EPISODE 15: Brendan Norman on How AI is Transforming Content and Advertising

Speaker A: Time you're building something. You know, the common founder trope is like, you go really deep into the tech because you built this thing and you're excited to share how it works. And the reality is knowing who the other person is on the other end of that, what they care about are results typically. And if you can help them understand why this is relevant while also understanding the tech. You know, there's a smaller subset of data nerds out here who love to dive into the, you know, like, how big is this data set? You know, all of the backend systems that are like, fun to talk about for a very small niche group of people.

Speaker B: Welcome to the Last Word on product Marketing. I'm Liza Chakowski, a B2B product marketing consultant and host of this podcast. Today's episode is about what it actually takes to bring a complex technical product to market, especially in a category that's evolving as quickly as ad tech. My guest is Brendan Norman, CEO and co founder of Classify. With Classify, he's building a contextual intelligence layer for the open web that's helping advertisers reach the right audiences without using cookies and helping publishers unlock more value from their content. Before founding Classify, Brendan was at Facebook where he helped launch the audience network and scale it into a multi billion dollar business. What I think is especially interesting about what Brendan is doing now is that it sits right at the intersection of a few big shifts happening at the same time. AI, contextual targeting and the agentic web where machines, not just humans, are now consuming and interpreting content, which is exciting and scary. So I'm excited to dig into what it looks like to build and market something like this in real time. So thanks for being here, Brendan.

Speaker A: Yeah, thanks for having me.

Speaker B: Sure. So you've been in ad tech through a few major shifts from Facebook's audience network, like I just mentioned, to now building Classify in a post cookie AI world. So what's changed in the market that made you feel like Classify, uh, needed to exist at this moment?

Speaker A: Yeah, that's a, that's a big question. Um, just like breaking out ad tech because I've been chatting a lot about like the different flavors and kind of different areas of ad tech. Like it's a huge world and mobile, mobile publishing, mobile gaming, you know, is very different than the traditional, you know, programmatic open web, ctv. The agency world, you know, is very different than the gaming focused world. So I've kind of had the luxury of sitting across a lot of different, uh, areas in this and I think some of the key observations that that I've made are the mobile gaming space is probably the most sophisticated. Um, they are obsessed with like every little data point and probably the most obscure data points that it's a completely separate language than in this open web programmatic space, um, where it's a lot of positioning and kind of longer term storytelling. Mobile gaming side are just really hardcore data nerds. So they're fully calculating metrics around what does it cost to acquire a new user, how can I monetize them? And looking at all different components of where do I put an ad, how do I track the efficiency of that ad, how do I make sure that I'm maximizing the value of a CPM for that specific ad versus you know, in addition to the user's retention, their engagement metrics. And they're kind of looking at this balance of like is the user in my game really enjoying this game and am I making a lot of money from them and do I get to keep them for a long time? And there's a lot of very complex math on the back end that goes into kind of figuring out all those things and kind of figuring out those balance. But I think that the ad tech in more the programmatic open website is kind of a different flavor and it's kind of interesting that there isn't as much of a focus on a lot of those types of metrics. You know, it's more around like how many impressions are we driving? Are we, is this campaign performing well? Are we making money from it? Great. And I think what's the interesting positioning is to kind of help to bridge some of that gap where there was a lack of, you know, sophistication and kind of technological, uh, advancement in how one, the web is being categorized and classified and organized. And then two, from a monetization standpoint and realizing that, you know, it has been fairly unsophisticated, folks have been very heavy audience targeting for a long time using things like cookies, the traditional type of identity based targeting where you're organizing around this very probabilistic. I think Liza might be this person, she falls into these buckets. Let's target these buckets to reach her. Uh, we took a very different approach and said let's just focus on the content for now. And that's a big missing piece of the. Serving the right ad to the right person at the right time at the right time is difficult to do using legacy tools where a lot of the web has been inaccurately categorized using a very kind of basic categorization in terms of Understanding this webpage is about sports and it's about soccer. And then we might also extract a couple of keywords around Messi and Adidas soccer, and that's it. That entire article might get reduced down to those couple of data points. We took a very different approach. And if it's an article that has a thousand words, there's a lot of nuance, there's a lot of potential keywords that might get blocked by one advertiser that might actually want to reach that article. There's a lot of specific semantic understanding that a human would read that content and interpret it in a very different way depending on their frame of mind. That would help an advertiser show the right ad to that person in the right moment. But you can't really do that if you're only looking at it and reduced to like sports, soccer, Adidas, messy. And we took a very different approach of saying, let's really understand what's going on in the page. Let's classify it, let's deeply understand what's happening on the page. And then two, the ability to search and find similar pages, um, that are similar semantically according to the way that we're trying to find similarity and then organize the whole web through that construct of, you know, deep classifications. The output of all of that backend data, science and tech is that it allows the right ad to show up to the right person at the right

