We also support Hindi language, do you want change to it?
हम यह भी हिन्दी भाषा का समर्थन है, आप इसे करने के लिए परिवर्तन करना चाहते हैं?Yes(ह)
Episode 72 of Tech Buzz China features product expert Eugene Wei in conversation with our co-host Rui Ma on the timely topic of TikTok and the mechanics of what makes the app so successful. Tune in to learn why Rui calls TikTok an “algorithmic entertainment platform” and why Eugene terms TikTok’s algorithm the “Sorting Hat.” You’ll also learn why the app was well situated in coming out of China, and what the future ownership and trajectory of TikTok in the U.S. might be. This is a special episode for Tech Buzz — so tune in!
We recorded today’s program on the heels of President Trump’s executive order banning “any transactions” with TikTok or its parent company, ByteDance, by any U.S. resident. Though Tech Buzz usually covers Chinese tech companies in China — for example, our popular episode on ByteDance’s overall rise does just that — this headline was too significant for us to pass without comment. After all, TikTok is the first non-utility app and consumer internet product made by a Chinese company that has really hit it big globally.
Not to mention, Rui is currently working on an e-book on ByteDance! You can get updates on it by subscribing to our newsletter, at techbuzzchina.com. As always, past transcripts and other content are also viewable at pandaily.com and supchina.com. If you enjoy our work, please do let us know by leaving us an iTunes review, and by tweeting at us at @techbuzzchina. We also read your emails, at [email protected] and [email protected]. Thank you to our growing community for your always valuable feedback!
We are grateful for our talented producers, Caiwei Chen and Kaiser Kuo. Our intern Ji Junyi at Pandaily helped with the transcript this week.
[0:00]Hi Tech Buzzers! So if you haven’t heard, United States President Donald Trump issued an Executive Order banning “any transactions” with TikTok or its parent company ByteDance by any U.S. resident. The definition of a transaction is “to be defined,” but basically, almost everyone is interpreting this as a true ban of the app and that unless it is sold to a U.S. buyer, like has been rumored, there’s no way to continue using the service after September 20, which is when the order will take effect. And oh yeah, by the way, the administration also announced the same for Tencent’s WeChat app.
Yeah. While we usually cover Chinese tech companies in China — for example, even our coverage on ByteDance has mostly been on what the company is doing there, with its various video apps and forays in education, gaming, enterprise software, and lately, even fintech, this was just too big of a story to not talk about here on Tech Buzz China. TikTok is, after all, the first consumer internet product made by a Chinese company that’s really hit it big globally. In the U.S., we’re looking at estimates of 80-100mm daily active users. But even using the low end of that estimate, that’s quite a lot of people!
[1:25]For comparison, Snapchat’s most recent quarter showed 90mm DAU across all of North America. Sure that’s probably mostly the U.S., but still, TikTok is estimated to be higher than that. No wonder that now it’s up for sale, there’s a lot of talk around its value. But, for whatever reason, there is still this lingering belief — not everyone, but a good portion of people online it seems, believe that it’s not that mainstream, that it’s still just a tween lip syncing or dancing app, like its predecessor, musical.ly. Which by the way was also a Chinese app but acquired by ByteDance in 2017 for about a billion dollars.
We’ve covered the musical.ly acquisition and ByteDance’s rise overall on Tech Buzz before, and it’s one of our most popular episodes, Episode 28. And of course Rui is currently working an e-book on it, which you should subscribe to our newsletter to get updates on. But our point is that, if you can’t tell by now, obviously it isn’t all Gen Z’ers using TikTok. There are only 67mm Gen Z’ers in the U.S., after all. But you might be surprised by just how diverse the demographic of TikTok users is. According to Comscore, from May 2019, just 26% of TikTok’ers in the U.S. are under the age of 25.
[02:50]Which was a bit surprising for me to hear, but not that surprising. After all, looking at my own feed, there is a ton of content from creators who look about my age or even older. I’m certainly not getting tween dancing videos.
But that’s because TikTok is tailored to you, Rui. That’s its claim to fame, that its algorithm can very quickly figure out what you want to watch.
It really does. I mean, I’m not loving everything I’m watching, but a good amount. And that’s kind of where a lot of the magic is, in this algorithm where “the content finds you,” versus in more traditional platforms, where you have to “go find the content.” That’s a key distinction of the app, and probably the main one, but that’s hardly the only reason for its success.
[3:39]If you ever want to show off to your friends how much you understand about the mechanics that make TikTok successful, this is the episode to listen to. We have written a ton about the company — translated some of the most informative and revealing interviews of the founder Zhang Yiming and current China CEO, Kelly Zhang. So we’re drawing on that research, but we also have joining us today, one of the most thoughtful product experts here in the Bay Area, Eugene Wei.
If you haven’t heard of Eugene, I’m pretty sure after today you’ll be thanking us for helping you discover him. But today is where we veer off a bit from the typical business analysis that we do and really stick to product and design — partly because TikTok just doesn’t have much revenue here in the U.S. yet — the rumor is that ByteDance is expecting $500mm from the US operations this year — but also partly because we really want to hone in on why it’s just so different from what’s already out there, and also offer up our thoughts as to why it wasn’t created here in Silicon Valley. Bits and pieces, yes, but not in the coherent whole that you see in the current product.
[5:11]Hi everyone! We are TechBuzz China by Pandaily, powered by the Sinica Podcast Network by SupChina!
We are a biweekly podcast focused on giving you a peek into what’s buzzing within the tech community in China. We uncover and contextualize unique insights, perspectives and takeaways on headline tech news that don’t always make it into English language coverage. So you can be smarter about the world of China tech.
Tech Buzz China is a part of Pandaily.com, an English language site that tells you “everything about China’s innovation.” I’m one of your two co-hosts, Ying Lu.
And I’m your other co-host, Rui Ma. We are a part of the Sinica Podcast Network, created by SupChina! In addition to Tech Buzz, you can also find Sinica which covers current affairs. And we are also proud to be partnered with Financial Times’ Tech Scroll Asia, a newsletter on Asia tech news from one of the best publications in the business. Go to ft.com/tech-scroll-asia to sign up today!
[6:14]And as always, we are still looking for more reviews on Apple Podcasts! Send either of us a screenshot of your review, and we will gift you a free three month subscription to our Extra Buzz newsletter. You can send me an email at ying at techbuzzchina dot com. Several of you have already claimed this, and we love it when you do. Thanks so much!
