 And welcome. I'm glad we're all back at Slash. It's happening again. You've got another person from Sifted moderating another panel. I hope you're all happy with that. I'm here with Gustaf from Spotify. You've been at Spotify for over a decade. A long time. Can you give a very brief whistle-stop tour through what you've done in that time? The two-hour version for the four-hour version. Two-minute version, please. All right. So I joined Spotify in 2008, 2009 to head up mobile. Because back then, mobile was actually a thing. It was sort of a skill. Today, always say, if you don't do mobile, mobile doesn't even mean anything. Back then, it was a skill. So I got lucky. I joined to head up mobile product development. And then after a few years, I took over all of Spotify's product development. And today, my role is sort of Chief Product Officer, Chief Technology Officer for Spotify. So I'm responsible for all the products, the applications that you use, all the platforms underneath, all the machine learning and servers and that kind of stuff. That's the short version of it. Very good. You followed your instructions precisely. And so a few years ago, you launched Spotify's kind of first, next product, which was podcasts. Why, like, when did, I guess, why did you decide to launch a new product to start with? Just keep doing what you were doing forever and ever and ever. Yeah, it's funny that you call it the, you struggle there. It's like the first, second, and we actually had a project internally called Spotify's first, second, to figure out what the second product was. And this is a tricky thing because doing your next product or laddering up, as some people would call it, if it works, people are going to say it's brilliant. And if it doesn't work, they're going to say you're unfocused and you should have kept doing what you were doing. So you kind of never know until afterwards. And Spotify had a few failed attempts at trying to sort of ladder up. And so there was actually a lot of fear internally in, and if we were going to do this. However, we had a lot of signs on the platform that this was something that people wanted to do. Both externally, we actually saw people sort of uploading non-music content as music to the platform, so audiobooks and podcasts. They were sort of trying to hack the platform. We also saw our own employees, which time and time again have been a really good predictor actually of the future. Many of our employees during hack days and hack weeks that we have, they started simply implementing an RSS podcast player into their music player. So we saw these signs that, you know, our sample size of a thousand employees, they wanted it. We saw many people trying to do it. So we decided to take it seriously. We kind of investigated the market, gathered a lot of data and decided to go for it. We actually took it, we have this internal prior list called the company bets board, which is how we prioritize what we do at Spotify. And Daniel decided to simply make it the number one priority for two, three years almost. And were you, to take a step back, were you already in the mindset when you were like, we need to create a new product? Or was it more led by you were like, wait a sec, everyone seems to want podcasts? It wasn't that we needed to create a new product. I think that's dangerous to do. It was much more organic than that. The way we thought about it was we were all using podcasts ourselves. And there was this question, you know, about actually from the music industry even would podcasts compete with music listening eventually? And the way we thought about it is, well, maybe it doesn't, maybe it does. But either way, if it is going to compete with music listening, then it's going to compete with us anyway. We don't control the world. So it's better that we do it then. And so we thought about it that, you know, we had a pretty good idea of which moments part of I was relevant in. Those moments like when you're driving to work, when you're running and so forth. And it became pretty clear that these were exactly the same moments when you would listen to a podcast. And so for us, we just saw it as very adjacent to what we were doing. And we saw ourselves using these two things together. So from that point of view, it was actually quite organic. Then there was a big strategic decision to go for it in terms of, you know, would we lose focus on music and so forth. And so once you've reached that point, what happened then? Did you go and look at like what every single other podcast maker provider did? Or were you actually like we're going into a Spotify creation bubble and we're not looking at anyone else and we're starting from the basics. Like how did you run that process? That's a good question. You know, I talk about you should look at competition to understand what they do. But most of the day you can do something very different. And so we actually spend quite a lot of time looking at competition. But ideally not to copy what they do because they probably do it because they're great at it and you're going to be second or third to that. So we certainly looked at all the applications out there. We tried to understand what they were good at, what they were not good at, if there were opportunities there. We obviously saw that there were some applications that were vastly dominant, vastly dominant over the others. And even though there were other experiences that might have