Speaker B: time, which is what everyone has been claiming for a long time with cookies. Right. But it never really came to full fruition. Um, so it seems that AI presents the opportunity to truly be able to get that personalization layer and true context. Is that part of your thinking about something that seems really foundational, how content is actually understood and structured at, at scale?

Speaker A: Yeah. And, um, it is for a couple reasons. I, I think that the, the other missing layer to a lot of this is advertising, at the end of the day, is about good storytelling. And if you are reading an article and you see an ad that's relative or, uh, related to that specific article that you're reading, you're much more likely to eng. And even if you don't, you've had a better experience, you want to come back to that. But if you go to a website, you're reading an article, there's a bunch of just spammy content all over the place. And I won't name any names in the publisher space, but some just stack ads on top of each other and, you know, the actual content itself is minimal and you have to see 30 ads just to, to get the actual content terrible experience. And the reason that they have to do that is because they need to pay their own bills and they're doing it through this. Let's just put as many ads in there as possible. But the key lesson I learned in the mobile gaming space is you actually don't need to spam people with a ton of ads. If you just have really relevant ads, you know, the right advertiser that wants to reach that specific piece of content and that specific person, they're going to pay a lot more to do that. And user has a better experience. So all of this backend data science, understanding of what, what the content is, you know, just helps to facilitate a much better user experience, drive a better outcome for the advertiser and hopefully more money for the publisher.

Speaker B: Sure, sure. So when someone asks what does classify do, what's the hardest part to explain more clearly? And where do people kind of get lost? Because you've gotten even in this conversation, very general and then got very specific really quickly. Um, which is super, super interesting. But I'm just wondering how you kind of have thought about, uh, telling your

Speaker A: story super high level. Just saying that we help to organize content better so that ads can show up at the right time. That's a very high level piece.

Speaker B: Sure. Yeah.

Speaker A: It's fun. Anytime you're building something, you know, the, the common founder trope is like you, you go really deep into the tech because you built this thing and you're excited to share how it works. And the reality is knowing who the other person is on the other end of that, what they care about are results typically. And if you can help them understand why this is relevant while also understanding the tech. You know, there's a smaller subset of data nerds out here who love to dive into the, you know, like, how big is this data set? You know, all of the backend systems that are like fun to talk about for a very small niche group of people. To um, answer your question though, it has been an interesting challenge because we built one core engine that supports the advertiser world and the publisher world. And it is a very nuanced and kind of novel approach to organizing this data. Um, I'll share it with you afterward and feel free to drop it in the link. But there was a book published around 160 years ago called Flatland, and it's a, it's a Victorian novella that puts the reader in the shoes of a two dimensional character.

Speaker B: Mhm.

