402: A $2 Billion Industry Built on Digital Duct Tape
Download MP3Hey. It's Arvid, and this is the Bootstrap founder. Today, I will share my insights into the podcasting universe at large from a software founder perspective. There are so many untapped opportunities in this space where budgets and expectations are growing massively right now. So why not just talk about the most challenging issues and what could be gained if those were solved?
Arvid:The episode is sponsored by paddle.com, my merchant of record payment provider of choice, who's been helping me focus on PodScan, my own podcasting business from day one. They're taking care of all the things related to money so that founders like you and me can focus on building the things that only we can build, and Paddle handles all the rest, sales tax, credit cards failing. All of that, I don't have to take care of because they do. I highly recommend it, so please check out paddle.com. Over the last eighteen months or so, I've been building PodScan, which is a business that processes millions of podcast episodes every month, tens of thousands of new episodes every day.
Arvid:With 33,000,000 episodes right now in our database, it's quite something. We transcribe everything and do AI based analysis and allow for search and mention tracking for brands, marketers, PR people. That's kinda what this business is about. And here's what I've learned. The success stories in podcasting, they happen despite the infrastructure of the space, not because of it.
Arvid:Podcasting is not making it easy, and podcasting is clearly having a moment. It's a 2.1 ish billion dollar industry in advertising revenue alone, and that number is oscillating a little bit, which is one of the challenges that we actually have in the space. I'll get to that later. But there's hundreds of millions of people in The US listen to podcasts every month. Acquisitions are happening all the time and everywhere in this industry.
Arvid:But underneath all of this momentum is a technical infrastructure that feels like it's held together with digital duct tape or in 20 old protocols that were never designed for what we're asking them to do right now. So today, I wanna break down the biggest technical and social problems holding this particular industry back because every single one of these problems represents this massive opportunity for founders who wanna build something meaningful. And I'm right in there with my own solution to some of them. But let's start with the foundation here. Podcasting is built on RSS feeds.
Arvid:It's surprising, really, if you look into the space for the first time because you wouldn't expect that something that is so central as a concept, right, the podcast as an idea to be just so wildly distributed all over the place. And RSS is just that. It's a tech from the year 2000 that was designed around block syndication. It was meant to distribute text files or links to blog posts that were easily digestible as text, and now we're using it to distribute gigabyte audio files to millions of people all the time all over the place. So think about that for a moment.
Arvid:We're taking this protocol designed for sharing lightweight text content, and we're asking it to handle massive audio distribution at scale. Now this audio stuff is not in the RSS feed. It's just linked from there. But there is so much data exchange happening around these feeds that I don't think they're built for it, really. And this shows.
Arvid:There's little standardization in the whole podcasting ecosystem, even with efforts like podcasting two point o, which is a kind of a protocol or a kind of a convention of using certain tags inside this RSS feed for podcast stuff like transcripts or categorization, like iTunes, whatever, that those kind of things. Episode numbering is inconsistent or sometimes missing entirely. Description are often malformed. People just put HTML into fields that should be text, or they claim to include some kind of transcript but actually dump a full HTML page into that field too. Publish dates are completely arbitrary, and different platforms handle and interpret these RSS feeds by themselves differently as well.
Arvid:I'm exposed to all of this with Podscan because I'm following these 3,800,000 RSS feeds that currently exists in the podcasting universe every single day. I've seen everything, all kinds of weird things in there to this point. And every podcast hosting company, like all the companies where you can upload your audio file, distribute it to all these, like, services and to your listeners. Eventually, they have their own way of providing and updating and caching these feeds and the information within. And everything is decentralized, which sounds good in theory, but it obviously creates chaos in practice because people just do what's easy, not what is best for this community.
Arvid:So that's the one side of podcasting. It's already quite complicated. And then on the other side, you have the walled gardens. You have Apple and Spotify, YouTube, those places. These are proprietary systems with their own internal hosting and listening apps.
Arvid:Right? The Spotify app only plays Spotify content. Apple Podcast is exclusive to Apple's ecosystem. And from a founder's perspective, these platforms are quite difficult to work with. Like, it's hard to authenticate against.
Arvid:It's hard to rate limit, and it's hard to have just consistent access to their APIs because they don't want you to. It's their stuff. Right? They wanna keep it. And when you process podcasts at a scale that we do at PodScan, edge cases become the norm.