actually been better, but no one knew about them. So we tried to understand the market and we realized that we could certainly build a really good experience. But our biggest challenge would be to let get anyone to actually know that this experience even existed. So I think the most controversial decision when we try to map to lay out a plan for why should we be in podcast? What is it that we're going to bring to this world? And why could we be good at it? We figured out that our big opportunity was that if it was true that users actually wanted their podcast with their music, then we had a fantastic opportunity because we had hundreds of millions of people listening to music. But if it was true that they wanted a separate experience, then we had no right to play there. We wouldn't have any advantage over anyone else starting from scratch and certainly not against the big incumbents. So this was one major decision that we tried to figure out if that was an opportunity. We figured out that it was and then we chose to integrate instead of building two apps. And that was a painful decision because it's much, much harder to build a good experience when you have to balance music listening with podcast listening and the skip button should turn into a scrub button. Much more complicated than two apps, but we decided to do that. Then we also on the product and content strategy side saw an opportunity to work with exclusive content. The way that Netflix had worked with exclusive content, that doesn't exist in music because in music no one wants exclusivities, they would just break the play listing experience. So that's how we thought about it. We saw the space, we thought it was going to grow, but we don't just jump in. We try to figure out, can we do something unique here? So tell us a bit more about this, why you decided not to create a separate app. Because I guess also you have a challenge that lots of people here don't have yet which is that you have so many users that I guess the danger of creating something new and then mocking up what you already have is that much greater, isn't it? So I guess that also is a concern when it's being built into the app that millions of people are already using. The question for us was the obvious question. If you put the podcast experience into the music app, do you get the best of both worlds or do you get the worst of both worlds? I mean, it's a tricky thing. Often when people say it's the best of two worlds, it actually means that they're just doing the lowest common denominator and it's horrible at both things. So it was a hard decision and we prototyped a lot. But we saw that longer term, we actually think that it's going to be audio and we really believe that. And there was obviously lots of precedence for people mixing music and talk, for example, listening to radio. And so in our vision, we said that not only is it a quote-unquote distribution mechanism which sounds very businessy, we actually think they're going to be experiences where a host can seamlessly talk about a song and then you can legally listen to a song and royalties are paid out to the right artist in the same experience. And that couldn't be built if they were separate apps. So we had a vision for how this would evolve into something into a better experience and we made a bet on that. Then it was a huge challenge for our design teams to start creating these UIs that kind of change behavior dynamically which at the time wasn't very common, like morphing UIs. So I guess I'll just give it to them. A fantastic design team that took on a new challenge and executed on it. And how, when it comes to your team, they've been doing this one thing for a long time. It works well, customers love it. Like how do you kind of lead them through that? Okay, we're going to do a podcast now. This is going to be the thing. Like are people generally excited about that or are they kind of nervous? I would say both. In a big company, you're going to have people that are fully dedicated to one side. For example, to the music side. Of course, there will be fear of what if we lose focus now? Weren't we about music? What will artists think? What will users think? So there was certainly fear. But what we tend to do at Spotify is... But I would moderate that. I would say most people at Spotify, are quite excited about building new things and trying new things. So it's not that hard, especially not on the product side. But what we tend to do is we tend to try to debate quite a lot. Maybe versus other companies. It's not a very top-down company. And I think that comes from Daniel. He's just not a very top-down guy. So we try to debate a lot and anchor decisions quite a bit. And it's a slow process. It can be quite painful. And especially for American executives who are used to command and conquer and reporting lines and DRIs and all of these things. It can seem very frustrating. I've literally heard people saying, does everyone in this company get to have an opinion? And they actually do. So it takes time. But the benefit is if you take that time to debate, two things happen. The idea gets vastly better. Because the first idea is usually not as good as you think. But if you don't talk to it about anyone, it never gets shot down. So the idea gets pushed around and gets better and better. And if you're going to explain this and convince people, you have to sharpen your narrative. So after a while, the idea becomes