Balancing Tone and Sentiment in Podcast Advertising with Dr. Zubatiy

Join host Annalise Curvelo as she sits down with Dr. Tamara Zubatiy, Co-founder and CEO of Barometer, to discuss her journey from cognitive science to leading Barometer—a company at the forefront of brand suitability and contextual targeting in podcast advertising.

Dr. Zubatiy breaks down how Barometer’s sophisticated tools examine podcast content with a nuanced focus for tone and context. Discover how these tools can help brands make thoughtful advertising choices, ensuring their messages align with their values while steering clear of unwanted content.
Learn about the intricate world of podcast advertising, including the challenges of maintaining brand safety while engaging effectively with listeners. Dr. Zubatiy also offers an inside look at Barometer’s methods for assessing complex content, including true crime podcasts that might use a lighter or more entertaining tone for serious topics like crime and murder, adding an extra layer of nuance.

TRANSCRIPT

Annalise Curvelo
Welcome to Sightly’s Breaking Through the Mayhem podcast. I'm your host Annalise Curvelo, and we are thrilled today to have Doctor Tamara Zubatiy, co-founder and CEO of Barometer join us today. Barometer has built an AI powered omnichannel brand suitability and contextual targeting solution. They launched in the digital audio space in 2021 and became quickly the best in class solution for podcast brand suitability. Since then, Barometer has grown an omni channel footprint with a particular focus on news content and diverse voices. Tamara, you've had over ten years of experience working in applied machine learning across both academic and industry roles.

Welcome to the podcast. We're so excited. Thank you so much for joining us today.

Tamara Zubatiy
Thank you so much for having me. It's a pleasure to be here and great to see you Annalise.

Annalise Curvelo
Nice. Great to see you too. So could you start out by just telling us a little bit about your background, what were you up to before Barometer and how this all kind of happened?

Tamara Zubatiy
Yeah, totally. So my background started in my education. So, my undergrad was focused in cognitive science with specialization in machine learning and neural computation at UC San Diego. I feel like that really formed kind of my foundational lens for how I see the world. and then after that, I did a lot of human centered computing research primarily focused on the space of older adults and conversational AI technologies.

So thinking about like Google Home and Siri and what would the next kind of iteration of those technologies look like? And so my PhD was focused in that at Georgia Tech. I recently completed that. So that's kind of been my background is human centered AI and kind of, to put it in like colloquial way. It's like I always was frustrated that I had to learn a new language to communicate with the machine.

Why shouldn't the machine, like, learn our language to communicate with us? So that was kind of my why.

Annalise Curvelo
That's really interesting. Yeah, that's exactly right. That's kind of funny. You're back. I mean, your background in cognitive science has to give you such a cool and unique advantage in the AI space, and, you know, your perspective on machine learning.

How have you leveraged the understanding of the human brain to kind of do things to replicate exactly what you're saying? Like that language? speaking of, you know, learning and human decision making.

Tamara Zubatiy
Yeah. I feel like there's so much there. I always say, like, my first love was the brain. it's just so interesting. one of the biggest kind of learnings from studying the space for so long is just how little we actually understand about how the brain works, and there's so many competing, still theories.

And it's kind of amazing, honestly, how quickly the space is still evolving. So I would say that the current kind of implementations that we've built are based on the current understandings of how the brain works, but it might not necessarily be held up in the future. And so for a lot of reasons, I'm a huge fan of deterministic machine learning approaches.

I feel like there's a lot of really interesting opportunities there. And one of my other big learnings from cog side besides that, we don't know yet everything about the brain is, to be human is to be biased. So, like, that's the one thing that we can count on for being people. And one of my favorite biases that we have, or the impact of the biases that we have is our kind of challenge with consistency as people.

It's really hard to be really consistent at, like, analyzing something or, even like coding something to qualitative coding, an interview, something like that, to be like, oh, this is a mention of this, or this is a discussion about that. And so that was kind of one of the things that at a high level, I identified as, why should humans do the things that we're really bad at?

