We wrote about ByteDance back in 2019 from a journalistic account of China’s tech titans. Among the many companies scanned then, the one
that’s garnered the most news both about its obvious success (and also the unwelcome political kind in recent months) has been the TikTok, a subsidiary of
ByteDance. Matthew Brennan’s new book “ATTENTION FACTORY – The Story Of TikTok and China’s ByteDance” was a natural one to examine given our firm’s deep research interests in artificial intelligence driven applications markets.
As a former business strategy advisor who also ran the North America strategic operations of an Asian movie-streaming & mobile-gaming platform in a past life (part of one of the largest industrial-to-consumer conglomerates to emerge there), the author has operated deep within trenches of building business & platform strategy for one such consumer platform and managed complex funnel and monetization metrics that underlie such digital business models. Much of the insights from those experiences inform research frameworks his firm now leverages for analysis of subscription business models among enterprises today.
In just two years, TikTok has emerged to rival companies like Netflix, YouTube, Snapchat, and Facebook with more than one billion downloads in 150 markets worldwide and 75 languages. On the app, homemade videos showcase everything from comedy to lip syncs to dog grooming tips that users create and share on their phones. The scrappy, goofy, fast-moving content has hooked young audiences around the world. Since little translation is required, TikTok reaches well beyond other successful Chinese apps.
It’s the go-to video sharing app used predominantly by GenZ. It specializes in micro-targeted content served through slick recommendations. 60% of its users skew 16-24 years. It had 50M daily users and 91M monthly average users this year. Massively addictive content platform. Its amazing algorithms is the secret sauce that match content based on your own engagement with video content, does not depend on your social network or other influences or friends, contrasting with other platforms. In 2018, TikTok ranked fourth worldwide as the top non-game app downloaded, at 663 million behind only Facebook at 711 million and its related apps WhatsApp and Messenger, SensorTower data shows.
As the first China-made consumer internet product it has a genuinely far-reaching global profile. It also had to deploy a strategy of dual versions of TikTok – one for China’s internet censored market and another for the rest of the world. Interestingly, about one-quarter of TikTok’s downloads came from India. TikTok’s inroads in India and its young, mobile-savvy population is one of the big reasons for its early success s well. The app makes money through ads and from the sale of virtual goods such as emojis and stickers to fans. An easy-to-use interface combining click-baity news and entertainment with powerful AI to precisely match users rather than recommend content based on their viewing habits and “likes” have fueled the app’s success. In 2018, the app had gained some 30 million new users within three months.
ByteDance enjoyed a 9x growth, from 11M users from just couple of years ago to 100M users in 2020. Its revenue grew from 200-300M M (est.) in 2019 to about 1B in 2020. It is also expected to grow 5X to $6 B next year according to Wells Fargo research.
Contrast with Facebook whose revenue only doubled pre-IPO. WeChat has 15M US users, seen a Chinese version of WhatsApp in the US market. Used as the preferred mode of communication in China. Russia’s Telegram has a small 3M user base. Korea’s Line which is gaining usage due to Chinese users transitioning from TikTok has just 3M users in the US.
The primary appeal of Matthew Brennan’s new book is his meticulous effort in going behind the scenes, looking under the hood and taking a deeper look at the business and technology strategy underlying the immense success of this consumer content platform, and an analysis using popular business templates. Much more than what an average journalist could typically cover in a book.
We learn about the subtle interaction-design elements gleaned from focus groups, constant experimentation with various content elements and insights gained from intense machine-learning supported analysis of video clip types and genres that made TikTok such an addictive experience for its target market of teens and pre-teens.
We learn about the technical infrastructure elements that had to come together to enable TikTok’s vision of a video sharing platform – the launch of the first real smartphones (then the iPhone 4S) in China, the evolution of the mobile network to 4G that could handle the bandwidth of video streams seamlessly as well data plans affordable enough for people to enjoy the experience of enjoying videos on their mobiles. Ultimately mobile screens became the most hotly contested real estate in China and a formidable choke-point in gaining people’s attention.