[6:40]All right. Cool. Thanks for joining me today, Eugene. So happy to have you on the show to talk about TikTok. Please introduce yourself to our audience, to Tech Buzz.
Happy to and first of all, thank you for having me on. I’m a longtime listener, so this is really exciting for me. Most of my career has been at the intersection of technology and media.
I grew up in the age when the web and the internet got big and so I worked at Amazon.com for seven years in the early days, ‘97 to ‘04. Then, took a break from the tech world to go into filmmaking for a couple of years, which brought me first to New York and then to LA to film school, but then I got pulled back into the tech world.
I ran the product team at Hulu for its first three and a half years, moved back up to the Bay Area and did a number of different things over the years — did a startup, ran the product team at Flipboard, worked at Oculus on video related projects. That’s a quick summary of my career.
[7:47]Yeah, well Eugene is being very modest, his career is very impressive and you can find him at Eugenewei.com, where he has a collection of really interesting blogs. And he has a newsletter, which is actually where I first read one of the main articles that we’ll be talking about later today.
Let me explain briefly for those of you who are unfamiliar with how TikTok works. To describe for you, when you open up the app, it basically has these main features:
When you open it up, you see a full screen video feed. And this is your default homepage. It’s actually called the “For You” feed, or what many people refer to as the “For You” page. It’s an algorithmic feed. You can also skip out of that feed, and go to a feed where it’s exclusively composed of people you follow, you can do that as well. But the important thing to note is that the default is this “For You” automated algorithmic feed that the app just feeds to you.
[8:49]On each video of this screen, you can add or follow this creator, you can like this content, you can comment on it, you can forward it, or you can click through to the sound that they used to create this content and see what else is made with that sound. We’ll talk a little bit about that later, about how TikTok is really good at using sounds, specifically music, as a way to drive adoption and engagement. Other than that, there is also a “Discover” page. There’s some ads on top of what’s going on in the app and new features they want you to discover.
There are these trending hashtags. And these hashtags, as far as I can tell, are not by popularity — these are trending hashtags that are things that you might be interested in, or things that the app is recommending to you. And the hashtags could be about beauty tips, it could be about some political movement, it could be basically about anything, and people are using these hashtags and making content around these hashtags. So you can click through and see the content related to that hashtag.
And in the bottom of your app, there are five buttons. I just went through the first two. In the middle, you can click on this “plus” looking sign. And you can make your own video, so you can become a creator. Next to that is an inbox for messages where you can communicate with other people on the app. And then finally, the last button is your profile page. So it will show that everything you’ve created: it’ll show how many followers you have, how many pieces of content, any other profile information that you want is your sort of typical profile page.
[10:21]So one of the first things we want to address is that it’s often referred to in the media as a social media app. But it’s really not that “social.”
You know, I personally call it an “algorithmic entertainment platform.” And if that doesn’t really make sense to you, you can think of it like: it’s a short video, mobile first, a little bit interactive, and it’s more like YouTube than a lot of the other social networks out there. I don’t know if you agree with me, Eugene.
[10:49]Yeah. I think when people think of “social network” or use that term, they’re thinking more of something like Facebook or WhatsApp or, or an app where people are interacting with other people all the time. And definitely TikTok is more of an entertainment-first network. It’s only “social” in that it is loosely a collection of people that are connected through these videos.
Yeah. So, Eugene. Let’s just get right to it. One of the main reasons I wanted to have you on the show is because you wrote this really excellent piece called “TikTok and the Sorting Hat,” which everyone here really needs to read at Eugene’s blog.
[11:28]But why don’t you summarize for those of us who haven’t had a chance to read it yet? What are the main points, and what got you interested in writing this very long and detailed piece in the first place?
Certainly, I had been thinking about this for a while since 2018, when I had gone back to visit China and caught up with some of my old colleagues from the old Hulu Beijing office, some of whom worked at ByteDance now. And it just happened to be that TikTok has become super timely in the news recently, and I figured it was a good time to just finish my piece and get it out there, since everybody was talking about the app.
[12:05]But what got me initially interested in writing it was that — I always tell people that I’m a “cultural determinist,” which just means that I think culture explains a lot more phenomena in the world than people typically believe.
And I have been going back to China a lot over the past decade and observing their tech scene from you know, as an outsider, really. One of the things that I long thought was that China would struggle to have any of their apps really crack the Western market, largely because of the cultural barrier. I don’t think it’s anything particular about the Chinese tech scene; I would say the same of American tech companies trying to crack into the Chinese tech scene or the Indian tech scene. It really, I think, helps sometimes to understand local culture, and how you have to adapt and bend your business models to work in different cultures.
[13:00]So TikTok’s rise to some of those impressive stats that you shared earlier — that was really fascinating to me. I’m like, how did this app manage to become like the app for originating and generating so many memes in American culture?
It’s really fascinating to me. I have visited ByteDance offices in China and looked around at so many of the people working on the app, most of whom, you know, would say English is a second language. And that a lot of what’s on TikTok is totally bizarre to them.
And so that’s that’s kind of what got me started in looking at the app. And the more I looked at it, the more I realized, it was interesting for a variety of other reasons. And so that’s kind of what my piece is about.
[13:42]I think this story begins a little bit with musical.ly, which Alex [Zhu] and Louis [Yang] started way back in the day, after they were pivoting from this short video education app that wasn’t going anywhere. Alex and Louis had worked some in America too. So it wasn’t like America was this complete foreign concept to them. But I think even they would say that they didn’t necessarily set out to make an app that became so popular with teen girls in America.
The irony is that they had launched it — musical.ly — in both the U.S. and China. And, you know, strangely enough, it didn’t actually take off in China in the beginning. So they focused really on the American market. And they really became a popular phenomenon in that demographic in the U.S.
And Alex and Louis and musical.ly, I think, did a good job of listening to their users, as they asked for different features to help make it easier for them to make these lip sync videos.
[14:38]So that was going well. And the problem was, ultimately, there are only so many teenage girls who want to do lip sync videos and there is only so large a market for people who want to watch that. And so, when they had stopped growing after a while, they eventually did sell to ByteDance, which had approached them previously.
ByteDance in China had actually built a musical.ly clone called Douyin and it had taken off in China. I don’t know what was different between the first time Alex and Louis tried to do musical.ly and when ByteDance launched Douyin, but Douyin really became a huge hit. There’s something very funny about that circular journey where they cloned musical.ly, and they eventually bought musically and rebranded it as TIkTok.