Why can't we find machine help for helping us compensate for the biases that we know we have? And so that was a really kind of philosophical motivator, if you will, for Barometer.

Annalise Curvelo
That's really cool because it's funny. And I know we were talking about this a little bit before we hopped in, but I think there are a lot of synergies in our two companies.

And, you know, one of the things one of our big missions and driving factors as well is how we can help brands stay authentic and consistent in the market. So by everything that they're doing, because we want to help them say, okay, if you're making moves, you know, a lot, a big topic in our industry right now is sustainability.

It's something we're really excited about. If you're a as a brand, are making moves around, you know, becoming more sustainable and adding those practices into your organization, but you're only telling your audience about it on Earth Day, right? You're only doing something about it like, no, we gotta help you. We gotta help you be more consistent. And that's actually one way we're also leveraging, you know, AI and machine learning.

So that's really that's really I have never, you know, thought about connecting it back to what humans are not good at. But that's kind of what we're doing is helping brands become more human. And therefore they are going to naturally have those challenges as well. Exactly.

Tamara Zubatiy
And it might seem so obvious after you bring it up to them that, hey, like this is you say that this is like something you really value, but you look at it like you only care about it for one day of the year.

I think that's such a great example. Yeah.

Annalise Curvelo
Yeah. So another area where I think we're really aligned Sightly and Barometer is we're helping advertisers gain the confidence to play in spaces that have been typically and historically and still to this day, kind of blanket blocked. Right. So like news and politics. So at Sightly our brand mentality platform is breaking down content in a very nuanced and granular way.

And then we're providing opportunities for brands to lean in where it makes sense and kind of stay away from areas where it doesn't make sense based on their unique perspective on the world, which is what we call their brand mentality. Right. so, for example, if a brand wants to stay away from news about the catastrophe side of a natural disaster, but they want to lean into the warm, fuzzy stories about pets getting saved, right?

That’s where our platform is able to break that down and kind of provide that option for them, which is really cool. So I love to hear a little bit more about how you guys are also creating the confidence for brands to play in these types of spaces, and specifically in the podcast area.

Tamara Zubatiy
Yeah, absolutely. I think, one of the ways that we have well, firstly, I really I really like that example, and I think it's so cool that you guys are able to make that distinction because it is just so important to support the news.

And yeah, we definitely do a lot of things to try to granular arise or like make additional like, columns, if you will, for people to be able to evaluate a piece of content with, not just like one category, like debated sensitive social issues comes to mind, it can be so easy to just be like, no risk for debated sensitive social issues.

But one of the things that we've taken great care to do is to unpack some of the GARM definitions and to really elaborate on them in places where they're maybe like, could use some more elaboration and also to just interpret them the way that they're actually written. For example, with news being specifically called out in informative right, as investigative pieces about like different things that are happening, breaking news, all of those different specific news related terms are actually mentioned directly in the GARM framework.

And we feel like one of the things that we started doing is actually trying to figure out, like, label, is this a dramatic discussion of a current event, or is this like an informative investigative discussion? And so we do this in a couple of ways. the first way is flagging all of the different topics being discussed instead of just bucketing them into potentially contro, and this allows advertisers to be like you were saying a little bit, maybe aligned more with their brand values.

So for example, there's like thousands of potentially debated sensitive social issues, but maybe the brand has a case about sustainability and they want to align with content about global warming and climate change because they feel like that's actually, one of their purposes. And so their purpose driven, like brand message should be aligned with that kind of content, but maybe they're not comfortable with political discussions.

And so you can tell immediately how it could get dicey. Maybe there's a political discussion talking about global warming and so then we start kind of further unpacking and adding in additional layers. We have an interesting layer about elections and political people. You can add specific granularity about, like maybe you only are comfortable when one side of the aisle is talking about discussions about climate change.