Also existing were some significant cultural differences between mobile video creation/consumption habits of Western teenagers and their Chinese counterparts who are often too busy studying or preparing for exams while in school to find time to create videos. It didn’t help either that censorship as a risk exists for all media companies in China to deal with, with the key to survival being an ability to understand and anticipate the authorities’ playbook.
We also learn that TikTok did not turn into a runaway success overnight. Like most other startup ventures that went through its share of app development heartbreak and funding struggles targeting a highly competitive and saturated female pre-teens market with its peculiar brand of short silly meaningless videos.
ByteDance managed to differentiate itself from the current crop of apps with high code-bloat and their resulting poor user experience. It did so by building one that was light enough on tiny lower resolution screens of early mobile devices that operated on lower bandwidth networks, but constantly updated itself by connecting to its back-end systems via the network.
Product-market fit may well be the hardest things to nail down for an early startup, and also one of the key steps paving the way to successful venture funding. ByteDance hit its stride with a clear platform strategy, an excellent founding team, and an intense focus and discipline that’s the hallmark of all mega-successful consumer internet platforms we see around us today.
CHOOSING A BOLD UNDERLYING TECHNOLOGY & HUNTING THE BEST TALENT FOR ITS FOUNDING TEAM
ByteDance founder Zhang Yiming’s genius was in recognizing very early the shift in the way information was distributed and consumed in the newly emerged mobile medium. He made the bold choice to move away from human-centered content selection and distribution, to a platform almost entirely enabled by artificial intelligence based on machine learning on user-generated big data. He was convinced about the big shift taking place from people searching for content to “content searching for people”.
In August 2012, five months after founding ByteDance, Zhang launched his first mobile app, Toutiao or Today’s Headlines, an AI-powered daily curated feed of news content personalized to users. In 2016, Zhang added to his product lineup by introducing a video sharing app, Douyin, for the Chinese market. He rolled out an overseas equivalent of the Douyin video app, dubbed TikTok, in 2017. That same year, ByteDance paid an estimated $900 million to acquire Musical.ly, a social video app based in Shanghai with more than 200 million users worldwide and a large following in the U.S. The deal combined TikTok’s AI‑fed streams and monetization track record with Musical.ly’s product innovation and grasp of users’ needs and tastes in the West.
Zhang’s other genius was in recognizing that interest graphs that could be built around news consumption patterns could also be repurposed for short videos. The user would typically swipe or tap the screen multiple times per minute, and each interaction revealed a little more about the user and their content preferences that could be used to further enrich their interest graphs. Long form videos (as served up by Netflixes and Tencents) provided much sparser data. People would watch a 45-minute episode of a drama without touching the screen even once.
SCRAPPY GROWTH-HACKING COMBINED WITH A SERIOUS QUEST FOR MONETIZATION
TikTok stepped out to play in market that while vast did not have a player that had successfully built a monetizable business. It was a long-tail end market with many small niche players with fierce competition, and where most of the spoils were divided up by the big players. No investor in their right mind would be willing to fund yet another entrant in a seemingly over-served market. It was a “red-ocean”.
One of our frameworks for subscription business-model strategies revolves around the “R-E-M Framework”. This forms the fundamental basis for funnel-management and the continuum that connects success with garnering eyeballs to success on the financial bottom-line. It begins with Reach that feeds into Engagement, and finally leads to successful Monetization. Each element in the chain is equal-weighted for criticality to business success, and must follow in that sequence.
Maximizing reach via pre-installs. In its early years ByteDance acquired tens of millions of users using cheap Android devices, through aggressive pre-installs of its apps on new phones by influencing shipments leaving distribution warehouses, as well as retail store managers.
Leveraging its past pedigree in recommending items into news feeds. It recognized the potential for short-videos to tap into a goldmine of user data on genre preferences, community preferences, even pointed to existence of teen-subcultures to allow better content segmentation. ByteDance’s system centers around three profiles – the content profile, the user-profile and the environmental profile.