Speaker A: And this two dimensional character literally lives just on One flat plane. They're all these different shapes and they have their own nuance and their own perspective because they can only see one dimension or two dimensions. And they feel superior to the one dimensional creatures that they've observed and they feel even more superior to the, the no dimension creatures observed. And then all of a sudden one day there's a three dimensional creature that passes through this two dimensional plane. They don't know what the heck's going on. And the two dimensional character gets to go experience three dimensions for a short time, comes back and tries to explain it to um, the rest of the folks living in two dimensions. And they think one, he's crazy. Two, it's scary to let him communicate that there's more out there than they've observed. So they lock him up in prison and they basically control his voice. Fast forward to a parallel to like kind of the ad tech industry, you know, when we've organized data into just using keywords, it's kind of one dimensional, two dimensional, somewhat categorical taxonomy or a nested taxonomy of, you know, it's sport, soccer, you know, in addition to these keywords. And three dimensions are kind of understanding the relationships between how the specific piece of content relates to other pieces of content. It's kind of a dimensional framework for not just realizing like this piece of content is one thing. We realized that it's a lot more complex than just one thing. And then the other layer is when that ad shows up against it. How are you, Liza, interacting, uh, with both that piece of content and that ad. So there's even kind of a fourth dimension around entailment of understanding who the user is, what their experience is with that ad. And then we can kind of layer in all these different dimensions together that help really understand and then continuously optimize making sure that, you know, we are creating the best experience for users and we're driving the best outcomes for the advertisers.

Speaker B: Yeah. So it sounds like it's challenging to balance the depth of what you've built with telling a story that's going to land quickly and uh, it depends on the audience. Right. So do you find that you're mostly in front of more people who understand at a high level and want that level of information, or are you talking to more technical buyers

Speaker A: now? If we're talking to a media planning team or media buyers or traders, you just kind of jump into the very quick, this is how it's different than what you're used to.

Speaker B: Mhm.

Speaker A: There's a lot of contextual vendors that exist in the ecosystem. I originally thought that I'd left ad tech entirely and was building a platform for backcountry ski touring. And I went out and tried to market this platform and went out and chatted with a bunch of the contextual targeting providers and asked them to give me a sample of like, okay, I want to run ads. Show me which URLs and not just domains, but, like, which URLs you run this ad on. And nobody could do it at the URL level and nobody could do it with precision. You know, and backcountry ski touring is very different than resort touring, resort skiing, or cross country skiing. And there's a lot of nuance to this and I just didn't want to waste any money, any impressions on showing out to the wrong group. And nobody had that level of precision. So went out and figured out, you know, how to do it.

Speaker B: Oh, interesting. So that's kind of your origin story for Classified, that you were trying to run campaigns and found that the targeting was not specific enough and you didn't want to waste your money.

Speaker A: Exactly. And kind of thought that I was out of the ad tech world entirely and got pulled back in.

Speaker B: Got pulled back in.

Speaker A: One, solve this problem. And then two, realized that, like, oh, man, if this is a problem for my little project that I'm trying to build here, maybe it's a much bigger problem and maybe a lot, probably a lot more people are experiencing that too. So, um, it kind of evolved into, I put that project on the back burner for a while. Um, but yeah, realized that it's a helpful tool to have that level of granularity and accuracy. And if we can just extend that out, you know, we can kind of rebuild some of the best parts of, you know, Facebook still has one of the best ad tech platforms on the planet. When you go to Facebook or Meta now, I'm like, you show up with a campaign, you show up with the budget, you show up with who you want to, who you want to reach and what the outcomes you're trying to look for are. They're really good at knowing how to balance that across Instagram and Facebook and WhatsApp and the inventory that they have because they have so much data. And it's a good combination of knowing a whole lot about you, but it's also knowing what you're looking at at that moment. And it was kind of shocking that the rest of the open web has all these disparate pieces. It's a very confusing ecosystem. And if you look at the Lumascape, there are a lot of buckets for folks that play in the space, but nobody has really cracked the nut uh, of figuring out how to put all these pieces together. I'd argue that some of the large agencies and even some of the independent large agencies have been building their own and they've got amazing intelligence platforms that allow their traders to kind of use a lot of these tools together. And I think what's also really exciting right now about some of the new Magentic uh, protocols, it allows systems to be able to work much faster together using APIs, using natural language and not having to have all these very complex hops or having to migrate back and forth between web UIs where there's a lot of disconnect and a lot of data leakage, but there's some really cool new stuff that's like growing rapidly. And I think it's, it's an exciting time to be in this space again.

