Matthew Brennan has published:
The slides mentioned are below.
Related podcasts and articles:
- SMILE Marathon 2: Machine Learning / AI and Zero-Human Operations
- SMILE Marathon 4: Rate of Learning
- Learning Advantages
- Social Capital
Companies for this class:
- TikTok / Douyin
I write, speak and consult about how to win (and not lose) in digital strategy and transformation.
I am the founder of TechMoat Consulting, a boutique consulting firm that helps retailers, brands, and technology companies exploit digital change to grow faster, innovate better and build digital moats. Get in touch here.
My book series Moats and Marathons is one-of-a-kind framework for building and measuring competitive advantages in digital businesses.
Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.
Welcome, welcome everybody. My name is Jeff Towson and this is the Tech Strategy Podcast. And the topic for today, why the bite dance attention factory is a threat to Facebook. Now that phrase, the attention factory, that’s a, it’s a new book that’s just come out by a buddy of mine named Matthew Brennan, who I think some of you are familiar with. He lives out in Chengdu. He writes about 10 cent a lot. knows what he’s talking about. Anyway, he has a new book that just came out about ByteDance which is called The Attention Factory. And that’s, I think that’s a really apt description and I want to kind of just talk about what it is this company is really doing and why it’s a threat to Facebook in particular. And I have put the link to the book down below which you can get at Amazon. It’s a very very good, definitely worth a read. Now for those of you who are subscribers, the Ideas here that I kind of want you to focus on are really rate of learning, which I’ve brought up in a couple contexts. In one, it’s in the the Smile Marathon, this SMILE, you know, one of the ways you compete on an ongoing basis. Your Endless Operational Marathon, one of those is rate of learning. The L in Smile is learning. So that’s kind of important in the media space in general, but at I’ve also mentioned rate of learning, speed of learning, as a competitive advantage, which I think it definitely can become at a certain point. And this is an old idea, going all the way back to Henry Ford and building Model Ts and realizing, you know, the more Model Ts you build cumulatively, the cheaper you get at doing it. Turns out people can learn, organizations can learn, and if one of the ways that can play out is in cost structure per unit, costs can decrease with cumulative experience, volume, and other types of learning. So this one really plays at both levels, and that’s kind of an important idea. And the email I sent out a couple days ago raised the question of prediction. Now one of the ideas for today is rate of learning. both as an operational dimension and as a competitive advantage. The other one is sort of machine learning and AI, which is basically cheap prediction, which I’ve talked about before. I’ve also sort of said, that’s another dimension of competition, the smile marathon. The M stands for machine learning. And I raised the question, you know, can machine learning, which is prediction, cheap prediction, automated prediction, can that also rise to the level? of a competitive advantage as well, like rate of learning has. And I’m gonna send you an email in a couple days. I’m still sort of working out my thinking on that one. So that’s the other idea for today. And this all goes on, so basically two learning goals for today, rate of learning and machine learning AI as an operational competitive dimension. This all goes under learning goal number 32, which is adaptation, innovation and resilience as competitive strategy. Okay, let’s get into the topic. Now I spoke about TikTok before, which was in podcast 20, which was what is TikTok slash Douyin’s secret. Douyin is the Chinese version of TikTok. TikTok is the international version of Douyin, all made by ByteDance. And I’ll give you a little quick summary of that, but that’s in podcast 20, and the link is in the show notes. But I kind of teed up a couple ideas there where I said, look, what’s going on with this company? Just the TikTok part, not their other apps, which I’m going to talk about today. And, you know, my standard approach to this stuff is, look, there’s so many factors you can look at, you know, so many dimensions we can look at, you know, price, we can look at cost, we can look at user experience, we can look at using money, we can look at, I mean, there’s a million factors any time you look at any company. What I always try and bring it down to where the rubber hits the road for me is I always take the consumer view, then I take the competitor view. And the consumer view is, you know, all business at the most granular level comes down to a decision, which is I walk into the store and how do I choose to buy this pair of shoes or this pair of shoes or not buy, right? That’s when it all plays out is in that one moment. So I always look in this case at the consumer view. What is the consumer experience? How do they come to make that decision? I’m gonna watch this or I’m not gonna watch this. That’s where the rubber hit reached the road. So I always try and look at it from the consumer view. And then I also take a look at it from the competitor view. Okay, this person is already, here’s a shoe company, Allbirds or whatever, and people are buying it. If I am a well-run, well-resourced competitor, could I take part of that business from them? assuming I’m well run and I have some skills and all that, or is there something going on there where it’s just too difficult? I can’t do what they do, or I can do it, but I can’t do it as well as they can do it, or I can’t do it at the same price point they can do it. That’s the other kind of rubber meets the road moment, the competitor view. Could I do it? How protected is this from a company like me or anybody else? Now, when I looked at TikTok before, I sort of said, look, the consumer view, You know, the consumer view, in this case, it would be the viewer. There’s actually another party, which is the content creator and the advertiser, because this is a platform business model, but hold that thought for now. The consumer view, the viewer, would be, okay, I thought it was a very addictive product, in particular. I thought short videos were particularly popular. I thought it was highly personalized, and it turns out, highly personalized short videos are addictive. much more than other types of media and that’s just the way some things are in life. If you’re trying to sell carrots in the supermarket it’s actually quite hard. If you’re trying to sell potato chips it’s easier because it turns out they got a lot of salt and stuff like that. It’s a more addictive engaging attractive product than selling bread. Okay highly personalized 15 second videos on your smartphone turns out That’s closer to potato chips than carrots as a product. Other part of that, I thought they had a very, very easy to use interface. I think they’re