better. But the second feature of that discussion is that everyone who participated in that now feels that it's actually also their idea. It's like they're not getting it handed to them. So we took quite a lot of time to strategize and discuss around this in bigger groups. So when we kind of pushed the button and made this the number one company bet, it wasn't that hard. We didn't do it against anyone's will. But we probably spent six, seven months debating the decision before we pulled the trigger. Do you ever spend too long debating stuff? For sure. Of course we do. But as I said about competition, we figure out what the competition does and then do something completely different. And specifically at the time, a lot of the talk in Silicon Valley and the competition was about code decides arguments, move fast, break things. Everything was pushed against spending minimum time on decisions and just don't discuss just code and see what happens. And so I wanted to go as much against that as possible, which means you sometimes go too far. So I deliberately coined another phrase, which is talk is cheap. So we should do a lot of it. It's much cheaper than writing code or being wrong with a live product. So yes, we over-indexed on it. And that probably means we're too slow on some decisions. But we also avoided some pretty big pitfalls in our history. You mentioned at the beginning that you had launched some other first, next product or whatever stupid phrase I'm using. What were they and what did you learn from those failures? So we tried a video, video slash audio quite seriously quite early several years before this. I don't remember if it was 2015, 2016 or something. And the big learning there was it looked great on paper strategically. You take content from different audience, your license, like the top content that performs the best and you put it together. Of course, if you take the cream and the crop of everything and put it together, it's going to be great, right? But it wasn't because the thing is that these things made sense stand alone like a few comedy shows and a few TV shows or something. But when you put them together, the product didn't make any sense to any user. So first of all, the user came in there saying like, I'm coming in here to listen to my playlist. Most users who open Spotify, especially back then, they already knew what they wanted to do. They went straight to the library. So you come in for a separate use case, which is music. You get interrupted on the way with something annoying and it says, now you're going to watch this comedy video. And then you're thinking, first of all, I didn't want to do that. Secondly, so is this down a comedy app? Do they have all the comedy? No? Just a few, okay? So, but is it a documentary app? No, it's just, it's like, according to Bumbling Theory and Strategy, it made sense. But I literally felt that as a consumer, it doesn't make that much sense and it also didn't work. So the big learning there was to not overly rely on abstract strategy, but also you have to be able to explain what the consumer is supposed to think about this and why it makes sense to them coherently. And with podcasts, it was so much clearer. People had been listening to talk and music all the time and you could clearly see a vision where it's even talking music in the same experience. And how do you test things out like that? Is it you roll it out to a subset of users? Like, what's your process generally there? It's case by case. Podcast was actually pretty easy because, as I said, people literally built podcasts using RSS into the client, so you could try it yourself. And one thing I've found is, even though there's a risk of looking at your own employees as a representative sample of the population, because they largely aren't, especially if you're only in Silicon Valley or only in Stockholm, it's still largely true that they are predictive of the future, right? It's just early adopters. Rather than being different, they're often just early. So we found that using your employees as a sample group and yourself is quite, using only yourself is dangerous. But if you use a thousand people, it's a decent sample size. So we actually test things on our employees all the time internally. And if you're lucky, like around podcasts, you can build this and it's not very expensive to get a feel for if what you have in your mind actually makes sense when you use it. So we do a lot of that. We test it internally. So when we actually roll it out, we are pretty certain, but we still roll it out carefully. We roll it out to 1%, try to understand what happens, and then adapt and roll it out. And it can take a long time for us to roll out big changes. I think someone said famously, changing an application is like going into someone's desk and just rearranging all the papers and saying, it's much better. Even if it is better, people are going to get upset. So these rollouts are very tricky. You want to communicate with your consumers about it. You want people to allow people or earlier doctors to opt in and like it and start talking about it and so forth. So how do you do that communication piece? Is it you like just, oh, this thing has changed in the app, or do you actually need to have, do you need to be more like upfront about it with people? We try to do PR around it. Not because many people read PR about our launches, but the early adopters do, and the super fans do. So you