Nurturing an active community of users by seeding good content that would draw user engagement by signing on key influencers and professional content creators, often with very lucrative cash subsidies.
Finally, it enabled average teen users by providing the right tools to enhance experience creating and editing short videos by the average teen mobile user (full-screen high definition, music, filters and special effects supported by newly emerged image-recognition and computer vision capabilities – something that TikTok’s western counterparts failed miserably in doing. Teens typically don’t have photography or editing skills or lifestyle necessary to be an Instagram influencer. Nor thee writing styles to
build a Twitter account. However they have the dancing and lip-synching skills, ByteDance’s tools and filters helped enhance their presentations.
INTENSE PLATFORM-BUILDING DISCIPLINE
The twin flywheel approach that ByteDance leverages for its growth is a rare one not available to all players in the business. Its early decision to build an industry leading recommendation engine played a major part. TikTok’s video classification system is able to accurately identify and classify all kind of subcultures and content automatically. TikTok uses the app’s algorithms to decide which videos to show users, dictates their feed entirely, and learns their preferences the more one uses it. Recommendations create a virtuous cycle of continuous improvement through a data-network effect leading to better engagement and more time spent on the app, further enriching the user’s profile
for the recommendation algorithm to get more accurate predictions. AI‑powered apps at ByteDance go to an extreme not common yet in the West.
ByteDance’s path to success also bucked the prevalent industry logic – once a startup reaches scale in the Chinese internet ecosystem it must accept investment from the BAT triad of industry giants, or risk getting crushed by competition.
THE POLITICAL STORMS SWIRLING AROUND TIK TOK
There are a few key differences between US govt. actions against WeChat versus that against ByteDance. The WeChat ban is more personal, while TikTok ban is more commercial in nature. WeChat has generally low relevance to the US, except to Chinese diaspora in the country who use it not only to stay connected with family in China. While WeChat is primarily seen as a messaging App here, it is a much bigger super-app platform that combines payment-processing along with many other consumer services that Chinese use in their day to day life.
ByteDance’s massive artificial intelligence driven recommendation capabilities and its ability to suck massive amounts of private user data opens up several interesting questions that include –
How will it use American user data in future?
Where does it stand as a source of the usual security threats such as Trojan hacking, or its data privacy due to standard invasive data collection?
Another risk area identified by experts point to TikTok’s individualized micro-targeted personal feed that makes it a slick vector for disinformation and influence.
Given that, how does the Chinese Communist Party seeks to use the platform to influence American population? Something that recently brought TikTok under the purview of US National Intelligence Law.
How deep & extensive are its data collection practices? It’s already known to collect location-data (GPS), IP addresses, Device Access Control/ MAC
addresses, data on user contacts, stored images, keystrokes rhythms and patterns, and several other pieces of data that qualify as Personally
Identifiable Information.
What groundwork is China laying for future use of American user-data? It is known that the Chinese govt. is actively deploying integrating biometric data and
other data in the controversial Uighur region of Xinjiang. Given our knowledge of how powerful are today’s surveillance tools, it’s easy to speculate how governments could leverage this data to control citizens.
Tools used by US govt. fall under the purview of CIFUS, a law designed to restrict US private investment in foreign companies. Protection of sensitive personal data is now seen as a national security tool, and made a case for the recent divestiture order against ByteDance. Additionally, questions remain around the use of another policy tool IEPA, for broader economic sanctions related to policy, economic and national security concerns, and which was invoked during earlier discussions around banning WeChat.
TikTok’s inroads in India and its young, mobile-savvy population is a big reason it’s soaring. About one-quarter of TikTok’s downloads came from India before it was unwittingly caught in a political controversy involving India-China military border confrontations and a resultant backlash against all things Chinese in the country (including about 43 highly popular mobile apps built in China).
Experts mention the TikTok platform also has censored content around Tiananmen Square, Tibetan independence, and the Hong Kong protests, captured and shared by young activists. Given its very younger skewing user-base, it stands to reason that TikTok could not lend itself to applying authoritarian principles for its content and user management.