very good at the UI, the user experience. You know, you flip the screen, you can react to it. They focus, and they’re not just doing short videos, they’re doing things with music, dancing, lip syncing, sing-offs. I mean, there’s a reason those TV shows are very popular. You know. singing competitions, very popular. So very addictive, very attractive sort of content they were selling, very nice user interface. That was a pretty good product from the consumer side. Okay, then you take the competitor view, what’s going on? Do they have any competitive advantages? I argued, I’m sort of recapping podcast 20. I argued, you know, this is an audience builder platform. I’ve identified sort of five types of platforms I think are very common. There’s others, but the five I think you see commonly, one of them is an audience builder, platform business model, simplified type of an ecosystem strategy. Audience builder has two user groups at least. User group one, the viewer, user group two, the content creator, you’re in the business of connecting them. People upload videos to YouTube, other people watch videos on YouTube. Turns out there’s a third user group on these platforms, which is advertisers. That’s another group. There’s actually more to it because now it’s becoming commerce and stuff, but there’s usually two. Okay, audience builder platform. The difference there between them and say, Facebook is it’s not a social network. You know, you don’t, on Facebook you see content. So Facebook is mostly an audience builder at this point. They also do messenger. There are a couple of things. I’ll talk about them in a sec, but. You know, you follow certain people and based on what your connections post, that’s what you might see. Now TikTok doesn’t do that. You don’t follow anybody. It just tees you up content that it would think you want to see. Twitter’s kind of halfway in between. Twitter shows you stuff, but it basically shows you people you choose to follow. So you’re choosing who you’re following and that’s who you see in your feed. Okay. That’s a halfway between, say, Facebook where you have to choose each other and it’s a connection. Well, TikTok is the other extreme. There’s no following, there’s no social graph. It just tees up one video after the next, regardless of who makes it, and that dramatically expands the pool of content they can show you. A lot of high frequency usage, a lot of users fast, and they’re also particularly powerful in converting their viewers into content creators. All right, and then they have a network effect, which is sort of, okay, fine. They’ve got a decent network effect going, similar to YouTube. The more people that create content, the more valuable this is as a platform to watch videos on because more content is better than less content. Okay, TikTok’s kind of like that. That was my basic take on what they’re doing. It’s just a very clever type of audience builder platform with a sort of new type of media format, short video. and they did it very, very well. Okay. As we bump from TikTok up to ByteDance, the parent company that built TikTok, which is the point of today’s podcast, you realize this isn’t the only app they’ve made. It’s not like they made one app and this one happened to be very successful, good for you. They have made tens, hundreds of apps. I mean, that’s hence the attention factory. They are just a factory for turning out apps. And this is their most successful, but they have a couple other super successful ones, and then they have like 31 others right now that they’ve got going, and they’ve tested another, who knows how many to find these. So they are constantly churning out apps, and then they, whatever is the most successful, they push those hard, they put a lot of money behind them, and stuff like that. So ByteDance as a parent company is an app factory or an attention factory. And then the most popular thing to come off their assembly line so far has been TikTok. Actually Douyin and TikTok, which are kind of different. Douyin is the Chinese version. It’s actually a lot better. It’s a little more sophisticated. The content creators have been doing it longer, so the content tends to be better. Like TikTok in Thailand is not terribly great. It’s okay, but Douyin in China is amazing. It’s something to watch. Now that’s the basic TikTok story I teed up. Now I wanna sort of take you up a level and talk about ByteDance. In the book, there’s a lot of history of ByteDance and how long they’ve been around. And it’s a really cool story because they’re a very well-run company, very aggressive. But… I want to sort of take the same framework. Okay, if they’re making lots of these products, these apps that all have the same goal, which is to get your attention. When Alibaba makes a product at Taobao, they generally want you to spend money. It’s a commerce driven model. Everything TikTok, ByteDance builds is based on getting your attention, which they then monetize by advertising. Okay. Across all the apps they have, we could look at these same viewpoints, the consumer view. the content creator view, the advertiser view, those would be the three user groups, because you gotta convince all of them to participate in what is mostly audience builder platforms. And then we would have the competitor view. Now, when it comes to the consumer view, we look back over the last two years in China, and ByteDance has basically taken a big bite, I guess, out of Tencent. I mean, that’s where the numbers moved. Go back down, this is from Matthew’s book, from 2017 June, how do Chinese consumers spend their time in online, well, 54% of their time is spent within 10 cent, 3.9% of their time is spent with bite dance. And then there’s Baidu, Alibaba, Sina and others. Go one year later, 10 cent drops about six to 7%, bite dance goes up the exact same amount. So 2018. ByteDance is up to 10% of people’s time, and it keeps growing. And the others, Alibaba took a little hit. I’m sorry, not Alibaba, Baidu took a little hit, Alibaba was pretty stable. But really, that’s who ByteDance is taking time from, is its 10 cent. So they’re a problem for them in China. And if you think about what a hard game that is, how fast technology changes, how fast fashion and fads change. You know, doing this sort of game, especially if you’re entertainment focused, how do you keep people’s time year after year? Maybe they’re listening to hip hop, but next year it’s going to be pop is going to be popular. And then maybe heavy metal will be popular. And, you know, fads change all the time and technology changes and types of media change. Short video is a very popular thing right now. It wasn’t five years ago and it may not be in five years from now. Maybe we’ll all be doing VR and AR in five years. Maybe we’ll be doing live streaming. I don’t know. I mean, all of this is changing very, very quickly. So it’s a hard game to play, especially in the media and entertainment space. So that’s user group number one for ByteDance, but they always play the same game as far as I can tell. We moved to their second user group on their audience builder platforms, and that’s advertisers. They’re monetizing in one simple way that makes it nice and easy to study. And in this case, this is also from Matthew’s book, short video as a category, you know, where do the digital marketing dollars go? About 17% in the last year goes to short videos. That’s where advertising dollars are going. 