Adding more and more of those granular kind of layers helps to open up those inventory sources that were previously completely blocked. One of the interesting things that we've seen is being able to show an advertiser like a huge swath of different types of news and show them. Initially you said maybe no news or specifically no news politics. We can get into how like sports is tagged as news, that everything else is tagged as news and all of that.

But just in this phase of politics, it can be really helpful to show a brand like comparison, a podcast that's like, very. I don't want to name drop any podcasts right now in this recording, just for a variety of reasons. But there is a lot of variation when it comes to the way that content is talked about.

And at Barometer, we think about it in like three levels presence of, context, and tone. And so I know we'll talk about tone a little bit more later. But at a high level, when somebody looks at the way that news podcasts are analyzed in Barometer, they can see it broken down by all the categories. And then they also get this qualitative layer about, well, how was the discussion?

Was the discussion negative and partisan and opinion or was the discussion neutral and informative, or was the discussion appropriately emotional for the subject matter? Those are kind of the three basic buckets. In addition to being able to get really granular about the subject matter. So that's kind of how we think about it. Like I like to say we're horizontally in terms of all of the different types of columns or features that you could assess in the content, and then also vertically in terms of the tone and the other types of like more qualitative, features that we can detect to contextualize how that topic is being discussed.

Annalise Curvelo
Wow. Okay. So I mean that is a very robust way to look at content, right? First of all, I feel like you should just take a moment and be proud because it's really cool. And not only is it a robust way to look at content, but you're doing this in long form content like that's very difficult.

So I'm assuming I'm not, I don't have the machine learning background. we have a really amazing team at Sightly that, you know, does all of that. And, and they're so talented and I'm learning from them every day. But I know that just I'm curious how you got into podcasts specifically being that is such a challenging territory with its long form nature.

So yeah. What was the decision to venture into that format?

Tamara Zubatiy
Yeah, that's a good question. It was partially a passion and partially like a need. On the passion side, we're really passionate about freedom of speech at Barometer. I was born in the Ukraine. Most of our team was born abroad and lived in countries where freedom of speech wasn't guaranteed and freedom of press maybe didn't exist.

So one of the things that we observed is the waning trust in certain media institutions and the sort of flight to independent sources and podcasting being one of them, amongst consumers. So there's a lot of data that shows that there's been a huge increase in consumer listening, of podcasts and audio content. It's something like 30% of adult American media time is spent on digital audio today, which is shocking.

The other data points are about, like news consumers and, news consumers tend to engage with like, print and digital news, but also tend to engage in podcasts and there's, podcasts was for a time the fastest growing digital, advertising channel. So from kind of like the passion and interest point of view, that's kind of why it seemed like it was popping off and it seemed philosophically interesting, because all of these conversations were being had there that weren't being had anywhere else.

It also seemed like kind of the bastion for the purple people who didn't fit nicely into any of the sides, especially several years ago, like in 2020. And then, of course, the pandemic shut down a lot of the other types of traditional recording. And so that also created this huge wave of influencers, individuals and famous people, frankly, moving their full production into podcasts, which changed the space for being like ham radio

People who wanted to have like a podcast show to these are multimillion dollar productions. This is done in the studio. Like, these are scripted. it's it's a higher budget than many people would like first assume. And so that's kind of the whole other side of podcasting is the business of podcasting, aside from the many, many millions of individuals podcasts that have been put out, the other part of why we picked podcasting was out of necessity.

So getting access to data is really hard. so getting access to like, print data, we licensed news and print data from sources. and that was really hard and expensive. And it takes time, for CTV data and walled guard data that that stuff is locked down and regulated for how it can be accessed. One of the amazing things about podcasts is it's syndicated publicly on the open internet using RSS.

And so all of the audio files live in the RSS feeds. And so we basically made a very large financial investment in transcribing the content that was available using the best transcription technology that money could buy. And I think we're up to about 200 years of audio content transcribed. So that was a huge factor, is we got access to those conversations, and it was these premium long form conversations, in addition to the premium digital content that we were getting from the news.