And that’s where the story gets really interesting because, you know, musical.ly was on sale. And it was snapped up for a billion dollars by ByteDance. But there were a lot of other companies in the U.S. that could have bought musically and didn’t, largely because its growth has stalled out.
[15:36]ByteDance did two things, I think, to sort of get the growth going for the app that was now called TikTok. One is, they spent just an ungodly amount of money on marketing and advertising. For a year, I saw TikTok ads everywhere I looked, both online and in the real world. They were TikTok ads everywhere. And these ads are really kind of cryptic, it’s really hard to tell what the app is from your average TikTok ad. But that poured a lot of people into the top of the funnel for them.
And the second thing, and what I refer to as their sorting hat, which I think was critical here was their algorithm: this much fabled algorithm that helps to determine what videos are shown on every user’s “For You” feed or “For You” page. It’s one thing to pour people into your app, but you also have to help them have a good experience, to get them to stay around.
The rumor was in the beginning that all the people they were sending in their app weren’t retaining that well. A lot of people thought they were crazy to advertise so much. And I even was somewhat skeptical that that spend would work out in the beginning. The rumor was that the 30-day retention on all these new users was, you know, sub 10 percent or something really low. Their effective acquisition cost per user was super high back then.
[16:58]But I argue that it was important they did that because it’s not easy to break out as a network from some local maximum when you’re stuck at some set of users that kind of dominate the app. And this was the case with the teen girl lip sync video phenomenon.
You know, you look at apps like Pinterest, and how much they’ve struggled to broaden out from their initial sort of like female user base. Networks kind of have a path dependence. So the culture and the content that you have early on in a network can really establish the tone for what becomes dominant or acceptable in the moving forward.
So, they brought a lot of other types of demographics of users into the app. And then the algorithm effectively acts like the Sorting Hat from Harry Potter, in trying to help different groups of people find each other — people with similar tastes in videos to find each other.
[17:54]So if you’re a creator, and you don’t want to make a lip sync video — maybe you want to make a dance video, maybe you want to make a cooking video, or political humor, or any number of those subcultures — how quickly does your video find its way to the right audience? And that’s where, you know, I think TikTok’s algorithm is really the key to helping TikTok break out of its flatlined growth.
I’ve talked to people who have some access to some of the data. And they’ll even tell you that when they switched musical.ly onto the ByteDance backend and the algorithms, that the daily time spent per user spiked quite a bit. There really is something to this idea of the algorithm. That’s really important.
[18:41]The other key here is that if you think about competing with the social networking giants in the U.S., usually almost all of them follow the same path, which is that you have to try to get people to build up a social graph in your app to allow that app to then serve you a bunch of content. And the problem with that is that the social graphs that already exist in the West are so large. And people generally don’t enjoy the task of actually building up a graph to get a good user experience.
TikTok can bypass all of that in their app. They don’t need you to follow anyone to serve you good content; they just need you to watch some videos in the app and react to them. And because the videos are all very short, and generally pretty entertaining, it’s not that painful a process for any user to help train their “For You” page algorithm on their own tastes.
[19:33]The other important thing here is that I think that a lot of the social media giants in the West are running into some of the problems of trying to use their social graph as an approximation of an interest graph.
You see that clearly with Facebook where beyond some number of friends on Facebook, there’s a huge disincentive to actually post anything on Facebook because your audience is so broad. People often refer to that as context collapse or something like that.
But you even see it in other apps, right? If you look at something like Instagram, it’s sort of stuck in this in-between space. You may follow some people you know there. But you may also follow celebrities or brands and things like that: it’s sort of a mix of a social graph and an interest graph.
[20:20]I’d argue that TikTok, by sort of bypassing the social graph process altogether, can really just focus on your interests, and the interest graph. And really, if you think about it, where most of the money is actually made for social networking companies is actually on owning the interest graph. That’s how you decide what sort of ads to serve a user; that’s how you decide what sort of content to serve a user. And the social graph has always been an approximation of that.
That’s some of the magic of TikTok, is that it found a sort of orthogonal way to challenge the social media giants in the West. So that’s part of why the algorithm makes it interesting.
I also think that the algorithm is important in that it acts as kind of a quality filter. One problem with the Western social media networks with the follow graph or social graph model is that if you gain a lot of followers, you automatically get free distribution for your content, even if it’s not very good anymore. So if you have, you know, a million followers on Twitter, and you tweet, a good portion of them are going to see what you put out there. It’s going to get through their algorithm, even if you’ve stopped becoming interesting to them. So then you count on people muting that person or unfollowing that person to tune their own feeds.
[21:42]Whereas on TikTok, you’ll notice that there are some people who have a really popular video and if you tap into their profile, you’ll realize that’s actually their only popular video. They may have like dozens of other videos that have like no views.
And I think that’s good, because there’s kind of this weird meritocracy built into the algorithm where you have to continue to put out good content, even if you’re super popular, to get through to people’s For You pages. And that keeps their For You feed from becoming sort of polluted by people who are just really popular — which is a problem for Instagram and Facebook and networks like that. That’s another way in which the algorithm is important.
The other two points that I’ll make, which weren’t in this piece, but are in and pieces that I’m writing now is: well, how does this algorithm work? I think a lot of people talk about the algorithm as if it’s like, you know, what’s in that magic suitcase in Pulp Fiction or something, just some magical piece of code. It’s just like, better than everyone else’s algorithm.
[22:40]And I think if you ask most people who are experts in machine learning, deep learning, they would guess that the ByteDance TikTok algorithm isn’t necessarily based on anything revolutionary. It’s probably based on some very common and basic ideas in recommendation systems.
But what we’ve learned over the past decade, I think, is that it’s not just the algorithm itself, but the volume of good training data that you can train it on that really makes the difference. If you look at things like AlphaGo and GPT3, and some of the advances that have been made, we’re starting to learn that if you massively, massively increase the training data set, you can achieve some really great breakthrough results even if the algorithms themselves aren’t that much different than they were before.
[23:28]I think there’s a way in which the app of TikTok is designed in a way to feed a maximum amount of clean and useful training data into its algorithm. So you talked before about the UI, right? It opens to just one video at a time. This is actually pretty different and important here. Because when you’re only shown one video on the screen at a time, you are forced to make a choice to it. You can’t really have a neutral reaction.