15% goes to search engine, 10% goes to social media, 39% goes to e-commerce. So it’s pretty much the second biggest category of usage for ad dollars is in China is short video. And you go back two years and that number was about 4%. So short video has really sort of eaten into the ad budget spending of China. And that’s mostly come at the expense of search engines. And that’s Baidu. So if they’re taking time and attention in China, if they’re taking time and attention from 10 cent they’re taking ad dollars from by do and it’s been pretty brutal for by do By do okay user group number three the content creators now. This is actually where bite dance is really interesting Because you know I’ve sort of said okay, you’re an audience builder platform audience builders have some interesting dynamics on the content creation side you can You know if you’re doing Spotify you have to get the popular licensed music from the big music labels, because people don’t want to hear my little hit song that I just wrote today. No, they want to hear the famous artist going back in time. So you have to get sort of professionally generated content. And if you go to say Facebook, they’re kind of a mix. There’s a lot of user generated content, which is, hey, here’s what I had for lunch. Hey, here’s a picture of my cat. And then right after that in the newsfeed might be, here’s a New York Times or Wall Street Journal article. And they to a large degree have commoditized professionally created content, PGC, which has been brutal for the newspapers and the content creators, because they kind of made them just like everything else. So there’s a bit of a spectrum on the content side. And what TikTok has been very good at and what ByteDance has been very good at is, just using a massive spectrum of content. So TikTok is, you know, you upload little 15 second videos. That is pretty powerful, easy to do. Anyone can do it. Before that, they had Totiao, which was news headlines. So they would basically have their, you know, software’s hunting the web and searching for headlines and what people are reading. And then they would copy and paste them and put them in there. So that was another type of sort of news aggregation. This one is more short video aggregation and now they’re doing music streaming. They’re doing long form videos, they’re working on that. They have memes, they have jokes. They actually have quite a few different types of content that they create and then tee up to users under various apps. It just so happens that short video was the really popular one, but before that. the headline one was very popular as well. So there’s a lot of interesting stuff there. And this tees up another concept for today, which is the idea of social capital. And there’s a writer, I think he’s in Silicon Valley. I’m not sure, he’s in the US, guy named Eugene Wei, who writes really interesting articles about social capital. And why do people post on Facebook? Why do they post on Twitter? Why do they put up videos on YouTube? Why do they put up videos on TikTok? The content creators, what is their motivation? Whether, and it could be a professional, not a professional, say someone who’s actually making videos for YouTube, or just someone who’s using Facebook and is just reposting stuff, or putting little comments and things like, well, why do people do that? And… The argument from Eugene would be, we are status seeking monkeys, that we are highly motivated to seek status. And what these platforms give people the ability to do is to attain social capital, to become well known, to get recognition, to get likes and to get followers. And all of these sites give you very immediate feedback of how successful you are, how many people watched you. And social capital, like regular capital, can be very valuable. If you get a ton of followers, you can monetize that. You can get advertising deals. You can become well-known. You can sell books. You can do a lot of stuff. And so one of the questions is, okay, if you’re a content creator on any of these sites, Facebook, Twitter, TikTok, YouTube, how easy is it to attain social capital status? And the interesting thing about social capital is if everyone gets it, it’s not valuable. It’s kind of a it’s about exclusivity to some degree. So if it’s something everybody can do, like, hey, I’m going to post a little something on Twitter because it’s so easy to do the return on effort, the return on investment, how much time did it take? How much social capital did I get for my little cleverly thought out tweet is actually quite Now, because everyone can do it. Now what if I make a YouTube video? Well, and I make a YouTube video of me, I don’t know, talking about something, traveling. That takes more effort, but it also requires some level of skill. And it’s by having some level of skill that I can distinguish myself. I’m just not like every other person tweeting. I’m actually, you know, a good magician, or I’m a good juggler, or I’m a good singer, or I’m a good dancer. or if you’re on Instagram, I’m really attractive. There is that sort of distinction by skill level that is what creates status and social capital. So you don’t want people posting stuff if everyone can do it, because there’s no way to rise from to the top. But you don’t want a skill or an activity that’s so difficult that most people can’t do it. And that’s a little bit of the YouTube problem, is it actually takes some time and effort. and you gotta have a studio and cameras to post on YouTube. So that one’s a little bit too hard. Well, it turns out TikTok, ByteDance, and a lot of their products are perfectly suited to that. That making a 15 second short video is actually not that hard, all you need is your smartphone. Okay, but what skill are you gonna show that most people can do, but if you’re… good, you can actually stand out. Well, lip syncing, singing, dancing, most people can do that stuff. And if you’re clever and fun, you can, so you can stand out. So it’s right sort of threading the needle where there is a skill to be shown such that you can stand out and get some status and get some social capital and get followers. And suddenly you got a big following on TikTok, but it’s not so esoteric. or difficult that most people can’t do it. And who really cares about social capital? Who’s super obsessed with achieving status and recognition in this world? Teenagers. They are crazy about status and getting followers and being known in a way that older people really aren’t as much. Now you could argue we’re all status-seeking monkeys. But when you get older in life, there’s other ways you can demonstrate status. You can have a nice car, you can