Speaker B: Yeah. What are you thinking about the kind of future of advertising or what have you learned by being a part building within um, a space that's so emerging? Um, what do you think the future holds?

Speaker A: You know, as someone who's an optimist and someone who's, who's realized that this entire industry like didn't exist, like the ad, the tech side of this industry didn't exist 50 years ago, 30 years ago. I think humans are really good at finding more stuff to work on. So while there will absolutely be some changes in terms of folks job functions with AI, I think ultimately we're all going to get a lot smarter, have a lot more data to drive better outcomes. And I see the industry overall in 10, 15 years being bigger than it is today, with a lot more data, with a lot more efficiency. And I'm excited about a lot of these new agentic protocols that are just allowing these systems to work well together because they allow it to talk faster, they allow it to reduce some of these middleman. And I think one of the stats is that when I was working at Audience Network, an advertiser would put a dollar in and Facebook would take about 30 cents of margin and paid the publisher out 70 cents. Today across the open web, an advertiser puts a dollar in and the publisher only makes around 30 cents. So 70% margin is going to a bunch of different DSPs, SSPs, data providers, like all across the board. And what is kind of exciting is that the folks who are focused on building more platform based plays that include a lot of pieces to this puzzle, whether they're building or acquiring are going to be able to provide a really outcomes based uh, product that is efficient, that doesn't take 70% margin where they can still take a healthy margin without having to use all these different pieces. And I think the streamlining, uh, and the cleaning up and the cutting out of the middlemen I think will be the exciting piece. And then it becomes much more of a competition around who is driving the best outcomes, who is the most transparent, you know, who is the easiest to work with and things like that. And that's going to become more of it as opposed to just justifying um, you know, their means.

Speaker B: So this is good news for publishers.

Speaker A: You think it should be good news for everybody. There's a lot of folks on the publisher, um, protection side and I think, I think that's a very important piece to the puzzle. There's a lot of tech platforms out there that are basically scraping and repackaging and selling publisher content. A lot of that information is getting used by developers. It's going back into context. Uh, Windows and chats, AI chats, those folks, you know, are basically stealing publisher content. And that's not great. There absolutely need to be a lot more protections for that level of interacting with publisher content. On the advertiser side, I mean I definitely believe in a, in a fair value exchange and working with publishers, working with advertisers, you know, sitting in that area kind of in between those two to make sure that with better data, uh, with better partnership on the publisher side, publishers can actually increase monetization.

Speaker B: Great, that's good. Um, yeah, they need help, the publisher side. Um, so this is a product marketing focused podcast. So um, certainly curious about how you're showing up in, in market, um, and in emerging categories. A lot of times the product clicks a little bit once people see it. So how much of your go to market relies on a demo and how much are you working kind of to tell the story to preface the demo? Um, um, I go, I see a lot of founder demos, um, and I think a lot of them are kind of reluctant to set the tone and the context for the problem they're solving. They set up, they can set up the demo a little bit better to really resonate, to solve that problem. So I'm just curious how you've approached um, that when you're selling in market.

Speaker A: That's another fun question to tackle because um, I think that you know, if you just go to our website, which is triclassify ah.com, anybody can build a contextual segment really quickly. So you can just pop in, you know, using natural language. I want to sell, you know, Nike shoes to people in Brooklyn.

Speaker B: Mhm.