And TikTok the app is going to observe everything that you do: Do you let the video loop more than once? Do you scroll away before the video even finishes? Do you share the video out by tapping on the share button? Do you tap in to look at the musical track, to see what other videos are made with that musical trap? Do you look at the profile of the person who made it? You know, all these things are signal and they’re very clean signals on that specific video.
Contrast that to other social media apps in the West, most of which adopt an infinite scroll, an infinite vertical scrolling feed. You know, on Twitter, on Instagram, on Facebook, you might have two, three, four stories on the screen at a time. It’s really hard, when someone is just scrolling up really quickly, to tell how they feel about any single thing on their screen at a time.
[24:54]And what TikTok really needs to know about any video they show you is: they really want to get a “thumbs up” or “thumbs down” from you on that video, so they know how much to promote that video to other people in the future. Other people like you, or other people with different tastes.
It almost reminds me a little of Tinder in that way. Tinder is probably, in the West, the most famous version of this: we’re only going to show you one thing at a time, because we really need to know how you feel about this person. And that’s why “swipe right” and “swipe left” has become a general part of our vernacular for just “approval” and “disapproval.” That’s part of it.
[25:35]The other part is that none of this would be all that useful if they didn’t have more than teen girl lip synching videos, because they’re not going to learn much more about you if they only have one type of video to show you. So the fact that they have that broader demographic of users, and they have more subcultures now, means that they have a diverse set of content to train things on.
We know that they have a large team that focuses on data labeling as well. So, they are going to characterize all the different features of the video, you know, this video is a cat video, this is a political meme, this is a dance video, all these things. So they have a really diverse set of content, and they’re getting clean signal back on all that content. So that’s the other point is just: their app design is what I call algorithm friendly.
[26:27]My last point on TikTok, which I find fascinating, is that I actually think they have creative network effects. What I mean by that is, because of the tools they have in the app, and the way the app is designed, every additional creator they add to the app makes creativity easier for all the other creators on the app. And this happens in any number of ways.
One is at the level of idea generation. So, if you asked me to make a TikTok today, it would probably take me a long time to just think of what I would do. I’m not going to do a dance video, you know, do I do a magic trick? It’s really hard to come up with creative ideas, because TikTok has a fairly high skill bar for its content.
But the nice thing about TikTok is because it’s this large community, most videos that are made are actually just adaptations of other people’s videos. So you don’t have to come up with your own idea. You can do the renegade dance, you can do whatever the latest trendy thing is, and just do your version of it. You don’t actually have to come up with your own idea. So that’s one part of it.
[27:32]The other part is that the app makes it really easy for you to remix or reuse elements from other people’s videos. You can borrow the soundtrack, you can do a duet with someone else’s video. You can use someone else’s video in a duet as kind of like a sequential thing in a story, you have these long linked duet chains that are really fascinating.
A lot of the tools and the camera tools are designed to make this really easy: it can all be done in the software layer. I think of that as a form of what’s called “assisted evolution.” It’s sort of like sped up evolution through human intervention.
The “Discover” tab is the last element of this creative network effect. You mentioned the Discover tab before. What it serves as is a coordination mechanism. These hashtags that they’re promoting are shown to everybody, and that tells the creative community: “Hey, if you make a video right now, with this hashtag, there’s a good chance that it will get seen because we’re promoting that hashtag through the Discover page.”
[28:35]And so, they can sort of focus and concentrate the efforts of the creative community continually on new challenges, whether they’re corporate sponsor challenges or just organic challenges. And that makes it so that it’s not just a chaotic thing — that there’s sort of a “school of fish” effect with their creative community. And these are all really unique elements that I haven’t seen in other social networks in the West.
[29:02]Okay. That was that was really, really insightful and really, really good.
I really liked — I wanted to jump in and interrupt you, in case there were listeners who didn’t know what a Sorting Hat was, that it’s from Harry Potter. But specifically why I thought Eugene, your metaphor of the Sorting Hat was so great is because the Sorting Hat tells you which House to go to without you telling it information. You just put the Sorting Hat on your head… and then it tells you, right. There’s this magic, like, “I can read your mind” element. And that’s kind of what the algorithm is doing.
[29:35]Seriously, like after I read your article — I have a copy of Harry Potter in my house, because my husband is a huge Harry Potter fan — I actually went back and read that passage, just to make sure I was remembering it properly.
That’s exactly sort of what the algorithm is doing. It’s not asking you — although now TikTok does, it does have that screen where you can tell it your interests when you first start and register — but you can skip it, unlike many other places where you have to choose three interests, or five interests, or whatever to continue. TikTok is like “whatever, I don’t care.” My understanding is, they put that in because leaving it out was confusing for U.S. users in particular, who are used to that interface; they don’t need it at all.
[30:19]Yeah, those interfaces are, to me, a little bit of a barrier. I try out so many new social apps, and they’re always trying to get a head start by forcing me to go through some long “interest selection” screen. And it’s not even clear that those are that effective anyway, so you feel a little bit like you’re forced to do work, and then you don’t get the payoff.
The nice thing about TikTok is that even the training process is relatively short and painless. If these videos were all like 10 minutes long or 20 minutes long, then it might not work as well, because you wouldn’t want to sit through so many of them before the app learned what your tastes were. But it’s so painless to do it in TikTok.
[30:59]The other thing I was going to respond to was the Tinder thing, which is hilarious. I thought it was very clever as well. But also, I think what’s a little bit different is that TikTok, in this case, is trying to match you to an algorithm where you are very happy, whereas I’m not sure Tinder’s purpose is to kick you off the platform and match you magically.
That, and I think the sort of single-video interface is something that they kind of stumbled upon. Reading their past interviews, it was partly because at the time Douyin was getting started, they could see musical.ly, but also this is the year — 2016 — they started developing it in early 2016, this is when big screen smartphones are all the rage.
So, how do you really leverage this thing that every smartphone maker is pushing as their core advantage: “You should get a bigger phone,” you know, “we have no edges.” That was one of the things that you could really capitalize on.
[32:03]Yeah, no, I’m certain there’s parts of these that are just accidental. I sometimes write about things and I talk to people at companies, or insiders will email me and be like: “that was just an accident!”