have a nice job, you can have some money, you can have some nice clothes. If you’re 15, you don’t have any of those things. But if you can get some followers on TikTok, okay. So there’s kind of this interesting idea buried within the motivation of content creators and why teenagers in particular are super enthusiastic about posting videos on… a lot of the products. Anyway, so I think they hit that quite well. But the framework that Eugene presents is when you look at say social networks or things like that, usually there’s three axes. There’s achieving status, social capital. There’s a utility and there’s entertainment. And that’s what motivates people. Some things are more about a utility. You know, it’s a useful thing. A lot of WeChat is like that. It’s hard to get status on WeChat, but there’s a lot of useful utilities. YouTube is a lot more about entertainment. It’s hard for most people actually to make high quality videos. And then you could say Facebook and TikTok are mostly about status, social capital. Anyways, I think that’s an interesting aspect of this to think about. And that’s sort of the three user groups. And the thing I think that jumps out at you, is in a lot of bite dances stuff, they really nail all three of those in a very clever way. I’m gonna go through some of their actual apps. So I just want to sort of tee up the frameworks. And within those ideas, what is Facebook? Have you ever, I’m not really convinced there is such a thing as a social network. And Facebook is the social network. Hey, you know, the network, the platform business model is. created by people on their smartphones all connecting. So if you were to chart out the network, the nodes would be people. And then the platform type would mostly be an audience builder where people would behave either as a user, I’m sorry, a viewer or a content creator. You can be the same person in a network and be both types of users. Because you can take on different roles. I could be reading stories and then I could create stories. So even though it’s the same person, those user groups can actually be different roles as opposed to different people. You could say a lot of what Facebook is doing is an audience builder now. It’s a lot of people posting content and then other people view that content and everyone’s trying to get feedback and likes and the newsfeed would be the key there. But then they also have WhatsApp and that’s… basically a messenger tool, which would be another type of platform closer to, say, Zoom than an audience builder. But they also kind of have this idea of a social network. I won’t even get into Instagram. That’s different. They have this idea of a social network. What is that? I tend to think that’s not a real thing. I think this is sort of just one of these words that started to get used when we didn’t have the language to describe what was going on. I think they’re mostly an audience builder. and I think they’re a communications platform. I think that’s what they’re doing. But I’ve always had them as sort of a strange animal. But if you think about the user groups they’re trying to attract, it’s effectively the same ones that ByteDance goes after. It’s people who are viewing stuff, and it’s people who are creating content, and it’s advertisers. It’s the same three. Okay. Now within all of this, there’s sort of two, I think, big… problems and then I’m going to explain why I think ByteDance solved them and Facebook has not. But there’s two major problems and the first problem is there is a growing mismatch gap between the supply of information and content and media in the world and the demand for that. information and that’s basically your smartphone screen. That’s the attention economy. That’s the fact that look there’s only so many people in the world they’re only going to consume content of various types for so many minutes per day. That number is pretty fixed and the supply of content, media, short video, messages, photos, all of that just keeps exploding. And this is this is not just limited to entertainment, we see this in every aspect of sort of the digital age that there’s so many webpages with information, how do we find that information? Well, that’s the search engine, that was a solution to that problem. If you wanna find a video to watch, how do you find it? Because there’s so many videos, that gap between supply and demand, that matching function, that… That’s a problem in itself and the supply just keeps exploding. But the number of the hours in the day doesn’t change. And this is something in the book that I think Matthew really explained well, that Jiang Yimin, the founder of ByteDance, at least from the book, has been focused on this particular question for most of his career, that how do you match the world’s information with people’s attention? And that’s really what Alan Zhang of WeChat talks about as well, and that’s kind of what Mark Zuckerberg is doing as well. And that problem is just getting more and more. And it’s not just that the quantity is increasing, but what about the quality? I mean, there’s great quality content and there’s bad quality content. How do I know which videos I’m going to see? I can’t watch them all. I can’t search them all. It’s a real problem, that mismatch. And… The other way to think about this question would be, who controls the information flows? And I’ve talked about this before that, we are increasingly shaped in our thinking and our feelings by the information coming through our phones, as opposed to the information we access in our daily lives, walking around the street, seeing things, talking to people, traveling. It’s the info through the phones that seems to be the most powerful. And it’s really shaping how people live their lives and what their experiences are and what they think. So who controls that? Does anyone control it? Do you control it? How much information do you consume through your phone that you actually searched and said, I wanna know X, versus how much was just passively pushed to you through notifications in a newsfeed? And the answer is the vast majority is being pushed to you and you’re just consuming it passively without choosing what you want to read or see. Well, who’s deciding what is pushed to you? Is it people? Is it algorithms? Well, who’s writing the algorithm? Is it reviews? Is it being gamed? You know, that inflammation flow is a really important question. So, ByteDance has been focused on this question from day one. The other problem was the one I teed up before. Okay, even within this, if I’m a particular company in this attention game, How do I keep people’s attention? I might get it now, I might make some cool videos and everyone thinks I’m cool, but are they gonna watch me next year? If you’re a business, you have to keep people’s attention year after year. It seems like that is getting harder and harder as well, especially as the supply keeps changing and people are changing. So against those two problems, we see bite dance emerge as the attention factory. And that’s, I like this phrase. I