Speaker A: Cool. Or dog owners in Brooklyn. Um, and you know, what we'll do is we'll generate a quick small list of the top 20 URLs that we think are the most relevant. Um, so you can kind of quickly see that and then a couple sliders to understand, you know, this is what I want to spend a month. This is our estimated impression volume, users that will reach some of the outcomes and then you know, the next steps to activate that segment. But to your, I guess to your question, the agency world does operate a little differently. And I think the ability to tell the story is also a very important piece for some people. So from a personal standpoint, learn that pretty quickly, you know, depending on who I'm talking to. Some folks want much more of the background story, how it works, you know, what we're doing, and then to see the demo, some folks just want to jump into it. I built a whole suite of uh, different calculators that kind of also help to estimate things like, you know, when you're running deeper, uh, URL level contextual targeting using this technology. Here are some expected outcomes that you can have in terms of increasing, um, outcomes. And on the flip side of that, also deeper classification on the publisher side or SSP side when you do really understand what's going on inside of your content. So I also love the idea of giving folks the tools to just check it out for themselves and see, oh man, I'm really missing out on, you know, hundreds of thousands, millions of dollars of revenue by not using this thing. That's the easiest way to just help explain, you know, the power of this tech.

Speaker B: Sure. I think that having interactive tools the way you've described is a really cool way to go about it because you can show it in your meeting and then you can give them something to hang on to and to play around with afterwards, which I think is a, it's a great way to build authority, um, you know, and just make it a little bit more experiential. I wish more people would come up with a more experiential demo. So you're building in a space that's obviously building really quickly. How do you balance for what the market understands today and where you think it's kind of going?

Speaker A: Having yeah, good case studies and just showing people, you know, whether it's their own data or whether it's other folks data, uh, the results and what's important I think is like what to lead with first. So you know, a lot of the time if it's an individual conversation, it's understanding. Have you tried contextual targeting on the advertising side? What, what are you looking to do? First of all, what tools are you using and then how effective are those tools and then that can kind of help inform more the direction of travel.

Speaker B: But the gap, showing that gap, I think is a really good way to go about it immediately.

Speaker A: Yeah, exactly. And then I mean to your point about product marketing as more like a thought leadership standpoint, um, I've also like spent some time chatting with some of the other folks that are also building similar tools. And it's cool because we all share the same or, and, or similar visions of the future. And I think that the desire to streamline tooling, the desire to make the process of just advertising more efficient and easier, is a very shared perspective for a lot of folks that are building in the space. And you know, every time I see a competitors and I, um, won't name names but like there's some really cool folks in the space that I really deeply respect in terms of, you know, their thought leadership, um, in terms of what they're building, how they're building it and you know, engage with them to try to figure out ways that we can help educate the market on, you know, this stage of this more cutting edge technology that everybody should be using.

Speaker B: Yeah, that's a huge opportunity. As much as the media, I feel like makes people think that everyone has an immediate deep understanding of all of these things. I think that there's tons of education to be done. As we all get used to working this way, there's always more to know. So thinking about how you decided to launch a company, you have all this experience scaling Facebook's audience network into a massive business. What from that experience, um, has helped you, um, in building this company.

Speaker A: So I think some of the biggest things that I learned were build long term strategies around where you want to get to and then also realize that there's a lot of testing and there's a lot of like fast iteration to kind of get to that more sustainable place. Um, realizing that building relationships with folks who are willing to test, you know, who wanted to see results quickly but were okay with a lot less stable product and you know, not having enterprise features baked in, um, and then just working really tightly with product and engineering to make sure that we were building the right tools at the right uh, stability metrics and just kind of being a conduit between, between a whole bunch of different marketing. We have to be out in market selling the thing that we're going to be building. And you know, we need to be working really tightly with the folks that we're already supporting to just really understand what their needs are. At the end of the day, I think it's just a lot of listening. It's kind of taking the consultancy mindset instead of saying you need to buy this thing because you know you should buy. Uh, was always help me understand your problems and help me understand, like, what are you currently using? What would you design in a perfect world? What do you not like about it? If you can make it better, how would you make it better? Uh, you know, and just keep peeling back the onion and understanding overhaul lots and lots and lots of relationships with folks in that space. You get some really important through lines. And I think you realize that you can never be just the one stop for everything, solving everybody's problem. But if you understand the market, the competition really well, to help them not, um, only get the best usage out of what we built, but also the competition. My goal is to help them make money and have a good user experience. And I knew that we were really strong in certain areas and maybe not as strong in other areas. So I build a lot more trust and credibility when I am able to say we're going to crush it here and we're not going to crush it here. And like, these guys are great for this problem. And on the flip side of that, translating that into ways that product and engineering teams and marketing teams can understand so that from a building standpoint and um, from a positioning standpoint, you know, we're all aligned on what we're saying in the short term, what we're building towards in the long term.