So those were some of the really important factors for us, and why we chose podcasts, like as a joke, we would say internally it just the hard one, like it's really hard. So why not do that one? Because it's really hard. And then we could learn and branch out from that.

Annalise Curvelo
You like a challenge. I love that. Wow, that's really cool. Thank you for sharing all that I was that was more that was personally one of the questions that I was most curious about too. Besides like more about your personal background. So thanks for sharing all of that.

Tamara Zubatiy
For sure.

Annalise Curvelo
Okay, so let's get into the tone conversation because you brought that up a little bit.

So I know Barometer doesn't just analyze sentiment but also tone. You were kind of talking about the different horizontal and vertical layers that you guys get into. So how do you differentiate the two? And then can you tell us a little bit about why it's so important to differentiate the two? I know you kind of got into that already a little bit, but I just love to hear a little bit more about that topic.

Tamara Zubatiy
Totally. Yeah. We think that like sentiment and tone are quite related. So the way that we look at tone is we think about it in terms of emotions. So what emotions are present in how something is talked about. And to characterize tone, you could look at all of the different emotions that are possible and kind of the different levels at which they are there.

It's kind of like Shazam and how it looks at like the waveform of the sound. so you could look at like, what's the fingerprint of sarcasm? What's the fingerprint of like genuine happiness or something like that? Right. In terms of what emotions are present. But then when you layer on what's being talked about, it adds a really interesting quality.

So for example, crime. So you're talking about crime, but the emotions that you're using to talk about crime are happiness. And joy and like enthusiasm.

Annalise Curvelo
It’s like all my friends that that love the true crime. They get so much joy out of the true crime drug. I’ts not for me, I get nightmares. But that's I feel like that's really interesting.

Tamara Zubatiy
Exactly. And there's literally shows that are like my favorite murder, right? Like, so that's one example where the subject area is crime. So we know that that's kind of like in terms of sentiment, that's a negative sentiment around that topic. In terms of tone. How are they talking about it. They're using these enthuses as happiness and joy.

And what's the sentiment of those emotions? So that's the connection that we make is each emotion and tone has a sentiment. And maybe that sentiment in this case is positive. And then what we try to understand is what's the alignment between the topic and the sentiment with which it's being discussed, which is characterized by these different tones. And so that's when we get into what does glamorizing mean?

And so the glamorizing of true crime would be like making a light of crime. And so that's exactly that is you are talking about this really dark thing, but you're using these like really happy words and it's it's a misalignment.

Annalise Curvelo
Wow. Yeah. So do you find that advertisers want to stay away from true true crime or do they lean into it? I'm just curious.

Tamara Zubatiy
That's such a great point. So we're trying. One of the things that we are coining is the concept of contextual inclusion. so what's appropriate in one context might not be appropriate in another context. So an advertiser who might be afraid of discussions of violent crimes going on in their neighborhood might be totally fine with a fiction podcast or an investigative true crime podcast where they're talking about it in the context of a true crime.
And so in, in Barometer, we actually have ways for people to make contextual exceptions, for inclusion. So you can say generally I want medium risk for true crime or medium risk for crime, medium risk for death, medium risk for arms and ammunition. Except for in true crime, I'm okay with whatever discussions are in true crime because that's contextually suitable.

Similar for like history. This comes up all the time where people are like, I want to avoid, discussions of current wars. And so they block all the history. And then when we really press them on that, we're like, well, that's not a current war. Are you sure you wouldn't be okay with running on that? And so that's one of the opportunities that arises.

Annalise Curvelo
That's really interesting.

Tamara Zubatiy
Yeah. And then another thing I would add is, there's like the basics, like the GARMs and the IABs and then there's things that are really specific to advertisers, like, say, an advertiser is dealing with like elder eldercare or like is a nursing home, say, the advertisers, a nursing home. They would maybe want to stay away from certain types of true crime and not others, or say it's a preschool.