Yeah, “we’re not that smart, we’re not that smart.” Exactly. And actually, that’s part of the thesis I have going for TikTok, and why it’s so easy to use. And why did [it come up with] some of these at least “product level” innovations — maybe not the technology — because I agree, as you said, TikTok does not seem to have some amazing deep learning algorithm that no one else can figure out. This is all pretty standard knowledge.
[32:40]The reason why they were able to come up with something that was so intuitive was because they were dealing with Chinese users who are notoriously demanding, but also not that savvy. So when you think about smartphone users coming on board en masse starting in 2011, that’s quite a bit behind the U.S. And the fact that the U.S. had a lot of PC users already, so people had a certain expectation of what “social networking” means, what sharing and creating content means, et cetera.
So, for the Chinese users, many of them are receiving this and don’t know what to make. Some of the mechanisms that you talked about, which are hashtags, campaigns, the discovery page, etc., that really solved that problem. A lot of these things were kind of solving problems for, in my opinion — I don’t want to call them dumb — but let’s call it “less savvy” Chinese user, that happens to just work for everyone. Who doesn’t want a more simple, intuitive interface where I don’t have to think very hard about what to make or what to consume?
[33:48]Yeah. I would argue that this is a case where — you know, people in the West now talk a lot about how American tech companies should learn from China. But this was almost a case where an app that was in the West was bought by a company in China that was learning from its own market, and could see where Douyin you had already gone. And could envision that future for TikTok in the West, because they had already seen something that had worked.
I’d argue that the Chinese user base being newer to smartphones and the internet than the West actually really helped China in a number of ways, because it allowed them to sort of leapfrog concepts. And one of them was to leapfrog to video, as a platform or as a medium for more things.
[34:26]One of the things I’m struck with whenever I look at my — I keep a separate phone for all the Chinese apps just, you know, to keep up on them and study them. But also, when I go to China, I’m always looking at all my friends’ phones over there and looking at the different apps.
I think China is first to capitalize on video as a platform for everything: commerce, communication, dating, you name it. In the West, we’ve gotten stuck a little bit on more “first-generation” mobile app design paradigms.
[35:35]Yeah, I mean, it’s worked out really well for many companies. But at the same time, I think it’s true that, like you were saying, video is something that China has been excelling at.
A lot of the things you said are really great points. I do want to add to what you’ve said, because a lot of what you said was about the product itself — you have great product insight. And also a lot about the second-order effects of designing a product this way, or that way.
I think what’s sometimes I fear is a little bit missing in the conversation, which we’ve been trying to add here at Tech Buzz by translating firsthand presentations, interviews, etc. from the ByteDance management — and hopefully, if we have time, other amazing product managers and executives as well — is that there’s actually a really heavy operational element to TikTok. And I want to highlight this before we go on to the next section to talk about TikTok’s future fate.
[36:34]Because, when you look at Chinese companies, you have product managers, you have developers, you have sales, and then you have this group called “operations,” which is sort of growth and marketing and all this stuff at once, but really very, very hands-on in a sense.
So I wanted to give people some ideas of how Douyin first started its operations, and why that helped it become successful. Of course, as we all know now, Douyin did — I think it’s very reasonable to say that it copied a lot of the interfaces from musical.ly. Including what it said: it decided that it wanted to do HD fullscreen, which was, I think, musical.ly was one of the only ones.
And then, it picked music as a primary theme for it to go on, and making sounds being a huge portion of the product initially. So they would hire very experienced musicians and creators to make music especially for the app.
[37:29]And then they used lots of special effects filters — but that was, they will even admit at the time, because all this is from the interview from Douyin’s original creator, Kelly Zhang — and she says “Yeah, we did a lot of filters because it was just obvious. Look at the App Store rankings in China at the time: people love filters, so we did that.” And ByteDance is quite strong in that, because they have an AI Lab that allowed Douyin to leverage that very quickly.
But what’s really important is that aside from making all these “right” product decisions, like you said. So, I have this great product, it’s easy to make videos… but you have the cold start problem, right? Who’s going to come on and do stuff? Yes, I can create some campaigns. But how do I get more and more people involved?
What they did was they actually co-created a bunch of stuff with their early users. One of their most viral campaigns early on was this… it’s actually like a “finger dance”. I don’t see this as very common on TikTok, but in China in the Chinese version, in Douyin, actually that remains the most popular dance for millennials like myself. So we can’t dance, but we can like, move our fingers. So this was a concept that a super user early on co-created with the team. And that was one of the first viral videos, a “gesture dance.”
[38:51]And then really taking the time to be very close and good to their early users, and then actually taking their ideas and running campaigns with them, really helped them achieve early success.
They also did a ton of offline events. They did a lot of clever collaborations with organizations or objects that you wouldn’t associate with a music app. One of their most popular ones was with a series of classical paintings, where they took all these old paintings and made them into Douyin filters.
And also, just generally, operation-wise in China, for Internet companies, you’re talking about details: down to going in and commenting on fan forums, or on the equivalent of Twitter, and direct messaging individuals. So a lot of this hands on stuff, where you’re looking at conversion across every single channel, but then actively going out every day “operating” and executing on these very tedious things that aren’t automated yet.
[39:57]An example I like to give is Pinduoduo, a different app, a social commerce app that we’ve covered on Tech Buzz. They actually have people in large WeChat groups responding to customers and trying to get them to get on the app and buy promotions — that’s in addition to everything else that happens in-app. There is a real person doing this.
That’s the level of operational intensity that for many reasons that you don’t see here in the U.S. I think part of it is the labor costs, obviously. But part of it is also this… I love your idea of cultural determinism, but part of it is this idea from Silicon Valley that “if we can’t solve it with technology, then something’s wrong with our product” versus like, “oh, maybe there are some things that are better solved manually.”
So, instead of defaulting to a technical solution, there are lots of things in China that are still done “un-scalably,” or at least manually at this point. But that allows for community building early on, and retention — because you are interacting with a team member.
[41:02]Yeah. In the West, a common piece of advice that you’ll hear is that sometimes in the beginning when you’re building something, it’s actually good to do things that are non-scalable, because it’ll be harder for your competitors to match that. But China is the tech community that really lives out that credo to a much greater extent.
I was always fascinated, in the past few years going back to China, when I would meet with different companies, there was always a line item for operations. It’s just this amorphous group of people who do all sorts of things that are manual in nature, that need to get done.
[41:41]And you hit the nail on the head: part of it is that labor costs are cheaper, so you can afford to have a larger operations team. But also, it’s an attitude that it’s okay — if a problem needs to be solved, it needs to be solved and if manual is the best way to do it, you’re going to do that.
