Ranjan here, and today I'll be talking TikTok. Or more accurately, TikTok's owner, ByteDance.
I worked on a personalized news app startup from 2014-17 (Informerly, RIP). We didn't pull it off, but this newsletter isn’t going to be some Medium.com soliloquy on startup failure. It’s about that mysterious "news app" in China I kept reading about during the journey.
That Elusive Newsfeed
I first came across Toutiao when reading them raising $100 million in 2014. I couldn’t believe it. I had never heard of it and everyone I asked in NYC hadn’t either. How the hell do you have a $100mm Series B for a news app (and before the unicorn craze, too)? I tried reaching out to some business school alumni from China, with the universal response being "we don't really use it - it's kind of trashy".
This went on for a few years. I would ask my wife's Taiwanese family members. I would ask anyone I met in the media world from China. I messaged people on LinkedIn. No one could give me a straight answer of what exactly “it was” other than a kind of low-brow content thing.
My co-host Can wrote about the "splinternet" earlier this week, and this was one of those moments where you realize what a truly divided internet looks like. Here was a startup that supposedly did exactly what we were trying to do, that had raised an ungodly amount of money, and we were completely blind to what it actually was.
They finally crossed the ocean with the November 2017 acquisition of Musically and subsequent launch of TikTok, which is so damn good:
But first, Toutiao, which is supposedly sitting on some incredibly advanced AI technology thanks to the trillions of content data points they've amassed. They've been reported to pay top engineering talent $3 million and were early on the creating a "robot reporter" during the 2016 Olympics.
But what is most notable about their approach to content recommendation is that there’s no social graph. This October 2017 deep dive from Y-combinator dug into why this was so important:
Without any explicit user inputs, social graph, or product purchase history to rely on, Toutiao offers a personalized, high quality-content feed for each user that is powered by machine and deep learning algorithms.
Toutiao’s underlying technology learns about readers through their usage – taps, swipes, time spent on each article, time of the day the user reads, pauses, comments, interactions with the content and location – but doesn’t require any explicit input from the user and is not built on their social graph. Today, each user is measured across millions of dimensions and the result is a personalized, extensive, and high-quality content feed for every user, each time they open the app.
This was not Facebook and is kind of what @Jack wants Twitter to be. Contextual or topical personalization was exactly what we had been striving for with our startup, and this was the vision realized, at massive scale.
But I still couldn't get a clear picture on what this magical machine learning really felt like. A VC friend in China gave me the disclaimer, "it doesn't take too much AI to personalize girls in bikinis and people falling down stairs at scale". Was it just porn-y, addictive content, or was there something truly advanced in how they created personalized newsfeeds?
ByteDance vs. the BATs
Toutiao was launched in 2012. They quietly grew to an astounding $75 billion valuation by 2018 without ever taking money from Tencent or Alibaba. Instead, they had investors like Yuri Milner (DST) and Sequoia funnel in Softbank-ian levels of capital.
The investment source is worth noting because most of the recent big Chinese IPOs were all either spinoffs, or heavily connected to one of the BATs (Baidu, Alibaba, Tencent). Bytedance has managed to chart a unique path. Their relationship with the Chinese government is also not one of clearly anointed domestic champion, as they’ve had some major run-ins. They had their apps completely shut down in early 2018 for distribution of "vulgar" content.
This was remedied with an apology from the founder which was incredibly intense (I’m imagining Zuck or Sundar saying these words). This is a short analysis of how it had a self-confession feel to it, along with full English translation.
Jinri Toutiao's available numbers are sparse, but what’s public is interesting, to say the least. They were reportedly on track to make $7.2 billion in revenue in 2018, yet they are not profitable. I'm generally skeptical when it comes to a lot of money-losing startup giants (Happy Uber IPO Day to my cohost Can!), but this is really baffling.
Bytedance isn't Uber, it's Facebook. Facebook lost money before it built a flywheel1ish ad business. They have raised about $4 billion in capital, and have billions in cash flow. If you're projecting $7 billion in revenue, you're a very mature ad business and you'd think they would be playing at Facebook's ~30% profit margins. It’s a fully automated aggregation app, what are their costs!?
Unless, of course, you're heavily investing in something. Which brings us back to TikTok.
This was a fortuitous week for The Margins to cover the globalization challenges of the internet, given the backdrop of the trade war escalation. This is something that makes the U.S. popularity of TikTok so fascinating. Alibaba never made it in America. Tencent's WeChat Messenger never really attempted it (other than among the Chinese diaspora). Once again, Bytedance took a different path.
A very quick history of what is now TikTok:
Bytedance had launched a short-video app called Douyin in Sept 2016.
They began their overseas expansion of this app in Southeast Asia as TikTok in Sept 2017.
