Bart Roselli: The Evolution of Audio Advertising From ‘Serial’ to Synthetic Voice

Veritone One’s SVP Growth Talks Strategy and Tactics For MultiChannel Audio Advertising

“Change is inevitable. At Veritone, we always say, if you're not driving forward, you're reacting. And if you're reacting, you're losing, whether it's share of voice or creative messaging.”

By E.B. Moss

Bart Roselli has seen the audio space expand exponentially in his 10 years at Veritone, Inc., just as he’s seen Veritone itself grow into a publicly traded entity supporting podcasting, radio, YouTube, digital streaming, search and attribution software, and even synthetic voice offerings from their Marvel AI group.

Appropriately titled, Roselli is SVP Growth for Veritone One, the division focused on enhancing agency-media vendor relationships and fulfilling client goals across all audio opportunities. In this conversation, he offers both a primer on podcast ad strategies and the evolution of the space, as well as how audio supports both brand and demand marketing goals, from creative to targeting to greater sophistication in tracking response.

---

Let’s start with the growth you’ve seen in your years in audio.

First, brands need to think of audio outside of ‘just radio’: It’s also podcasts, it's streaming, it’s YouTube. And if you're not thinking that way, you're falling behind and probably missing a large chunk of your potential customer base. Since change is inevitable, at Veritone we always say, if you're not driving it forward, you're reacting. And if you're reacting, you're losing, whether it's share of voice or creative messaging.

That sounds like what Joe Jackman recently told The Continuum, too, about the need for brands to be “reinventionists.”

Yes, even interactions with brands have changed. Five or six years ago they were like, ‘Podcasting? Interesting, but weird.’ Now they’re coming to us saying, ‘Wow, we really need to be a part of this industry.’ But they’re overwhelmed by how big it is because it's not capped by affiliate distribution, so they wonder, ‘How the heck do you navigate hundreds of thousands of shows?’ We know not all of them are monetizable, but we're buying 30,000 unique shows on a monthly basis and that number is obsolete every single week. It's a billion-dollar plus industry. It's going to be multi-billion in a few years. So, from that perspective, how do we think differently outside of just the traditional ways of doing it and find growth from that?

---

“brands need to think of audio outside of ‘just radio.’ It’s also podcasts, it's streaming, it’s YouTube. And if you're not thinking that way, you're falling behind and probably missing a large chunk of your potential customer base.”

---

How do you advise your clients on navigating that and understanding the proper place to put their message?

It can be like fishing in the ocean. But our Veritone Performance Index takes all of our data—the hundreds of millions of dollars that we have in the data—and aggregates it to a variance based on like demographics. So, if a client comes in and wants to target females 25 to 44, we can pull up targeting based off of that and offer the types of shows that have been hitting performance metrics for B2C type companies.

It’s also about understanding the growth trajectory of shows. We all know a show can start out with 5,000 to 100,000 downloads a year, but where does it start to see a point of diminishing returns? And at what point do you adjust your strategy? Does it become top of the funnel, or does it become five to ten-ish percent of your budget, or is there a way to strategically flight the media differently so you're not just taking a one-size-fits-all approach?

The data helps us guide that. So when we put stuff on a plan, we know that these shows are working for our direct response clients in a variety of different verticals. We know the audience is engaged. We know they have money because they're responding and purchasing things on a heavy subscription or a heavy ROAS model, and you can do it with confidence.

And then, as you start to grow, you can start taking advantage of new shows. Sure, they bring some risks, but we also kind of classify them as undervalued stock—because if you get in a show and it works and grows, you get to ride that growth, and that's where the magic really happens.

I think back about the story some years ago about Serial, one of the first podcasts to really go mainstream. Not knowing what a hit they had, they sold sponsorships at low CPMs, and they were embedded ads—or “baked in”—because we didn't have digital ad insertion at the time. So those advertisers that got in early, say a Squarespace or MailChimp, ran to the bank. Tell us more about when to consider “embedded” versus digitally inserted ads.

The industry was built on embedded ads, which do live in perpetuity, and really helps classify the “audio on demand” element. It’s where the direct-to-consumer universe helped build the podcast business, because it works and you can have response coming in after the ads run longer out.

Obviously their delayed response is less aligned with, say, timely news or certain sports content, but we'll get orders from ads placed six, nine or twelve months prior. So people are still engaging with that content. That’s powerful because as we know with radio if you're not right there hitting them close to purchase time or hitting them at least three times that engagement isn’t there. That doesn't discount the value of radio because it's still the largest reach medium out there. But it helps you look at podcasts in a different way.

Now the podcast networks are looking at how to monetize every single impression. So, with Serial, they had the embedded ads, but now they’ll also sell new ads in the back catalogue since people can go back and listen to older episodes. As a result, embedded is not going anywhere.

Digital ad insertion (DAI) technology enables shows to sell back catalog, upfront catalog, or to insert ads at different spots in the show, and also helps from a production perspective, with the ability to go in and out. It’s almost like radio that way: you hear that impression and once it's gone, it's gone. So you have to devise different tactics and consider a frequency share-of-voice model.

DAI becomes very beneficial to heavy-up in a specific timeframe during huge promotional pushes, like a Black Friday or Cyber Monday. Clients can't wait the four or five weeks in the way embedded ads typically drive response.