And especially with social things, or things that have network effects, I do think the ripple effects can be much larger from small manual interventions. If you look at the social media companies in the West now, I think you could say that most of them regret not having a larger moderation operation early on.
[42:20]Yeah, and this is not to say that operations is all manual. They actually use a ton of software to keep track of these many, many analytics. I’m just saying that the final execution sometimes is delivered quite manually.
Also, if you look at the KOL ecosystem in China, it’s in many ways much more professionalized and operationalized than it is in the West. And that happened even sooner there, that it turned into this industrial operation. And I think the West is sort of just catching up to that now in the influencer community. More professionalization and tools are being built out for that for that ecosystem.
[43:09]Well, that was a lot on TikTok and the product: why it’s successful and maybe a little bit of my theories on why it came from China, and what made the first parts of its growth so successful.
Let’s now talk about the fate of TikTok. So TikTok is, as we know, under Trump’s executive order to be banned from the United States, as well as its parent, ByteDance. So far, the most likely acquirer, if it is acquired, is Microsoft. Trump has also said, for those of you who have been paying attention to the news, that he wants a cut of this transaction for the Treasury. But either way, we’re looking at a potential Microsoft acquisition of TikTok’s overseas operations, including at least the U.S., but also potentially the U.K., India, Australia, I believe these are all territories that have been thrown into the mix as well. Let’s talk about what we think about that. I’ll go first.
So, if I take ByteDance’s stance — because many people have asked me about this: “What about Microsoft buying Tiktok?”
[44:15]Well, if I’m the side of ByteDance, originally, I was thinking it’d be horrible to sell. But if the acquirer is Microsoft, because Microsoft is not directly competing with ByteDance in China in any way, and so far not yet internationally either, I don’t think it’s actually a horrible outcome. I think it might actually be fine, right? If it’s going to be banned, then at least get some cash from this company that clearly has a lot of it.
Obviously, it would be a terrible idea for ByteDance to sell TikTok to Facebook or Google, even assuming they could, right? Because they probably mostly likely can’t. And you have Amazon, Apple, etc., who’ve all clearly stated that they’re not interested.
From Microsoft’s perspective, though, I do think there are real risks. It’s been already mentioned in Chinese media that if Microsoft takes over this content platform, what if there are, you know, eventual content that surface that are not friendly to Chinese people? Because there could be Chinese people who look at this content and personally feel offended, and are like, “Okay, I want to boycott Microsoft now,” right? So there is a risk there. Maybe not a big risk, but it could happen.
[45:30]There’s also discussion of, you know, amongst Chinese people where they feel like their private information is potentially in danger, because who knows what Microsoft is buying: the algorithm? What data exactly? And is there some way that by buying this algorithm, I don’t know, that Microsoft can access Douyin data? I think that’s very unlikely, doesn’t really make much sense to me, but these are concerns people have raised.
That, and of course earlier, we were talking about the operationally heavy element of an app like TikTok, like Douyin. And it just doesn’t seem like Microsoft has a lot of experience with that.
[46:10]Yeah. Microsoft is… certainly, I don’t think it’s the first name I would think of as a logical acquirer. On the other hand, you know, perhaps because they don’t have a long consumer DNA, they’d be more inclined to leave TikTok alone and let it run independently, kind of like a LinkedIn or something like that. And that might be good for the TikTok team, to be left alone.
It’s unclear whether Microsoft really understands what TikTok is or what it’s buying. But in its own right, it’s a valuable asset. So I can certainly see why Microsoft would love the chance to pick it up on the cheap.
[46:46]Microsoft seems to have good relationships with this administration and also better relationships in China than most of the competing buyers. So they might be one of the only ones with the chance to pull this off.
There aren’t a ton of natural synergies. I mean, there might be some with Xbox, which is really kind of like, what I think was Microsoft’s only real consumer business.
But, you know, ultimately Microsoft brings a lot of technical infrastructure and money and resources. And if they were willing to support TikTok and let it run somewhat independently, I think it could actually be sort of like a friendly owner.
[47:28]Yeah. Aside from Xbox, Minecraft, you know, I don’t see them as having too much more on the entertainment side. I don’t think you can call GitHub “entertainment” unless you’re a very specific population, so I’d agree with you there.
It’s a little bit odd from the lack of consumer DNA. But I think everybody can see TikTok’s recent trending growth, it’s DAUs. If you look at its revenue as sort of a lagging indicator, then you would expect its revenue to follow more of a Douyin-like trajectory of growth in the future. And just for that in itself, at the different purchase prices I’ve heard so far, I can certainly see why any tech company would just want to own it as an addition to their portfolio.
Yeah. What about Instagram Reels? So a lot of people have been asking, “Hey, does Instagram Reels have a shot against TikTok?” I’ll let you go first on this one.
[48:25]I will admit to not having used Reels a ton yet. I’ve just been playing with it a little bit just for research purposes and flipping through it. I think there’s a broader problem for Instagram right now, which is that they are trying to be so many things at once, that their app is just sort of overloaded with things. Reels doesn’t have a dedicated Home. It’s not like a button on a tab anywhere; for me, it’s like I only get to it by tapping into the “Discover” page and then tapping into the top featured item, which is Reels right now.
So it’s a little bit weirdly buried inside an app which already is overloaded with Stories and the main feed and a messaging app. And, you know, their creation tools sort of differ by category, you have different creation tools for Stories, versus Reels, versus posting to the main feed.
And in some ways, this is a little bit of the classic Innovator’s Dilemma, where you never want to give up on the thing that got you there, but eventually that means that you’ve built like 80 different things piled on top. And a pure play competitor like TikTok can come along and be very clean and simple and focus on one thing and be less confusing than you as an app.
[49:47]The second thing that I would say is tricky for Reels is that Facebook’s inclination and Instagram’s inclination will always be: “Well, let’s copy someone else’s thing, and then slap it on top of our social graph, which is already massive. And that will give it a head start and a jumpstart.”
The problem is that I think sometimes their graph is a detriment to the way a new ecosystem evolves. In the case of Instagram, some people do use it as part social communication product and part as interest graph product. But it’s not clear to me that the people I follow socially actually are the ones whose Reels I want to see. And so, so far to me, the Reels Algorithm hasn’t felt as tightly fit to my taste. What do you think?