Speaker B: Yeah, absolutely. I think the unlock is when you truly start to understand that every conversation is about the customer. I mean, yes, you're offering a solution, but you, it's essential make sure that the solution you're offering is, like you said, what's the roi? How is it going to improve, you know, their experience, make them look good? It always comes down to the results that they're going to get. And I think people get a little tripped up and. But we have this amazing product and you should use it, you know, um, which like you said, it might be true, but it's not compelling enough to, um, make some on by. So this is kind of the section where I get slightly more personal with, with people.

Speaker A: Totally.

Speaker B: Um, what's a piece of advice you received that has stuck with you or something you've learned the hard way since becoming a founder, which even as a consultant, I can tell you is very different than being an employee.

Speaker A: So professionally there's a couple different types of folks. Some of those folks think in terms of, you know, this is my company and I'm just gonna, you know, I wear a lot of hats and I think dimensionally outside of this and I'll see something and I want to just get there, um, in the more startup mentality. And then some folks are more kind of corporate, long term, stable, task oriented. Um, I think it's great that we're all built differently.

Speaker B: Yeah.

Speaker A: And I think that few people are kind of built for this crazy founder world of building something where you are hair on fire, every moment is precious, it's your thing, you're so excited about it. Um, and then a lot of folks are amazing at what they do and they also like to work work, you know, in a big organization. They like to know consistently, like, here's my swim lane, here's what I do. Um, figuring that out is a really important thing to figure out. And I have worked in big companies and always kind of thought through the lens of like seeing everything at the same time knowing that if my swim lane was here, I might have ideas and everybody else's swim lane too. I have been really lucky to work on teams where that overlap of discussion around here's where we're headed has been very encouraged. So cross functional teams where you're working together to solve a similar problem and sharing ideas across your specific, you know, focus areas are the types of environments that I've always thrived in. Although the more I'm into this, the more I realize that like, I'm much more designed for this type of work or at least this ability to kind of context switch across a lot of different areas. Um, and then a lot of folks aren't. It's just knowing that about yourself. You know, if you're in a startup and you really love the big corporate world of like longer term planning, having a swim lane, you know, you kind of need people to help you set goals or set goals for you. And conversely, you know, if you wake up and you're thinking across, you know, ten different, uh, animals, you know, but you're supposed to just do this one thing and like kind of keep your mouth shut, like you're going to get frustrated very quickly if you aren't in an environment that helps support how you just design like your, your DNA. Yeah, I think that the best piece of advice was, was from a friend that I used to work with who left, built his big company, transitioned out of it after they had a huge success and had a discussion with him about that of just knowing like who you are and kind of what makes you tick.

Speaker B: I think there's people with more of an entrepreneurial spirit and get energized by that and people who feel over overwhelmed in that environment and really love being part of something bigger than yourself and uh, having your role to play environment. The important thing is to understand which one you prefer so that you can make the best choice for you. A lot of times though, you have to go through the process of the bigger organization in order to understand which lane you belong in. So as you wrap uh, up, how can people get in touch with you? Is it via LinkedIn or.

Speaker A: Yeah, uh, LinkedIn's great. Email's great. Either one. Happy to chat. If anybody has questions, you know, and wants to nerd out on this stuff or have ideas or poke around like, I love discussing this. So yeah, really appreciate you bringing me on and having a discussion.

Speaker B: Yeah, same here. Thanks so much for being on. I will, uh, put your contact information in the show notes along with your book recommendations and your, your company URL for triclassify. Uh, I love getting your honest perspective on what it's like to build a company at this stage. Be sure to subscribe to the podcast on all the major platforms and check out the YouTube channel for the videos. Thanks everyone for listening.