DAI gives you more of a tactical approach and also is going to help with psychographic qualitative targeting by getting more of that first-party data. It just starts to build campaigns which appeal to the brand side because, in the end, there’s still not one universal standard of tracking, like a download or impression.

---

“The industry was built on embedded ads, which do live in perpetuity. you can have response coming in after the ads run longer out.”

---

What’s key to podcasting’s evolution in measurement and pricing now?  

I think the biggest growth area is the continued development of tracking. That’s from a selling and operational perspective, but brands have always been hesitant to come into something that's not standardized. So, you want them engaging with that and understanding.

The IAB has done a great job with getting it to that 24-hour lookback window, and it'll continue to evolve, but it's still not standardized. Until it's standardized, you still look at things as apples, oranges, and grapefruits a little bit. If you don't have a database to reference, a $7 CPM might sound really appealing, but we'll tell you you might want to pay $25 for this one because this audience responds. CPM is directional from that perspective.

Is there a typical CPM range?

There is. Embedded has lived in the $20 to $30 range for shows for quite some time. Shows that are more of a niche target are going to be higher CPMs for a smaller but highly qualified audience.

The industry is starting to realize that there's a different valuation for DAI. Is the host actually doing the ad read versus a producer? That's a huge difference in CPM. What are the number of ads that are in rotation, or the share of voice you're doing? Typically—though not universally—it’s going to be a lower CPM, in around the $15 range. However, that's not always the case if you’re targeting big, key, high-profile shows with celebrities, etc.

---

“we classify New Shows as undervalued stock—because if you get in a show and it works and grows, you get to ride that growth, and that's where the magic really happens.”

---

Would you say embedded ads are best for brand or awareness marketing? More of the long-tail kind of build, while digitally inserted ads are for the performance or demand side, with more instant trackability and often a call to action?

Honestly, you could argue both sides when it comes to awareness and performance goals. But the first thing to understand is what the client's goals are. For a brand like Coke, embedded ads typically have been the better performers. But there're also clients that want to test a lot of creative and understand the right awareness messaging. Embedded doesn't really afford you that creative test mentality because, while it's extremely powerful and integrated with the content and the influencers giving that endorsement, you're running copy points, and that’s very different from a produced or scripted ad.

Clients will often come in saying, ‘I need some kind of return to validate the channels.’ A lot of the strategy depends on if they are constantly changing offers and trying to do a lot of that testing. With an embedded ad, that might mean an offer that's living in some form of perpetuity that might be obsolete, and DAI becomes the right approach.

This ecosystem is going to work synergistically together. It's going to be impossible to just do one or the other, and frankly, from a beneficial perspective, you're going to want to do both. And clients just need to determine when they can start to validate that top of the funnel 10–15% of their budget that may not tie directly to, ‘I spent this much and it drove this much.’ But, of course, it's aiding brand search and things of that nature, which helps the overall pie of what they're trying to do with their transactional-based buy.

With the advancement of AI pixel tracking, it's not going away. It's just going to be what makes the right sense for client A versus client B versus client C.

As you mentioned, there are many channels to consider. How is your Marvel AI division supporting audio growth?

It’s true. You can't just buy two channels and cross your fingers and have a nice day. So, how do you diversify your portfolio? I'm focusing on how we test multiple channels from a cross-channel attribution element and make sure that we're not just thinking myopically, knowing that the exact person they're trying to hit might hear the podcast ad once. We need to hit them when they're listening to Spotify, Pandora, iHeart, etc., or also podcasts they consume on YouTube.

It's also building a sphere of influence that, when you start to layer additional channels, aids retention and not only builds the audience, but also creates chatter for a brand. From a DTC perspective, you tap the power of influencer marketing.

Then there’s Marvel.ai, our synthetic Voice-as-a-Service solution. It’s really amazing technology for creating a person's voice by generating enough of their audio to do text-to-speech or speech-to-speech elements. It's perfect for celebrity talent, voiceovers, or people that don’t have time to record.

For example, we do a lot of voiceover spots for founders with synthetic voice. They’re running a company so it's hard to get their time to record. But we can do text-to-speech that puts it out in their voice. The flexibility and nimbleness not only accelerates return times, but you can even apply accents and translate into different languages, which supports diversity and inclusion. That helps you amplify the power of an influencer or a brand ambassador who, these days, may not be able to get into a studio. Or, if you have a huge national campaign running in a lot of local markets that takes three, four hours every single week to dedicate to tags, we just take that element and still get the power of it out there.

---

“we're buying 30,000 unique shows on a monthly basis and that number is obsolete every single week. It can be like fishing in the ocean. So how do we think differently and find growth?”

---

How do you protect against white hat use of synthetic voice versus “deep fakes?”

By putting watermarks and protection elements in place where people are being exposed to a synthetic voice. And that's the balance of pushing the technology and growing it, but doing it from an ethical perspective to make sure that everyone is aware of what they're engaging with. It’s something we’re providing to our clients already.

For example, DraftKings spokesperson, Ross Tucker, a prominent influencer in sports, doesn't always have the time to do all those tags. So, we started running it with Marvel AI and we’ve seen nothing but success.

It’s also becoming extremely cost efficient from a pricing, and also logistical, perspective. It just opens up a world where the goal everybody wants is to constantly evolve their messaging, whether it's chronologically or telling a story.

December 1, 2021