[50:41]Same as you, I would have to say I was disappointed when it was released just because you know, it’s not like Facebook doesn’t have the resources or talent — I have many talented friends who work at Facebook. And also the fact that Reels actually launched in Brazil late last year, and India earlier this year, so it has been around for a while. So the U.S. and some other countries were the last batch to get it. So I wasn’t super impressed, given that it’s been out for a while.
I also think that the algorithm baffled me a little bit, because I was also seeing, not like in a row content from the same creator, but I would say every couple of reels, I would get the same creator again, which rarely happens on TikTok. So again, I’m not a heavy user either. And I do understand that the product is going to evolve, there’s lots of people working on it. I just have to say that it wasn’t super impressive.
I think the main point, though, that I have is that short video is not going to leave anytime soon, just because of the fact that it’s much easier to create than some 10-minute YouTube explainer, which has a relatively high bar.
[51:54]Whereas on TikTok — the reason why it’s so popular, as you alluded to earlier, is because the creation element, even though it’s harder than doing Instagram Photo, is still relatively easy. And the tool has been so optimized for that, and the community, and with all the hashtags, that you know what to create, and how to create it relatively quickly. I haven’t posted any TikToks, but I have tried using the tool, and it’s pretty good.
I think what we do see is that eventually, I think what’s probably going to hit up against both TikTok and Instagram is that for higher quality content — and we already see this in Douyin in China, where they’ve announced they will have videos of up to 15 minutes — is that there will be longer and longer videos. On TikTok, you could already do 60-second videos, and Douyin can do I think 5-minute ones, with up to 15 minutes later.
But basically you see that the sweet spot though, as we can see from China — ByteDance has the Xigua app — the sweet spot is like four minutes. Their most popular videos seem to be less than five minutes. So this short video concept is here to stay, and so far Facebook doesn’t have anything in that category. So Reels might be their best bet. Even if it’s an awkward fit currently, I don’t see that giving up on this entire category. They’re definitely going to put more into it.
[53:25]Right. Yeah. I think there’s two things that I think. One is that I don’t think that Facebook or Instagram would have ever invested the amount of money that ByteDance did into TikTok. There’s rumors that the year ending March this year, through that advertising spend, that there were nearly a billion new users introduced to download TikTok the first time.
I’m trying to imagine that level of commitment from Facebook or Instagram to a product that they’re only building to knock off a product that’s doing well in the marketplace. I’m not sure they internally have their own conviction to put that much into making something like this work.
[54:14]The second thing is, I think that one of Facebook and Instagram’s problems traditionally with video as a category is that their monetization for creators is so poor. There doesn’t seem to be anything on the horizon to really effectively change that. When they started paying for videos, or more premium videos in Facebook, they famously did not want any pre-roll ads so that they could serve the short previews in the Newsfeed. But since there were no pre-roll ads, and most people weren’t watching the videos for too long, they never made it to any other ad units.
So people were making all these videos for Facebook and making no money off of them. Facebook was helping to subsidize their creation costs, but I tend to think creators in the long run are going to go where they can monetize. And I feel like ByteDance and TikTok, in general, have been better about working with their creators to try to help them with monetization. Sometimes they set up marketplaces to connect creators with brands — they just had that billion dollar [$200 to start] Creator Fund.
I think in general, China as a tech scene has done a better job of monetizing video content. Even the versions of Twitch in China monetize better than Twitch does for creators in the U.S. In the long run, I do think that matters because the creators are one side of that multi-sided entertainment marketplace.
[55:39]Yeah, absolutely. That’s interesting that you said you didn’t believe Facebook is that committed to it. Maybe I’ve just been too deep in researching for my ByteDance e-book and starting to take on that company’s thinking, which is a very typical Chinese company’s thinking, which is that “if this is a big market, I cannot let anyone else have it without a fight,” right? Not necessarily that I’m going to win this, but it’s a huge market, I have to put up a really good effort. It can’t just be like, “Oh, well, that’s that’s Alibaba’s thing,” you know.
I was going to ask you about YouTube. I personally have no opinion at all. The only thing I’ve heard that kind of struck a chord a was that it seems like not a ton of creators are super happy with YouTube. So if someone were to come in and treat the creators a lot better, which is which was your last point, then maybe it’s just not much of a threat.
[56:34]Yeah, YouTube. Well, there’s this rumor of YouTube Shorts. I don’t hold out too much hope for whatever that app will be. YouTube, shockingly, hasn’t really shipped any creator tools of any note for forever. I mean, they have the ability for you to upload from your phone. I think it’s the only thing I can think of that’s anything in the line of a creation tool.
And in terms of monetization and support for their top influencers, I think most influencers realize eventually that they really have to take control of their own fate: Go sign deals with brands, sell merchandise. All this transacting is happening off of YouTube.
Even the really basic things that China has: the ability to charge a subscription, the ability to take digital tips and digital gifts — YouTube doesn’t really have anything on any of those fronts. And so I think it’s been really, really slow in the past to really capitalize on their position in video. And that doesn’t give me too much hope that moving forward, they will suddenly get a culture of shipping quickly and being more creator friendly. So I’m not overly concerned about them.
[57:55]Which is unfortunate because out of all these apps the only one I pay for, because it has a premium option, is YouTube. It’s got like, some of the most interesting content — and some of the most annoying ads.
You know, in some ways people would say its algorithm is similar to TikTok’s, in that they both sort of exploit heavy algorithms. It’s like, if you like sugar, we’re going to give you a dozen Krispy Kreme Doughnuts. It’s very much about giving people more of what they’ve indicated they want.
But their algorithm seems easily gameable. And a lot of creators will tell you that like, okay, yes, I know it’s all about click-through rates and how long into my video people watch. So you’ve noticed that YouTube videos have gotten much longer. From a lot of creators I know, they’ll do tactics like delaying the key reveals or something til later in the video so that you have to keep watching into these really long videos with like 10 minutes of filler content, which drives me nuts. One of the nice things about the TikTok algorithm is that it doesn’t have any of these perverse incentives that are being gamed right now.
[59:03]What about Twitter? Because people were so surprised that Twitter was in the conversation. I’m pretty sure everyone who has been in the conversation does not make sense — as someone who formerly worked in banking, I’m sure that even if you have no interest, your banker’s going to come and try to convince you to have some interest.