can communicate with them through PR, social media, Twitter, et cetera. Then as I said, we also try to do opt-ins where the people who are excited about trying new things, they get to try it. And we see how it behaves with them and you get ambassadors and advocates. And then we have a huge and fantastic customer support team who also spends a lot of time not just answering complaints but trying to talk to, because the most passionate consumers will go there. So it's a good representative of the people who care the most. So they go on Twitter and they go on the forums and they discuss with our audience sometimes about how we thought about decisions and why and so forth. Interesting. And so another part of, I guess, launching podcasts was that you made some acquisitions. So you bought Gimlet Media and you bought Anchor Howl. Again, what's the thinking around when it's better to build something in-house versus go and get something great that's already out there? Yeah, so part of our podcast strategy was obviously that we decided to integrate with music, as I said, which gave us a good distribution channel and the potential to build experiences that could be built separately. But another part, as you mentioned, we made a bet that exclusive content would make sense in podcasts, whereas it didn't make sense in music. And that a podcast was much more like a Netflix series. Like, Narcos kind of makes sense, even though it's exclusive to Netflix. But having one episode of Narcos on Netflix and one on HBO wouldn't make sense, you know? So we decided then that we need to become a content company. And we were very much a technology company, although we have probably the most musicians of any technology company. Every developer at Spotify plays some instrument, basically. But we were still a technology company. So we decided to acquire our way into this new skill, to hire and acquire. So we first hired Don Ostrov, who comes from the media side, to head up the content department. Then she made a series of acquisitions of studios to get the skill into the company. So that was on the content side to get the skill to build original and exclusive content. But then we also did something else that was very different from music. In the music business, we dealt mostly with labels, right? A few big ones and many smaller ones. But we very rarely dealt with artists ourselves, themselves back then. We had no direct connection. We focused most on the consumer. And we made these deals with labels. In podcasting, because it's not controlled by a few labels, it's direct publishing, we both needed to and wanted to get into direct relationship with creators. And because we had been building software for consumers for so long, but never for creators, we didn't really have any skill. So we also made another acquisition in a company named Anchor. That was by then one of the larger, not the largest, but at least one of the larger, back then, podcasting creating solutions. And they took a novel take on that. They were mobile first. They had free hosting and they were attracting a lot of creators. So we made a bet on sort of buying a product team that had focused on what creators want, not just what consumers want. And that turned out to be very fortuitous for us because we quickly went from zero to at least one. And now I think we're starting to get pretty good at building software for creators as well. Both in podcasts and also in music nowadays, actually. But back then we were a pure consumer company. So I think when you want to enter a new space, you can accelerate that through acquisitions. Okay. And talking of new spaces, you're now doing your next, next thing, which is audio books. Why? Well, first of all, from what I just said, the pattern is pretty clear, right? So, you know, we have music creators, we have podcast creators, and we made acquisitions to quickly get people to understand these spaces, right? And so it's the same with audio books. We announced this acquisition, Find A Way, which gets us from zero to one from not understanding that much about the audio book space as well to through them understanding quite a lot and having relationships. So it's the same strategy, you know, when you don't have a skill, it can take a long time to build up. So that's the reason to do the acquisition. The reason why we're going into audio books, if you think of that, you could think of it as a third product after podcast, but the other way to think of it as is that it's actually more adjacent. You could think of audio books as it's just talk audio, like podcasts, but it's paid. And there are two ways to monetize talk audio. One is through advertising, which is what almost 100% of podcasters do. The other is by charging for it, which is what audio book authors do. And so we actually started already something like one and a half years ago to let podcasters charge for their podcasts. So today on Spotify, if you sign up through Anchor, most people use advertising, either what is called DAI, standalone solution or our advertising solution. But they can also choose to charge consumers for the podcast. So you can subscribe to one host. And we're also allowing third parties, like Benton, for example, to charge for his podcast on Spotify. So in fact, we kind of already have two business models for audio, free and paid. And I think of audio books as it's the largest library of paid audio, right? And it's very clear that consumers