Arvid:All those weird little podcasts with their individual quirks, they just add up to become a systematic problem. Here's what I mean. Over 60% of shows are missing basic metadata, like proper categorization. And even if they have categorization, it's not necessarily correct because people self select into categories. And so they pick the ones that they think are gonna get the most listeners, not the ones that are actually accurate to the content of the show.
Arvid:They don't have good descriptions. They just copy and paste something, or they have an AI draft something that doesn't really know what the show is about, or maybe worse, that doesn't speak the language of the listener. So you have a description that doesn't accurately reflect what people are gonna get from the show. And contact emails are often hit or miss, so you can't really reach the show. Show titles, episodes are formatted completely differently across different platforms.
Arvid:Some people use markdown, some platforms do, some don't, some HTML, or some use no formatting at all. So you see all the formatting if you copy and paste markdown in there. Horrible. Data is a problem. Data quality is a problem.
Arvid:So that's kind of my my number one problem in the field that I see that should be worked on a little bit is data quality. Language detection is another example. And I'm just rattling off all these examples here, but this is stuff that I have come to have to deal with in ingesting all of this data that I think, if solved, might actually be a net benefit to the podcasting ecosystem at large and individual contributors as well. Right? We've processed episodes that claim to be English speaking but are actually Spanish or the other way around, and there's no reliable way to determine explicit content as well without actually analyzing the audio itself, which is what we do at PodScan.
Arvid:But there's also something quite devious in there in terms of data quality, which is duplication. Like, popular podcasts that are out there and have been out there for years, well, they likely have switched hosting companies over their lifetime. Often, that leads them to a point where they have duplicate feeds. You will have one feed with the first 200 episodes of a show, and then at that point, they switched providers and created a new feed and then brought over most of their older episodes. And then the new episodes are on there, so you you have slightly overlapping duplicate content scattered across multiple feeds.
Arvid:And there's no clear way to deduplicate or connect them because people just forget about it, or they don't update them anymore because they nudge their listeners to go to the new feeder, and most of them do, and that's fine. So there's a lot of legacy data in the ecosystem as it stands that is just duplicate information, and it's hard to figure this out in an automated fashion. But why I'm saying this is there's probably a way that AI could detect this. Right? You could even like, if I think about it, you could probably go into the PodScan API and fetch a lot of podcasts and deduplicate it just from the data that we have in there already.
Arvid:We don't do this behind the scenes because that is not the idea of the platform. We just track all the feeds out there, but there could be a service in just deduplicating this and making this kind of information available. But that's just the data quality side of things. I think what is an even bigger and maybe even more frustrating problem for anybody in the field is when they're trying to build a business around podcasts. Because in web content, you can track exactly, like, how much somebody reads on a website, which parts of your page they engage with, right, of a blog, how long they spent on each section.
Arvid:This can be easily gathered, this kind of information with JavaScript tooling, this screen recording. All these things exist in the world of the web. And if you look at videos, it's the same. YouTube knows precisely where somebody started watching, where they paused, how long they stayed engaged, if they started scrolling through the comments. Like, they get this.
Arvid:But podcasting has a measurement black hole. With RSS based distribution, there's no way around this right now. All you know as a podcast owner, as a publisher, is that someone's device requested the MP3 file. That's all you get in terms of analytics. You don't know if they actually listened to it.
Arvid:You don't know how long they listened. You don't even know if it was a human that listened because Podscan for example technically downloads all these files for analysis which is one of many automated systems that are crawling podcast feeds and maybe someone subscribes to your show but never actually listens or maybe they're your biggest fan and listen to every episode immediately you have no idea you don't know because there's no unified way to report listening behavior. And Apple and Spotify have this information. They do know who listens and where and for how long and where they stop and where they restart, all of this, because they control both the file delivery and the player in which the file is played they know everything when you pause when you skip whether you're using headphones or whether the app is in the foreground or playing in the background but this information stays locked in their walled gardens That's why they are so aggressively acquiring companies to get more data in is to make it available to themselves and their users, but nobody else. The rest of the ecosystem is trying to get this data from somewhere, and they rely on this complex chain of tracking links, like URLs that bounce through multiple analytics companies before finally delivering the audio file.
Arvid:It's like a digital Rube Goldberg machine. It's really wild. It's the only way to get any measurement at all at this point. Look at this If you ever go to a website that hosts podcast episodes, like, just as an m p three or that has links to them, like, any large podcast that has a website where you can listen to these files online, check out the actual URLs of where the audio file is hosted. And you will see this cascade of eight to 16 services that all have domains that can bounce the link from one domain to the other for each to have a tracking event happen.