I think with Twitter, it’s obviously a lot smaller. You have some of the same risks as Microsoft in terms of content moderation, but they’re a much better match in terms of the product, right. What do you think?
[59:34]Yeah. From the perspective of being an app that I think could be an interest graph — I think Twitter’s strength has always been that it could be a competitor on the interest graph space — in that way it would make sense as a match. On the other hand, like you said, they don’t have enough money, so they’d have to bring in outside wallets to join them in putting together a bid.
But secondly, they owned Vine at one point. And they kind of just let it wither and then die out. They have a really poor track record of shipping quickly. They’re one of the more slower moving companies in the social media space.
And so if I’m TikTok, I’m not sure I want to be owned by Twitter, which, you know, let Vine just sort of like die off. And, I feel like Twitter just has so many other problems of their own to solve. So it’s not like an illogical match. I’m just not sure… I worry about any company that would get bought by Twitter at this point. Haha.
[60:36]Right. And they don’t have the best government relations… or maybe you treat it as they do have the best government relations, I don’t know, depending on what you think.
Yeah. I mean, I would love to see Twitter learn some from TikTok in terms of just the ability to push harder towards an interest graph based model. I think that’s always made more sense for Twitter.
I think Twitter onboarding today, this idea that you still have to follow accounts and everything is just so outdated. I don’t know what the Twitter onboarding looks like as of this very moment, but the last time I checked into it, it was still sort of like, “Hey, here’s some accounts we think you should follow. It’s like, you know, Donald Trump and Katy Perry” and I’m just like, it’s super bizarre that it hasn’t gotten better over time.But yeah, that’s kind of the story with Twitter.
[61:31]Right. And then the last thing I was gonna say was, you know, there’s been some names thrown out. The only one I personally thought was interesting with Disney. Quite a few friends floated this, but it would be hilarious if Disney acquired it, and then Kevin Mayer went right back to Disney… In fact, they were like, “maybe that’s why he was hired?” I don’t think so.
But I think that basically, they probably have a huge content risk that this would probably not fly. But strategically that would have been interesting. It would be nice to see other companies try to get into this new emerging digital media space, other than sort of your same familiar three, four players.
[62:17]Yeah. If I were any media company I would be interested in TikTok. It’s got such a foothold with younger users that it would help some of these media companies that are seeing their audiences maybe age up a little bit.
But you’re right — Apple and Disney pride themselves in sort of like family friendly type of content. And it’s not to say that all the content in TikTok is salacious or anything, but they certainly have more edgy content than an Apple or a Disney as a brand are used to, and so that that would make it a challenge. I think you know if you looked at Warner Media, if they weren’t you know more saddled with ATT debt or something, I’m sure they would be very happy to own a TikTok as well, as they’re trying to become more sort of future-facing.
I actually think Amazon, though they may not be interested, really should be in the bidding for TikTok. One, because Douyin’s such a commerce powerhouse. And Amazon tried a short video commerce thing that was, you know, around and then killed off so quickly that I never even got to use it. I don’t think it’s their strength to build that, but they certainly have the backend operations and payment operations and all of that to enable more commerce. And they also have Twitch and other video assets that I think could be a natural synergistic fit with TikTok. So I don’t know why they’re not bidding. I think TikTok actually brings a lot of skills and things that they could use.
[63:47]And look, Netflix may not be in TikTok’s space directly, but if people are saying that Microsoft could be a buyer, then I don’t think Netflix is any less logical of a buyer. They’re at least more in the entertainment space as a whole. They are always competing with Fortnite and things like that, and TikTok certainly is an attention space competitor to them that would diversify their portfolio.
Yeah, I think it would have to be one of these companies that can directly leverage the business. I think all these private equity bids and things that are trying to be assembled, it’s tough to see any of those win out. They’d probably have to ally themselves with one of these more logical buyers.
[64:30]Again, the asset being sold off right now is TikTok only, but if it were Douyin, which includes all these live streaming functions, ecommerce, etc., then it would make a lot of sense for Amazon — and for that matter, for Netflix as well — because it becomes such a bigger chunk of your entertainment.
We had a guest a couple episodes ago on live streaming, and he made a good point, which is that probably reality TV can be replaced by live streaming. So, still not your produced, cinematic content, but sort of your reality TV unscripted, not high production value stuff. Yeah, I could easily see it being replaced by live streaming. But that’s not what’s being sold. That’s not not here in the U.S. yet.
[65:18]There’s one other option, which I guess is off the table given the Trump administration’s sentiment toward this: which is just to spin out TikTok and let it be independent in the West. I actually would love to see that, because I think with the right management it makes for an interesting competitor to the other tech giants in the West. And I wouldn’t mind seeing that as well, but it doesn’t seem like that’s an option on the table.
Well, I think those are all my questions. Do you have any parting thoughts that you wanted to share with our audience?
Not anything specific to TikTok. I would only say and conclude that: I do think when you look at short video as a platform in China, it shows the potential for how short video can be a much larger platform in the West.
Regardless of what happens to TikTok in this transaction, I would love to see — and I believe we will see, in time — that the West will start to catch up with China a little bit on this front, and see short video for the broad platform it can be. Whether it’s for commerce, or education, or dating, or job market, or whatever it is. I think this is definitely one category where China is leading the way.
[66:41]Yeah, short I think short video and live streaming both are just not part of my life, for me or my friends’ lives here in the U.S. very much, but is very much part of Chinese citizens’.
Okay, Eugene. Thank you so much for coming on the podcast. Why don’t you tell our audience where they can find you, and how to keep up with everything that you’re doing.
Thank you so much, Eugene, and make sure you do go to Eugene’s website and sign up for his “Remains of the Day” newsletter. Okay, that’s all!
[67:36]All right! That was the bulk of my chat with Eugene. What did you think about what he had to say? Did you learn more about what makes TikTok Tick? Send us your feedback!
Thanks for listening and don’t forget to write us that review for your free Extra Buzz subscription. Have any questions or suggestions? Email us! We really enjoyed putting this together, and we’re always open to any comments or suggestions. You can find us on twitter at thepandaily, at techbuzzchina, and my personal Twitter account is YINGLU2020.
And my Twitter is spelled RUIMA. Tech Buzz China by Pandaily is powered by the Sinica Podcast Network on SupChina. Pandaily.com is an English language site that tells you “everything about China’s innovation.” Our producers are Caiwei Chen and Kaiser Kuo. Thank you for listening!