want it. They actually pay a lot of money for it. So that was the reason, the main reason was that we want to be the world's audio platform. That's music and talk, but not just the half of talk that is free, but the paid part as well. The business model thing is important to me because I think different business models drive different content. There's a lot of content that could only exist under an advertising model because enough people won't pay for it. But there's also the opposite. There's content that really can't get created from a broad enough, but it could exist if you were allowed to charge for it. So we want to allow both basically. So how do you know when it's time to kind of do the next thing? So obviously, if we're thinking about the audio space, there's so much. There's Clubhouse, which some people may disagree, but seems pretty dead. You know, there's that kind of audio out there. There's in the kind of, I know from the journalism space, there's Substack where kind of writers are creating their own thing there. Like, how do you Spotify, sort of, measure yourselves and not go, we're going to do all of those things because we are the audio platform right now. Like, what kind of steps and checks do you have to make sure your roadmap is a sensible one? Yeah, it's a great question and a tricky one to answer because when you're a smaller company, you just don't have the capacity to do many things. So you have to make bigger bets. And that bet is no way to do something. The bigger bet is often to just stay out of things completely because it's too expensive to do. As you get bigger, you might have the luxury to be able to afford to look at something even though it's, you know, it's maybe 70% chance there's not going to work. It's almost like, you know, VC investing or something. You can get statistical about your bets. Now, I think that's important. If you look at these really big companies, as a product person, you could look at some of the world's biggest companies if you put up all the failures they did, all the apps they launched and never became anything versus the successes. It looks like really bad product management, but if you think about it from a VC portfolio point of view, it makes all the sense in the world. You just have to produce that many bets and if one of these pays off, it's completely worth it, but you can't think like that when you're a startup. Spotify is now semi-large, so we're starting to be able to do some of these things where it actually makes sense even if we're not sure whereas only like two years ago, that didn't work. It was too expensive. We had to be really, really sure. And so previously, we just took less risk. We were sure of what we were doing and the mistakes were very costly and quite expensive, as I mentioned. Now we're trying to force ourselves to not be quite as sure because it makes sense for us to be a little bit more statistical about it. And I think live is a good example. We've also investigated and are investing in trying live, but if it was five years ago, it would have been too expensive. We would have waited and say either it pans out and then we'll have to catch up or not. So that's one way to think about it. It changes with how big the company is. Is now the risk that you don't move fast enough? So like say Clubhouse had come along and we were all still using it. People would be saying, why the hell wasn't Spotify doing this at the start of the pandemic? Exactly, that is the risk. One way to think about it, if you're a big company, many of the decisions isn't really should we do this and what if it works? The decision framework that I try to use is if you look at it in retrospect, as you just said, let's fast forward three years now. And it turns out that live didn't work and we invested this much. What's the outcome? What's the damage? People would be like, that's stung a little, that was bad. Will it have hurt us much? No. The opposite. It turns out three years from now, it was everything and we sat on the sidelines. How is that going to feel? And then you realize that we should probably look at this. Even if we think it's only like a 30% chance that it actually works. And that's not how we thought before because you can't afford to think like that. So I think you're right. But we're not doing it for like optical reasons, what the press will think. It's more like, how will we feel about this area? And that's a big shift. It's a shift from a startup that makes one bet on one thing to sort of a VC that starts betting on more things. I still think the adjacency is very important. So with live audio, the adjacency is still very clear. I mean, for us, what we saw was that a lot of the podcasters on our platform that we really love and that we listen to, they were creating what looked very much like podcasts. So it's clear that this is not a separate space. It's clear that it was very close. That's why we decided to look at it and go into it. So I did a bit of crowd sourcing of questions before this from people. And what I really liked was Adele tweeted some feedback to Spotify recently that she didn't want her album shuffled, right? And within like a day or something that had been sorted out. How did the team manage that? What was the magic that happened there? And why did you bother listening to Adele? So I think that the truth, it's a great story. I think the truth is that as most product people here know, few companies build