as a deeper societal problem. How do we not monetize content that’s objectively societally harmful? Not just content that he said, she said might be controversial, but I mean content that’s illegal. So that’s when they came up with this concept called the Brand Safety Floor.

So this is a designation, it’s a binary line in the sand and everything below the floor, there’s a definition for 12 different categories, things like adult content, the below the floor version of that is child pornography. Things like guns, selling illegal guns on your podcast, drugs. These apply to all media types, not just podcasts, but there are specifics instead of visual depictions, it might be discussions in a podcast, nuances like that. So that’s brand safety and it can be quite challenging. You may hear about the challenges of platform safety on large platforms like TikTok or Meta where they have to deal with user-generated content. It’s very similar to podcasting. User-generated content that might be very subversive, illegal, dangerous. So that is kind of the brand safety piece. Within top monetized content, there’s not a lot of stuff like that, that’s below the floor. So there’s a level of nuance that’s needed to really understand brand suitability, which is kind of like the next level up. So if we think about brand safety as how we exclude stuff, brand suitability is how we include stuff with a scalpel rather than an ax.
So to give you some context for how do advertisers do brand suitability before Barometer, there’s a person who sits and listens to episodes and the human makes decisions.

They might listen to a couple of minutes of the episode, they might be really hungry, they’re usually really qualified people, but it’s just a huge problem of scale. You just cannot listen to every single episode of every single show even that you’re running on as an advertiser, let alone the shows that you’re prospecting. So what we sought to do is build a barometer that would basically listen to all of the content in a consistent way, provide data points about these different categories so that advertisers could have and enforce a consistent brand suitability strategy when they buy a podcast. So that’s what we did. So what we do is we listen, our AI listens to all of the podcast episodes and then analyzes them based on these industry standard frameworks reporting on the data in a very transparent way.
Anything that we flag in the content, we highlight and mark up in the transcript so that it’s 100% auditable by any advertiser or publisher so that we could be held maximally accountable. Then we bubble it up at the level of the show. So after we analyze a certain number of episodes over time, we can actually make show level analysis as well. One other moment, content is just part of the brand safety and suitability discussion. It’s equally important, like who the influencer is or who the host is. So sometimes the content itself might be totally suitable at above reproach, but then in the background, the creator is committing fraud or doing some other crime. So that can also be very tricky to detect for advertisers. The current best practice is not always, but sometimes the same person that is listening to all of the episodes, it’s also Googling all of the hosts, Redditing all of the hosts, checking any controversial data that they can find, sometimes literally decades back. So it’s a huge challenge.

For that, our host intelligence solution helps, doesn’t fully replace the efforts of these very talented and true humans who are really working on these processes, but augments their process. So it’s more efficient by mining data from over 80,000 different news providers to understand how these hosts show up in the news from the perspective of sentiment. So finding actually all of the mentions of them in the media, how they’re being portrayed, and then providing transparently all of those contexts, all the articles, all of our sentiment data about them for that same auditable transparency.

Amanda Razani: So I have another question. I know if you watch the media, there have been some conflicting opinions about the use of AI in different industries. I think about efficiency and the ways that we are all trying to harness this AI podcast. There’s several that are put out on a steady stream. There’s of course, I guess images that have to be posted with those, sometimes transcripts. How can AI help with some of that automation, like with the images, transcripts and things like that?

Tamara Zubatiy: Yeah, I think it’s a huge democratizer. I feel like in the past if you were working with a large publisher, you had the luxury of having transcripts and having beautiful media assets and having gorgeous social media tiles. But now there’s so many new companies that are popping up that are making it much easier for individual creatives to really make the most of their podcast and of their audience. There’s companies, I’ll name-drop one that I’ve met at the podcast show in London called Clip Gen, and they literally generate the most click worthy clips of your podcast so that when you make your social media posts, you don’t have to go through editing and finding that perfect clip and all of that. I think the story that I like to tell is how it empowers the humans. So we truly believe that our technology doesn’t really replace humans. And similarly we think that AI’s role should be to amplify the natural intelligence and capabilities and creativity of humans. So I really see that as the opportunity.

Everything from being able to make more content, like if you previously had a weekly show, maybe you can use AI to summarize the research of the things that you were looking into to make it a daily show, to make your content more available. There’s a lot. From the transcription point of view that’s also very exciting. For accessibility, it wasn’t a standard in the past to have transcripts for all things, for all podcasts, like it was for visual media to have subtitles. But now that’s becoming a little bit more of a thing. Not only is the accuracy improving, but the costs are kind of going down. It’s becoming more ubiquitously available. I think Spreaker, I could be wrong about this, is offering free transcription to folks who host on their platform. That’s how ubiquitous it’s becoming, and I’m confident if it’s not already in some countries, it soon will be law for the disability accommodation.

Amanda Razani: That’s awesome. So last question, as rapidly as this technology is advancing, where do you see this industry in a year from now?

Tamara Zubatiy: That’s a great question. I think one of the things that has been challenging for us and as an industry is standing up a more programmatic kind of real-time bidding ecosystem, the same way that it exists in other media channels. I think it’s less than 2% of podcast ads are sold in a realtime exchange with open bidding. So obviously there’s a lot of room for that to change. I think that some of those improvements will be driven by AI. So I think that’s one perspective. I also think that there’s just going to be a lot more creative tools available, so I expect more people to start podcasts. It being easier to make podcasts and maybe be easier to monetize your podcast.
I think there’s been a real gap between let’s say the top 50,000 shows that people actually buy and the millions of shows that exist. I think that that gap has been due to a lack of availability to mine and target that content. Those creators, even if they have their niche audiences that listen to them regularly, they can’t monetize those audiences because nobody knows who they are. So I think that’s potentially something that could happen this year. I don’t know if it’ll happen in one year, but that’s the next frontier, is to unlock all of that content and then democratize the monetization at scale through mechanisms like programmatic.

Amanda Razani: Wonderful. Well, thank you for coming on our show today and sharing your insights.

Tamara Zubatiy: Thank you so much for having me.