actually thought that phrase up. Just to pat myself on the back. Matthew, as we were talking about the title, I said use attention factory. But it has two ideas there. One, it’s this idea of how do you capture people’s attention. And two, how do you do it as a factory approach where look, your product that’s popular this year may not be popular next year. So how do you do it as a factory approach where you’re just constantly churning things out? And. ByteDance is absolutely doing that. When is the, how many apps has Facebook created in the last 10 years? They bought Instagram, they bought WhatsApp, they tried to make a copy of TikTok with Reels, which stinks, by the way. When’s the last time they created anything that’s new? And then you compare that to say, ByteDance, one of the things that jumps out when you read the history is they were creating apps all the time. They were creating an app every couple weeks in the first years of their founding and then seeing what worked. So even if you look now how many apps, I’m counting them up. Within China, they have about 19 apps running right now. Internationally, they have about 13. I’ll give you some of the names. Like within China, they have ToTiao, which is a news aggregator that was launched in 2012. They have Douyin. short video that was launched in 2016. They have Shiguaw Video launched in 2017, that’s short video. They have CapCut, which is a video editing app launched in 2019. Duoshan, which is a messaging app launched in 2019. Peepee Shah, which is memes and joke videos launched in 2018. Feishu Enterprise app, 2019. Go Go Kid, that’s an education app. Launch 2018, Dongqidi, something automotive, I don’t even know what that one is. Wukong, Feixiu, Feiliao, Toteao Surge, Qingbei School, Guagua, Studywell, I mean, most everything I just read you was 2017, 2018, and that’s just their China ones. If I go international. 2017 they launched TikTok, 2020 CapCut, 2019 Lark, that’s enterprise stuff, Daily Hunt, FaceU, Reso, You Like, Babe. And then they have a whole long list of discontinued apps because for all the ones they’ve launched, they had a whole bunch more that they just didn’t see the traction happening and they killed them off. So they are constantly shooting out apps and seeing what works. That’s their factory. And what is the goal of all of these apps? To get your attention. And they measure it. So their first app was not Totiao and it wasn’t TikTok, Doughin. It was a meme app that they had this little funny thing where you could post comics, gifts, cute videos, and just little cartoons. And it was kind of, I’d say it was click bait, but it was pretty much just trashy. mindless clickbait and you know that’s potato chips and people liked it and they would just sort of systematically study what hit and what didn’t as an app and then within each app they would see what types of videos were hitting and what types weren’t and the ones that would get attention they would elevate and not and it included all these types of content video, short video, music, lip syncs, text. photos. So they were learning to sort of, and some of them were user generated like short videos, and others they were just scouring the web like articles into TOTYAU. They were just trying to bridge this gap between the world’s content and what people wanted to see. So I’m looking at sort of you know what type of content they would have. Pictures, videos, articles, music. The topics within this content might be home design, education, fun, beauty, knowledge, news, and it would either be user generated or they just have their back end systems crawling the web and putting it all together. And then they would see what hits. And it was out of that that they discovered that short videos were super popular. And not just that short videos were popular, but dancing, lip syncing, that was particularly popular. I mean, TikTok was not an accident. That was a very systematic, scientific approach to studying what gets people’s attention. One of the reasons this is so powerful is okay, so let’s say you’re doing this systematically. You’re testing things at the app level, but they’re actually doing more than that because within each app, what do you have? Within TikTok, you have all these user-generated videos within TikTok. So at that level, they’re also testing what works. So they’re experimenting at the app level, but then within the app level, they’re seeing what specific types of videos are popular today versus a month ago versus next month. So they’re experimenting at the individual content video level as well as at the app level all the time. And people are generating more and more content for them. They keep doing this. And out of this, they start to develop basically a social or a user interest graph where they start to map out all of their users what they like to see. So not just what’s popular but what’s popular to you. Not just what short videos you like to watch but what news articles you like because if you’re using multiple of those apps even if the app doesn’t do real well and maybe they cancel it. your usage of that app will populate a user graph, an interest graph for you, which they will then port over to their video app. So they’re generating all this useful data about what each individual person likes so they can start to tailor all types of content to you. So that was kind of the first thing they did, I think, that Facebook did not do. I mean, Facebook and ByteDance are… struggling with the same problem, this matching of information between a sea of information and what people want. And Facebook, I mean, ByteDance has solved this, I think, with two things. The first is this factory approach to apps and content within apps, which Facebook is not doing. They’re not rolling out apps every two weeks. And the second thing they’re doing is, this is the part I think Matthew really did great on in his book. He basically lays out this argument that this mismatch, this difficulty of matching people with the world’s information, we have seen one, two, three, four, five different attempts to solve this over the last 20 years. And ByteDance has really pioneered a new version of this. So, you know, it used to be in 1999, the websites that were most popular were portals. They were Yahoo. and what was Yahoo doing? It was curating content based manually with people based on what you wanted to see. So if you would go on Yahoo in 2000, 1999, there would be a sports tab, there’d be a news tab, there’d be a politics tab, and they would have chosen articles from around the web to show you. So that’s how they’re bridging this gap for information flow. That was sort of attempt number one, which was manual. And it was also sort of a push mechanism. You didn’t choose what to see on Yahoo. You went to Yahoo and you pulled open the sports page and they showed you 10 articles they had decided. So it was push and it was curated by people. All right, after that, 1995, 1997, after a new mechanism emerges to do this, which was the search engine. Starts with something like