anything in a day. That launches to like 100%. So something's off there. And the truth is that we have been working on this feature for quite some time. The background is that because Spotify's free tier is shuffle based, according to our licenses, you have to listen in shuffle mode for consistency reasons. It was a shuffle default in premium as well. But we had gotten this feedback from a lot of artists and a lot of listeners that, well, I'm premium. I'd actually like to listen in order. And that made a lot of sense to us. So we had implemented it and we got lucky. We were just ready right when she tweeted. So it looked like it was a one day build. But as it often the case, it was already in the works for a long time. So when it comes to customer feedback, you must get a lot these days. How do you figure out what's the process? Who do you listen to? Who do you not? Adele, I guess, you turn that into a marketing stuff. It's a very good question. And obviously many companies, this is why you turn to statistical methods of getting user feedback. So we do what everyone else does. We analyze our metrics and we try to understand user behavior quantitatively as well as qualitatively through service and so forth to get as much supporting data as possible. But then I still think the old, I don't know if it was Steve Jobb who said it or someone else, there's a dodge that users don't necessarily know what they want. They say what they want, but you have to interpret what it actually is that they want. And people would ask for a faster car or a faster horse instead of a car and all of these things. I think that's true. So the way we implement that is you try to gather as much data as possible. But then you still have to create some sort of model of the world. You have to have some hypothesis for where the world is going. Maybe it's based on new technology that is coming along. Maybe it's based on new creators that are coming along, new types of content or behaviors. And then you try to build a model of that and see if it fits the data. But you have to have a model of what you have to be able to answer why we do this. You can't just do things because people say them because then the product isn't going to make any sense after a while. It's just going to come in the nominator. So I try to bet on certain people in the company that sort of almost have the entire product in their head. So like when I go from the library to this I'm like, how does this all make sense? And then, you know, we try to discuss that model and have as much of a shared model as possible. And then you take all this quantitative feedback and you can adapt that model. But I think it's really important that the whole thing makes sense. So you can't just divide and conquer and just build on demand. It's going to feel good for a little while and then the thing is going to get very complicated. So final question, 52 seconds left on the clock. What are some of your other bets? You're looking into live. What are some of the other ways you see this world of audio changing in three years? And okay, some of them aren't going to pan out quite how you expect. Well, I think, as I said, I think audiobooks and paid audio is a big bet. And in general, I think the trend you see that we certainly didn't come up with, which is a shift towards from just consumer focus to creator focus. And starting to serve creators and build for creators, which we have for some time now, but I think it's accelerating. And it's now it's not just building software, but it's also offering different business models for creators. And you see it in written media as well with sub-stack and so forth. So that's a big bet that we're doing as well. This is why we're offering different types of business models for different creators rather than a one size fits all. Previously, it was basically like music. You're going to have to accept this. It's one model and that's it. That's not how we're thinking about it now. So that's one big bet. And within that, we think audiobooks is a big bet. Another thing that we did recently that I think is interesting is back to focus. We've actually started recently allowing anyone who wants to add video to their podcast, which is also a little bit controversial internally. Like, aren't we an audio company? Isn't this defocus? But what we saw when we started adding some podcast with video, like Joe Rogan, for example, we saw that actually consumers, there is this type of background of video or audio with video that is very popular where consumers actually watch something like large percent of the video. They sneak in and out of it and look at it, but they listen mostly in the background. And because our company is about being the best background application for all these use cases, we decided to let anyone who wants also add video. And that I think is interesting because it lets the format evolve. So we've invested quite a lot in things like feedback loops and ratings and comments and so forth on podcast, but I think this combination of being able to talk about something in the audio and then show it and you can glance at it is interesting. Amazing. I think that's happened with TV for a long time that you have kind of in the background. So the sign is flashing at me, but thank you so much. We'll be able to come back to Slash in three years time and see how good your crystal ball was for the future of audio. We'll see you from back. Thank you.