They might want to stay away from child abuse, but they might be okay on other types of true crime because they know moms would listen to true crime, but maybe they wouldn't want to run on the children, like it can get really, really nuanced, but that's actually an opportunity where maybe a category that would have been previously entirely blocked, that could actually perfectly hit your audience, could be selectively unblocked, whether it's to include the non glamorizing crime or the crime that doesn't specifically talk about children or whatever it is that's needed to allow that content to be included.

Annalise Curvelo
Wow. Yeah. And it's really really cool to hear you say that. Sometimes you do have to press them a little bit and give them these kinds of examples. How do you normally handle that?

Tamara Zubatiy
So we make it a data driven process and we formulate it in terms of proposals. So we start with whatever their existing brand risk profile is as they've articulated it.

And then we say basically we try to move one thing. So every time we propose we have a filter menu in our UI where you can enter their filters. And then it'll curate the inventory based on their filters. And then you can basically move one and say, okay, this is all of your settings. And then we're just going to move crime from low to medium and death from low to medium, and see which true crime shows are included.

And so there's like a couple filters we press and then we just screenshot that, pop it into a PowerPoint slide and say, these are the 30 titles that you would include if you could change this one component of your risk profile for this one context only. And then sometimes the publisher can help us and they'll give us like reach data or other unique data, like 100 million more unique listeners would be activated if you included this one podcast, because it's our biggest show and you're currently excluding it,

Annalise Curvelo
like their demo, if it's a mom, if it's in this age range, etc..

Tamara Zubatiy
Yeah.

Annalise Curvelo
Wow. That's actually really interesting. I'm very excited to also show you our platform at some point, because I feel like we're just doing very similar things in kind of in different spaces. So it sounds like I don't even think I told you this, but I lead our partnerships team, so I'm getting all sorts of partnership ideas spinning while we're chatting, which is very fun.

Tamara Zubatiy
That's awesome.

Annalise Curvelo
Yeah. Okay. So that's that's super interesting. Let's, let's talk a little bit more about you. You brought up, value based, you know, the importance of having your clients figure out what their values are. That's something that, again, we also talk about on a daily basis. So we tend to find that brands that are ready to think about advertising this way and to kind of handle these types of conversations and take all this on are more purpose driven brands.

So that's kind of who is really all into this concept, right, of figuring out what your values are and then making sure that they're contextually aligned with the media that they're, they're running next to and all of that. So, is that kind of the types of companies that you guys are also going after? And do you have like a couple of success stories, you don’t have to name brands.

But just so I'm just curious.

Tamara Zubatiy
Yeah, I think, so we like to work with both advertisers and publishers, and I think that's kind of a weird thing about us or not weird, but like unique in the space of, like, measurement vendors in terms of, like, we offer the exact same product to both advertisers and publishers, and we make no distinction in terms of the types of advertisers we like to work with.

There's two broad categories within podcast advertising. There's like broadly, the performance marketers, mostly interested in like bottom of funnel metrics, historically working with podcast agency. So there's a handful of like podcast specific marketing agencies out there. And these are really public brands that if you listen to podcasts, you've heard they're ad, like brands that come to mind are like Rocket Money or like BetterHelp or like SimpliSafe or Indeed.com.

Those types of brands like household names built on podcasts. and then there's the space of like, blue chip advertisers. One of the things that we observed when we got into the spaces, the overlap of the top ten advertisers in the world and the overlap of the top ten advertisers in podcasting is zero. And brand suitability is a huge part of that.

Like why they can't get into that, all of that. So that became kind of right is let's find the brands that have historically stayed out of podcasting and bring those brands in. And let's empower the brands that have always invested in podcasting to scale their programs more and responsibly. And so when we talk to a large brand that's been investing in podcasting for a long time, they might be running on thousands of podcasts in any given week.