Ask Jeeves and… Vista and then Google comes along eventually and Google basically Tries to solve the same problem. How do you access the world’s information? Well, you use a search engine But it’s a different type of mechanism because it’s not a push mechanism I don’t sit on Google and things just get teed up to me all the time I have to actually manually search for something. I want to know X So it’s a push mechanism, it’s an active intent mechanism, which is why their advertisements are so valuable because they show active intent by people what they wanna see. And it was done by technology. So they didn’t do it manually like Yahoo, they did it with algorithms and data search and web crawlers. So sort of active intent, technological solution. Fine, they do very, very well. But still, you don’t actually search for that many things in life information-wise. You sort of passively consume most. So then we come to another model, which would be subscriptions, where you sign up for someone’s email list, you sign up for their RSS feed. So you’re choosing, there’s an active intent to it. I wanna follow Jeff’s podcast. So there was an active intent, you chose something, and then based on that, then the information just starts to come to you passively. So that’s again, another solution to this problem. And then Mark Zuckerberg comes in with the idea of a social network, and the way you can think about a social network is a crude solution. It’s almost like a hack to this information problem. Okay. I don’t want you to have to search for everything, i.e. search engines. I’m going to push the content to you based on what I think you want to see and you’re just going to stare at the newsfeed and consume it. But how do I know what you want to see? I can’t curate it manually. Well, I use a proxy, which is I kind of make the assumption that I think you would be more interested in information put out by your friends. So I’m gonna have you connect with your friends and then whatever your friends post, I’m gonna show that to you. Because I think that’s sort of a crude hack to show you what you really want just based on what your friends post. And that’s kind of what the social network was. It’s automated, it’s a push model where you sit there and just passively stare at the newsfeed while it tees up one little post after another. But the content is kind of crudely assembled from your social network. and then they try and come up with your social graph and things like that. Okay, not awesome, but functional. So what ByteDance comes up with, and YouTube basically came up with the same solution, is we wanna do an automated push model where you stare at a newsfeed, but it’s not just coming from your friends because that’s a crude approximation for what you actually are interested in. We’re gonna use machine learning to accurately describe and choose what you wanna see. And we’re gonna be very specific. And YouTube started doing this in 2012 or so, where they started to replace the system where you choose YouTubers you wanna follow. I’ll subscribe to you, I’ll subscribe to you. And instead, they just start recommending video after video based on basically algorithms. And when they did that, the usage went, you know, started to really surge. That’s kind of, that’s basically what ByteDance is doing. When you log into ByteDance, any of their products, TikTok, Totiao, whatever, they never ask you what you like. And they never ask you who your friends are. It’s not a social network at all. As soon as you start watching their algorithms start to understand at a very detailed level, what you like, and they start to mechanically, algorithmically, search all the world’s content, not just your friend’s content, every short video that’s been posted by everybody to pick what you want most. And that’s a pretty powerful solution to this whole information problem. And that’s kind of their strength, machine learning. So I’ll put in two charts in the show notes that basically show this has two sort of powerful implications. One, we shift from a pull model where someone goes online, goes to a search engine, tells them what they want, the search engine searches the world’s information and brings it back to them, to a push model where the information, you just turn it on and you blankly stare at the screen and the algorithm starts to push to you what it thinks you want and based on what you do, it then adjusts its behavior, which is how most information is. consumed. And when you look at content, you know, people talk about the long tail. I’ll put the chart. These are Matthew’s charts. I’ll put a, you know, chart in there. There’s a small percentage of content, let’s say short videos, that everybody really likes. So the most popular 1% of content, people generally are, they can access that by a subscription model. They follow the most popular YouTubers. and they choose them and they subscribe to them, and that’s how they get the 1%. But you’re not subscribing to a thousand people. The social network and search, let’s say that gets you the next 20 to 30% of content that you search for something and it pulls from that or it just pulls from your friends. But then there’s another say 70% of content out there, random little tiny short videos made by farmers in India. and factory workers in Indonesia. This big, huge long tail of content is mostly not accessed by a social network or a search approach. But this algorithmic push model that ByteDance does, it pushes the long tail. It finds the funkiest little content from around the world that is hyper-specific and hyper-niche in what… what it is, but also in terms of what you specifically are interested. So it hyper-personalizes and hyper-customizes everything just to you, because it can use this algorithmic push model to do that. And that’s a lot of the strength of something like TikTok, is they’re very good at pushing the long-tail content. That’s why it feels so addictive. because it is showing you at an incredibly granular level stuff that just you wanna see. I think that’s where the addictive feeling comes in from that you get when you watch TikTok. And I think the user interface is quite cool. And that’s kind of, I think the most important part of this talk is to think about bridging that gap between information and people’s attention and that what ByteDance has really done is pushed a newer way of doing that. That’s why 50% of their staff are in R&D. 50% of their people, just as far as I can tell, are engineers and who seem to only do algorithms because they’re trying to solve that problem. And when they solve that problem with their algorithms, it doesn’t just work on videos, it works on music, it works on education, it works on text, it works on headlines. It can be applied to a lot of things, which is why this company is maybe incredibly valuable. But when, you know, so when you hear about ByteDance, the attention factor. I think there’s three important ideas there. One, that they have a factory approach to making apps. They’re making apps all the time. That’s