They leveraged their Nov 2017 Musically acquisition by moving all users over to its TikTok app in August 2018.
By November 2018 it was already #1 in the Apple app store.
The rocketship had launched, and quietly, a Chinese internet giant had landed on American shores. The moment I used it, I finally felt that magical Toutiao machine learning that had long evaded me.
I didn't even create an account. I just looked around the search page, watched some videos, scrolled past others, and downloaded a few videos (yes, TikTok is as free-flowing as content comes and lets you easily download everything away from the platform).
Within a few uses, my feed turned into this mesmerizing series of shuffling dance tutorials, East Asians dancing, South Asians dancing, people switching outfits in rapid succession, low-budget comedians, #DogsofTikTok and yeah...girls in bikinis. Six years of waiting, and I felt like I had finally tasted Toutiao. Which is so damn good:
Tariffs and TikTok
I still can't believe its not a bigger story that a Chinese-owned app is quietly taking over the American market. Think about all the times American tech firms have stepped over themselves to try to access the Chinese market (my favorite is imagining Xi Jinping's reaction when Zuck asked him to help give his kid a Chinese name).
Google's employees are revolting against its DragonFly project, Amazon just this week shut down another marketplace attempt, and Uber’s China operations were merged into some financial engineering whizbang corporate structure. Meanwhile TikTok has quietly surpassed even cool-kid Instagram in downloads.
Will TikTok end up caught in the US-China trade conflict?
Keep an eye on the organization CFIUS (The Committee on Foreign Investment in the US). It's the multiagency group that evaluates business transactions for national security threats, which just happened to see a major expansion of power last August.
CFIUS had previously prevented deals like a merger between the American MoneyGram and China's Ant Financial, but last month, we actually saw them break up an existing transaction. In April, they ordered the 2016 acquisition of Grindr by a Chinese internet company dissolved on data security grounds. Things are moving.
Is TikTok a Data Security Threat?
David Carroll wrote a great piece this week on the Terms & Conditions of TikTok and how it represents a possible "Cambridge Analytica data bomb waiting to explode".
Another great piece on the topic from Nick Frisch at the NY Times, "We Should Worry About How China Uses Apps Like TikTok":
The critical difference with China lies not in the wiring of chips or lines of code, but in history, culture and scale. By custom, Americans trust that the contents of their snail-mail letters are protected by the Fourth Amendment; government access requires a judge to issue a warrant. American tech firms routinely stand up to the United States government.
China is different. The People’s Republic has always reserved the right to open its citizens’ mail at any time, for any reason; there is no basis to believe its basic approach will differ because the technology is new. Ironically, the internet, which Western techno-utopians prophesied would liberalize China, may instead allow the party to indulge previously impossible fantasies of mass control.
I'm very curious to see what level of American headspace will have to be occupied by TikTok before it ends up on CFIUS's radar. Maybe their members just haven't downloaded it yet.
Eating away from the inside
Regarding business, not geopolitical, threats, I had linked a few editions back to this amazing Bloomberg piece on TikTok which gave this insane stat:
Over the past three months, for instance, 13 percent of all the ads seen by users of Facebook’s Android app were for TikTok
We know this incursion is a heavily-funded onslaught that's working perfectly. It reminds me of how brutally well executed Facebook's launch of Instagram Stories was, except instead of leveraging a social graph for growth, it's nonstop app install ads, and this time, Facebook is the one serving the ads.
I'm the first to call Facebook anti-competitive, but you have to give them credit for taking millions of dollars in ad revenue for install ads from a direct competitor.
This newsletter was a long one, but the ByteDance and TikTok saga has weirdly felt very personal. It's quite something to see a personalized news app startup, which happens to be the world's most valuable startup (at writing time it’s now back above Uber!), and which I think will find itself at the center of a global geopolitical conflict.
Maybe we can fight back with Flipboard….or Twitter could’ve just not shut down Vine.
My fascination with the Chinese market goes back further than trying to track down ByteDance. I covered East Asia during my trading days, and the Chinese Yuan was a focus. That led me to spending a summer in Beijing before business school in 2009, with that now cliched goal of learning Chinese (spoiler: I didn’t). There are sometimes those moments in life where it seems every intersection of your interests come together. It happened that summer.
I had been reading a China-focused financial blogger named Michael Pettis for years. It turned out he owned a rock club in Beijing called D-22. I went a few times, and in this music venue weirdly owned by my favorite finance blogger, I’ll never forget a performance by a band, Carsick Cars of this song Zhong Nan Hai. In perfect Chinese cultural fashion, it’s about a popular cigarette brand, while cheekily referencing the HQ of the Communist Party, of the same name.