Something that started as a couple of host red ads turns quickly into a program that can be really unwieldy and hard to manage. And so that's where we can really help, like performance marketers who really know what they're doing already. For larger advertisers, we'll typically work with like agencies. And we have a couple of vertical focuses. Finserv is a really big vertical in terms of they want to be in podcasts, but it's scary.

And we kind of help them provide some of those assurances. I'm sure you guys have similar experiences with regulated brands like insurance or like health care. Yeah. Auto often, we've noticed that different verticals have similar things that they care about, like auto brands might not want to be around like discussions of drunk driving or things like that.

A car crashes.

Annalise Curvelo
Yeah.

Tamara Zubatiy
And similar, like finserv clients might not want to be around like negative discussions about credit cards or fraud or something like that. so that's how we think about it on the advertiser side. In terms of like, how do we pick an advertiser to work with? It's kind of are they already investing in podcasts and are we growing their investment, or is it somebody new that we're really trying to build trust in the medium by introducing them, through this kind of controlled way of buying that allows them to use their digital savviness in, podcasts.

And then for the publishers, one of the things that we've observed is basically publishers are not included in the conversation, of brand suitability. The platforms are, but publishers aren't. And so we feel like the open internet publishers have kind of been done a disservice because of this. And so one of our big things is just working with the publishers and sharing the same exact data that advertisers use to evaluate their shows, so that they could write that email back and say, here's our 30 shows that almost meet your standard perfectly, but they can use the same language that the advertiser is using.

And we found that to be, like, totally transformative in the way that business is done and trust is built.

Annalise Curvelo
It sounds like you are connecting a lot of really important elements to our industry. It's really cool. okay. So we've covered a lot. I feel like I could keep talking to you for hours. This has been really fun.

But to end on a bit of a fun, what are some of your interests outside of all of this that we've been talking about?

Tamara Zubatiy
Yeah. So we just moved to New York like, a year ago with my husband. And so we've been really obsessed with New York. I think that our interest right now is going to all of the museums.

And we're big foodies. we love dining. we love to cook as well. And, fun fact, both, my husband and I, my husband's our CTO, Grant, we're both, yoga teachers, so, mndfulness.

Annalise Curvelo
Both of you?

Tamara Zubatiy
Yeah

Annalise Curvelo
Wow. That's awesome.

Tamara Zubatiy
Yeah. It's fun. What about you? I'm curious as well. And I also wanted to ask you what some of your favorite podcasts are, if any.

Annalise Curvelo
Okay, well, we can get into that. Yeah. I'm actually, I'm not the biggest podcast person. I think it's because I don't have a commute. When I did have a commute, I loved to listen to a podcast. Nowadays, I walk everywhere, and when I'm walking, I just kind of take it all in. I also, I was meant to kind of bring this up earlier.

San Diego is where I was living, and we just moved to Miami. So I'm kind of with you on this, like, new city, exciting thing. So we're also kind of doing the foodie stuff and going to the beach a lot and all of that. So it's been really fun. But, my favorite podcasts I don't even know. I have to get back to you on that.

I should get back, you know, but I have to look at my Spotify and see what ones I have because I haven't listened to any recently. Honestly, I guess I'm not in that 30% of my time, but almost everyone I know does. I feel like I listen to a podcast episode more often than I have, like a favorite, because people will just send them to me and then I'll kind of check it out.

And of course, the Breaking Through the Mayhem podcast, you know, I guess

Tamara Zubatiy
Add that to your auto downloads you guys.

Annalise Curvelo
Yes, I love it. So okay. Well, yeah, this has been really great. I would definitely want to hear about so much more. But what we will have to do in part 2 at some point, to wrap things up, where can our listeners go to find out more about you and Barometer?

Tamara Zubatiy
So you can visit our website at app.thebarometer.co or find me on LinkedIn, where we are very vocal and active. LinkedIn/tzubatiy

Annalise Curvelo
Amazing. Thank you so much again for joining us. It's been such a good time.

Tamara Zubatiy
Thank you so much for having me.