critical if you’re gonna keep people’s attention over the long term. Number two, within the apps, they’re also personalizing the content and they’re testing what works. These types of short videos versus not these types of short videos. And then number three, they have this sort of algorithmic push model. that is changing the way you consume information and that’s how they hold your attention. That’s why I think they’re a big threat to Facebook because I don’t think Facebook is able to replicate that. All right, let me pull this back to our frameworks now, but that’s a pretty important idea, I think. Now I’ve sort of talked to you about competition at three levels. Level one, well, say just highest level. is competitive advantage, having a structural advantage that’s very hard to replicate. Below that is this idea of just sort of an ongoing operational marathon, where like it does, even if you don’t have a structural advantage, most restaurants don’t, you just gotta do stuff every day, you gotta run, run, run. And I call that the smile marathon, and depending what business you’re in, you’re gonna focus on one or more of those dimensions, SMILE. Underneath that there’s probably another level which is just short-term moves and dirty tricks. You know, you can compete with another company by throwing money at the problem for a while. You can behave irrationally, you can do dirty tricks, you can do hacks, growth hacks. Those are tend to be short-term moves. You generally have to be aware how you compete on all three levels. Now, what I like about what TikTok and ByteDance is doing is within those levels, I’ve talked about rate of learning. You know, this is one of the key concepts. And I’ve described it as both a competitive advantage and as an operational dimension to compete upon. So, let’s say as an operational dimension, S-M-I-L-E, smile. The L stands for rate of learning. That in certain businesses, you know, if you can learn faster, if you get cumulative learning, that makes you cheaper. The Henry Ford Model T example. But there are other types of business where that can give you a real advantage, like McKinsey, IBM, Goldman. These companies are very well adapted to learn things very, very quickly. So let’s say there’s type one rate of learning, which we look, we do more of something, we get cheaper at it, fine. Type two of learning, I’ve used these, if you go back to the rate of learning lectures I’ve talked about this type two rate of learning, is our rate of learning is not that we’re doing one thing better and better, it’s that we’re doing new products all the time. And Steve Jobs was a type two learning guy. He comes up with the iPod. But then the next year he comes up with the iPad and then the iPhone and then the TV. I mean, he was always launching new products. That’s a type of learning when you get good at product development like that based on sort of the market. Now, it looks to me like ByteDance is doing both of those types of things. And there’s a third type, which we call sort of, this is a Boston Consulting Group thinking, they call it the Bionic Company. this idea of a self-tuning organization where certain parts of the organization are always learning. When you go on Netflix and you watch stuff, no human being is involved. The software just knows what you’re watching and they self-tune the content to show you more of what you like. So it’s kind of learning at an algorithmic level. So there’s at least three types of learning here, traditional learning, i.e. you’re becoming cheaper, type two, which is new product innovation, new launches, and type three is sort of… algorithmic, human plus machine, bionic learning. Okay, ByteDance is doing all three of those. I think they are actively studying on a minute by minute basis what gets people’s attention at the individual video level, at the app level. I think they’re very, very good at this and that’s a huge part of their strength. So I would say that one of their operational dimensions is rate of learning. Okay. I’ve also kind of said rate of learning can be a competitive advantage. That at a certain point when you are learning so much faster as an organization than others, like the Model T Ford factory, at a certain point you are actually cheaper than them on a per unit basis. That is a competitive advantage. Big factories can be cheaper than smaller ones in terms of volume. because they get hyper efficient at doing things. So that rate of learning as a competitive dimension becomes a structural competitive advantage and you can measure it. This one’s cheaper per unit than that one and this one can’t match them. A lot of times this gets baked into the idea of culture where certain companies are just faster or cheaper than other companies. And it has to do with. sort of rate of learning is a competitive barrier and you just simply can’t match what they do and people try and they can’t do it. So I’ve talked about rate of learning at both of those levels over the past year. Now, if I were to look at what ByteDance is doing, I think they’re doing both of those, but I also think they’re doing machine learning. That’s their other operational dimension, the SMILE marathon, S-M-I-L-E, the M stands for machine learning. I think this is where the vast majority of their staff are. I think they’re building these algorithms that connect information with viewers and maximize attention. So they’re both learning and then they’re adapting and building stuff based on that. That’s machine learning. Okay, and I’ve said machine learning is basically cheap prediction. So they’re getting very good at predicting what video you most wanna see next. And one of the questions, oh, I think they’re doing both of those as their operational dimension, rate of learning and machine learning. Those are their two main operational dimensions. But if rate of learning can become a competitive advantage at a certain point, can prediction also become a competitive advantage? And that’s the question I’m working on right now. And that’s what I teed up to the subscribers in the email a couple of days ago. At what point are you so much better at prediction that others can’t… copy it anymore. They can’t do it. So your video app is always more enjoyable and addictive than the other ones. Can it reach that level? I’m not totally sure. I’m working on that right now. But if rate of learning can become a competitive advantage, maybe prediction can also become a structural advantage. But either way, I think ByteDance is playing across the board. very very well. They’re very good at the operational marathon aspect, they’re very good at the competitive advantage level, and they’re actually good at the dirty trick stuff, because if someone like Facebook tries to throw money at them, which Facebook likes to do, they like to use money as a weapon, which they did against Instagram and Snap, you can’t do that on fight dance because they have deep pockets. So they’re kind of they’re playing on all three levels quite well, which I think they’re kind of a threat. Okay, so