Arvid:It's wild. Some for marketing, some for distribution, some for I don't know what. Like, there are just the wildest services out there, sometimes domains that you don't even know what service they belong to just to get some kind of analytics happening in this field. And that is incredibly hard for people to deal with because this measurement, this information does not exist. It's really hard to get by.
Arvid:And to just find this information, you often have to either go to the the big platforms or you have to do estimation, which is what we do at Podscan, just to get some semblance of analytics in this field. And this has side effects. Right? It makes it harder to know exactly what a podcast is and who is it for and where should we place it if people were searching for it. So discovery in the podcasting universe is also a fundamentally broken issue here.
Arvid:Most podcast platforms right now can only search titles and descriptions. That's where discovery is right now. You have a keyword and you find it in the title and the description. If not, you wouldn't know. But what if, I don't know, your competitor gets mentioned at minute 23 out of a forty five minute episode.
Arvid:How are you gonna find this? How are you gonna find this kind of content on established platforms? So at Podscan, we solve this right now because we are transcribing everything, like every single podcast out there, and we make the full content searchable both for humans to actually type and search and for automated systems to do alerting on keywords and semantic search as well. But imagine the possibilities if this was available everywhere. If this kind of search was facilitated for discovery, you could search semantically.
Arvid:Like, I wanna find a podcast for kids that talks about dinosaurs and how they came to be. Like, you wouldn't need exact keyword matches for specific dinosaurs. You could just say this, and the system would understand the intent and context. This would be a really interesting tool to have and to build other tools upon. And this discovery problem extends to competitive intelligence too if you wanna look at more from a brand angle.
Arvid:If you're trying to understand how your competitors are perceived, what brands are being discussed even around you in your ecosystem, or what kind of themes are trending in your industry, you're essentially flying blind unless you have sophisticated transcription and analysis tools. Again, that's what I'm trying to build here. But I'm not the only person to solve this, and there are other ways of solving this too. So I just wanna give you this problem of discovery as something to think about. There's so much valuable data buried in podcasts, both in the conversation and in the consumption around it.
Arvid:Demographics based insights, that's not enough. There is more. There's themes and entities and sponsors and relationships and topic trends, but it's all locked away because we can't and don't yet do effectively search or analyze the actual content of podcasts. And that has a cascading effect which is probably the biggest financial hurdle in this whole industry because monetizing in podcasting is so limited right now making money with a podcast is so much harder than making money with a newsletter or making money with a blog or making money on YouTube. Right now you have injected ads and sponsorship reads like inside the episode and that's basically it.
Arvid:Maybe there's Patreon as well where you can have people pay for a subscription to a podcast, but that is hard to do for shows that are, let's say, less niche where you can tap into communities that already exist. And I think the limitation stems directly from the measurement problem that I just described. Why we're stuck in a CPM world, like a cost per mil, cost per thousand impressions, when we should be moving towards a CPC world, a cost per click, or even better, cost per conversion. But you can't optimize for clicks or conversions when you don't even know if people are listening to the thing that you wanna track. Right?
Arvid:Spotify is, again, an example of doing that work. Like, they're doing interesting work here with their mobile app. I listen to it a lot to podcasts on Spotify. And so when there's an ad, you can now click through directly, and Spotify can track that engagement and see if the ad actually worked. But this doesn't exist in the broader, more chaotic world of RSS based podcasting.
Arvid:And I always have to look at this both from the side of somebody working in this field and as a consumer. As a consumer, I don't necessarily love ads, and I don't love injected ads, like randomly injected ads. But if there were ads in a podcast that then makes this podcast possible for me to listen to, then I would like those ads to be thematically relevant. I would like them to actually mean something in the context of the show, not just, you know, a bank telling me that I could put my money in their vault somewhere because they know that people have money and anybody listen to a podcast episode might have some money to put into their bank. That's not specific enough.
Arvid:If I'm listening to, I don't know, a Warhammer podcast, which I sometimes do. I listen to my my nerd stuff. I want nerd ads. Why don't I get ads for other podcasts in the gaming space, for example? Right?
Arvid:Why does this have to be read directly, like, through a sponsorship agreement between these two podcasts? Why can't that also be automatically injected and facilitate a good connection between the targets for ads and the people who run them? I think that should be possible. And right now, all we have is demographic targeting. And that is rough right now at best, like basic age, gender, location, if that even works.