that’s the last question for today. Can prediction be a competitive advantage? And prediction, cheap prediction, that’s machine learning. I don’t know, maybe. I’m gonna put in some of these, well, Matthew in his book, he put in some charts about why he thinks. ByteDance is such a good structure. One idea would be because they roll out so many apps that they get a lot of data from all of those apps, that data then goes into the user profiles. So even if you’re using TikTok, they’ll come up with a new app next month that will also be strangely attractive to you because they already know what you like to watch on TikTok. So they’ll use this sort of viewer information. to feed into every new app they do. So in theory, each new app could be getting smarter than the previous apps. And their algorithms could be getting better and better. So maybe their apps from two years ago, say Totiao, were kind of addictive, but maybe the apps this year are much more addictive because the algorithms are better. So the way I’ve been thinking about this is, do they have, the phrase I’ve been using in my head is, do they have economies of scope? in machine learning and user data. There’s economies of scale, economies of scope. Economies of scope is something like Procter and Gamble, Mars, Snickers, where they have a lot of power in distribution and marketing, and then they apply that to lots of products like this candy bar, that candy bar. You know, they can apply that scale to lots of products. Do they have the same economies of scope in machine learning and data? And they’re increasingly applying that. to product after app after app after app in this case. Maybe the fact that 50% of their people work in R&D and I think the vast majority of them are doing algorithms. I’m not sure there’s any other company that’s going after your attention app by app that has that pure scale of head count focused on this one question. So I like that. then they have some other strengths like they can transfer users. If they have you with TikTok, they can then, you know, sell you the other products so they can cross sell. But, you know, the one I’m thinking about is economies of scale and machine learning and user data. Does that give you an advantage in prediction that is structural? I don’t know. I’m, I’m, I’m trying to get to the answer. The way I’m trying to get to the answer is, is there a metric that plays out in market share? revenue or cost that is different. Like look, they are cheaper on a per unit basis than any other company, full stop. That would be a structural advantage. They can charge a premium. That’s an indication of a demand side advantage. They have exceptionally stable market share. That would be another. There’s no certain metrics that indicate the presence of a competitive advantage. I’m trying to see if this gets them there. That’s where I kind of am in my thinking. Okay that was kind of a bit of theory. I’m a little tired myself all that talking about theory. I think there’s some really cool questions here. The key takeaways just to bring it back, for those of you who are subscribers, the key concepts. This is under Learning Goal 32 which is Adaption, Innovation and Resilience as a Competitive Strategy. The key ideas, Social Capital, interesting idea to think about, Competitive Advantage, Rate of Learning and then Smile rate of learning, and you could also say smile marathon machine learning. I think those are the four to five ideas that I’m really sort of thinking about with regards to TikTok and bite dance more broadly. That’s the takeaways and they’re all in the show notes as well as the link to Matthew’s book, a couple of the slides which I’m summarizing which are his slides, and then the previous podcast 20 which was a little bit more basic about what is TikTok The takeaway I would have for all of this is, this is a strange competition. Like we know what Facebook is. Facebook is like the lion on the savanna. Very fearsome, top of the food chain, king of the beast, all of that. Tencent, also very, very good, but a different type of animal, which is, let’s say they’re a tiger. But they’re both after the same thing. They’re after your attention. That’s the game. So what is bite dance? Bite dance is like a dragon. It’s just different. The strengths are different. The structure is different. But it’s after the same thing. It’s your attention. Now, so if a dragon competes with a lion, who wins? I don’t know. They’re just different animals. They’re built differently. They have different strengths. And we’re gonna find out which is stronger and which is gonna win, because they’re going head to head now, at least in the US, assuming the government doesn’t shut them down. So yeah, it’s a new type of animal. Pretty compelling, pretty interesting. As for me, I’m doing pretty well. I’ve had a nice quiet week, been working away, meeting with some people. I’ve been talking with quite a few of you about this class and subscriptions and what’s working and what’s not. Getting great feedback on that. I think that’s really gonna make it jump forward and get a lot better, which is good. I’m starting to do some in-person meetings. I mean, I’ll tell you my dream, my goal, I guess. Education tends to be OMO, online merge offline. There’s a lot you can do online. That’s mostly what I’m doing. But I think the offline component, the in-person component is incredibly important as well. So that would be starting to organize in-person events, teams where people work on projects together and sort of take apart these questions themselves and then share their answers with others online. So maybe more of a community and started to do that in the last week to sort of see how it goes. And for those of you in Bangkok who were there on Saturday, my takeaway was that was awesome. Like it was great fun and really jumped forward in thinking and now we have people, you know, forming up in teams and putting together little slack groups and, you know, hopefully we’ll start to generate your own thinking. and maybe take some of these frameworks I’ve been throwing at you for a while and applying them yourselves, presenting them to each other, yelling at me, hey, you’re wrong, Prof, that’s totally wrong, right? That’s great, that’s what’s supposed to happen. So it’s trying to move more to an online community, offline events, things like that. We’ll see, we’re trying it out as of one, two, three days ago. And I think it went well, so that’s the hope. Let’s see, recommendations, carve-outs. I’ve been watching the Sarah Connor Chronicles, which was like the old Terminator TV show from 10 plus years ago. So I think it’s pretty great. Like I really liked the Terminator when I was growing up. I thought that was awesome. And then the movies just got awful. And now they’re really, you know, this TV show was quite good. So anyways, I’ve been watching it on Apple TV. Okay, I think that’s it for me. Thank you everyone for listening and a special thank you to all the subscribers. I really do appreciate it and I will talk to you next week.