Arvid:But there's no behavioral or interest based targeting here because it's so hard to gather that information. And context of these shows and of the audience of these shows is completely ignored. Ads are mostly inserted without understanding episode content, and that's honestly pretty stupid. If somebody's talking about productivity tools, that would be a perfect time for a productivity app ad. Right?
Arvid:It feels right to have this. I think particularly in the context of podcasts where there is so much context. Brand safety, another problem, still unsolved. There's no reliable way to ensure that ads don't appear next to inappropriate content for that particular company. And ROI tracking, forget about that completely.
Arvid:You don't know how many people actually hear your ad, let alone acted on it because there's no way to trace this. Or the people that are working on this still struggle and the infrastructure doesn't make it easy. There are ways, but it's hard. And I think we need more ways that make it easier. So here's why I'm excited about these problems, because I'm just complaining.
Arvid:Right? But every single one of these represents a massive business opportunity, I believe. The companies that will solve these infrastructure problems, particularly measurement problem that seems to be centric, the core problem for all the other ones too, they will enable the next wave of innovation in the podcasting ecosystem. Just like Google enabled the innovation in search by thinking about monetizing it through ads. We might not like ads, but they facilitated Google giving us the knowledge of the world, which is now being completely absorbed into LLM systems and AI companies.
Arvid:But that's a different conversation for a different day. I think we're still in the early days of what's possible in podcasting when the podcast content becomes fully searchable and analyzable. Just think about the applications you could be building. Right? We think about recommendation engines that are based on actual content, not just the metadata, not just the title and the description, but the vibe of a conversation or real competitive intelligence for brands monitoring dimensions.
Arvid:And we're building this at podscan.fm, obviously. It's not an ad for Podscan, although this whole podcast might be at this point. But it is something that can be taken even further in this space. Contextual advertising that actually makes sense. That's something I would really like to see because I hate noncontextual ads.
Arvid:Content intelligence that reveals trends as they happen, which then also facilitates better contextualization for marketing efforts, for sales efforts, for activism, for outreach. But if you know what people are talking about and you wanna find your people and do stuff in this world, then having tools that can pull this out of the conversations as they happen, that's valuable. And ultimately, tools that help creators understand their audience better and improve listener retention, that's a benefit to everybody involved. Listeners get more of what they want, creators get more listeners, and anybody between these facilitators, sponsors, advertisers get better access to the people that actually might need what they offer. So with better infrastructure and tooling, we could get much better at understanding the content that's out there and actually use it more effectively.
Arvid:Monitoring for brands, tracking social movements, figuring out who should be a guest on your show, emerging trends, all of this becomes possible with the right foundation. And here's something that I've learned from processing 33,000,000 podcast episodes, which is wild if you think about it. The more podcast data you process, the better your understanding becomes. Even though every feed has its own quirks, but individual show analysis, just downloading all episodes of one show, transcribing them and whatever, this could never reveal the insights that emerge from processing data at scale. So every company that gets deeper into podcasting data is building a data moat at this point.
Arvid:The patterns, the connections, anomalies, outliers, they only become visible when you're looking at the entire ecosystem. So for these problems that we're experiencing in this field, the question is not whether somebody will solve them. It's who will solve them first and who will be positioned to benefit from the market that then emerges on top of this. Because podcasting is growing. It's happening right now.
Arvid:And despite all of these massive technical limitations, people still tune in more and more. So Imagine what becomes possible when we fix the foundation or when we improve upon it. I think about this every day as I'm building Podscan. Every one of these problems is solvable with the right technical approach and the willingness to acknowledge that it won't be perfect from the start or maybe ever, but we can build something that enables new solutions to be built on top of the old ones. So if you're looking for a space to build something meaningful, to solve these real problems that affect millions of creators and, at this point probably billions of listeners, the podcasting infrastructure as a space is wide open.
Arvid:The industry is ready for solutions. I think the market is here. So the problems are real and well defined. The question really just is who's gonna step up and solve them. And that's it for today.
Arvid:Thank you so much for listening to the Bootstrap Founder. You can find me on Twitter at avidkal, a r v I d k a h l. If you wanna support me and this show, please share podscan.fm with your professional peers and those who you think will benefit from tracking mentions of their brands, businesses, and names on podcasts out there. We're near real time podcast database with a really good API and a lot of interesting use cases like competitive intelligence and urgent crisis management. So please share the word with those who need to stay on top of what is being talked about in the podcast ecosystem.
Arvid:Thank you so much for listening. Have a wonderful day and bye bye.
Creators and Guests
