TikTok Was Once Facebook's Most Honored Customer, Until It Hit 100 Million DAU
TikTok's most important early ad channel was Facebook, with promotion budgets growing over 100% every quarter; Facebook concluded its retention was too poor to keep users, and instead funneled traffic to it.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The first stop overseas wasn't Asia or Africa, it was the US and Japan
The starting point for going global was the domestic ceiling. ByteDance internally estimated that China's news feed market could reach a total DAU of 240 million; assuming winner-takes-all and splitting half the market, the DAU ceiling was about 120 million, and Zhang Yiming read a sense of crisis from that number. When choosing countries he again went against his peers: others did Asia and Africa, he did the US and Japan — "Asia and Africa's conditions are too poor. Even if you get a very large volume it's still too small. You can't wait for the market to get bigger." The US and Japan were hard, but if you held a piece, the share was large. He also didn't just look at DAU to measure market size; he required subordinates to calculate DAU multiplied by ARPU, that is, total daily active users multiplied by how much money a single daily active user can make, to derive the maximum commercial benefit a market could extract. That year ByteDance made internationalization one of its two highest-priority strategies.
— Zhang XiaojunPre-install in China costs 30 cents, acquiring a US user costs 3 dollars
TopBuzz ultimately died on its growth model. An operations executive did the math: in China, working with phone makers on pre-install, a user monetizes 30 cents a day, earning 15 yuan over a 50-day full lifecycle, while pre-install cost only 5 yuan, so it was profitable; in the US, a user cost 3 dollars, but over a 30-day full lifecycle could only generate nearly 1 dollar in revenue, so it couldn't recoup the cost. More fundamentally, in the overseas ecosystem both user growth and monetization channels were in the hands of Facebook and Google; the project got users from these two lines and monetized back through these two lines, making the growth model a dead loop. His conclusion: if you want to do content or social products overseas, the ceiling is Facebook — it is a giant crocodile.
— Zhang XiaojunZhang Lidong wanted to cut it, Zhang Yiming wanted to take another look
At a 2018 project review meeting, the big names sat around a square table, with mid-level staff behind them. Zhang Lidong, who normally controlled the purse strings, suddenly spoke up in a hoarse voice: "This project isn't worth doing, it doesn't make money." The room fell silent; no one dared speak. Zhang Yiming took the microphone: "This project still has prospects. We can take another look." After Liu Xinhua left, the product and operations of the internationalization business unit were taken over by Chen Lin and Zhao Tian, the corresponding heads at Jinri Toutiao, and the algorithm technology reported back to their respective middle platforms. In August 2018, the company resolved to cut news internationalization; TopBuzz ended tragically under the two mountains of copyright and growth, ByteDance's first business unit came to an end, and the only surviving fruit, Halo, later did not escape being banned by the Indian government.
— Zhang XiaojunByteDance's decisive weapon is finding a shell for its algorithm
For content products, ByteDance's decisive weapon is the recommendation algorithm — it often finds a product shell, pours in the long-trained algorithm, and adds the systemic combat capabilities of user growth and monetization. These three capabilities — recommendation algorithm, user growth, monetization — were all supported by middle platforms in the past, first validated and accumulated on Jinri Toutiao; but if it had stopped there, the company would have been just another news client. Musically came at exactly the right time: it offered ByteDance a second product shell and massive content fuel, and after merging with the algorithm it exploded with tenacious vitality. So this acquisition was not just icing on the cake — whether for ByteDance's short-video strategy or its globalization strategy, Musically was a must-buy.
— Zhang XiaojunByteDance bid the lowest, yet won the bidding war
Musically had weak algorithm accumulation and flat growth for a full year; in Q3 2016 the two founders flew specially to California to meet Zuckerberg. Zuckerberg saw it as a video version of Instagram and sent Instagram founder Systrom to Shanghai at least three times in person; finally Facebook verbally offered an all-cash acquisition offer of 1.6 billion US dollars; but in the later stages of negotiation, because it was uncertain how high the ceiling for short video really was, Facebook's attitude became hesitant. The second to act was Kuaishou, which, hampered by insufficient cash flow, proposed letting Musically retain independence and having Kuaishou hold over 40% of shares. The last to enter was ByteDance, whose offer was only 1 billion US dollars, lower than Facebook's, but could be cash plus stock; the founding team's three conditions — changing the name, integrating ByteDance's algorithm, and spending at least 1 billion US dollars on marketing — Zhang Yiming agreed to all of them.
— Zhang XiaojunBlack and white T were a horse race, not a heartwarming tale
After the integration, a saying circulated: Musically's icon was mainly white, TikTok's was black, so one was called White T and the other Black T; White T focused on Europe and America, Black T on East Asia, and later a heartwarming tale spread that the two products "intentionally froze and then reconciled." But neither side's original core team bought it: the TikTok team said focusing on East Asia was purely a strategic consideration, "Yiming wouldn't do such a stupid thing"; the Musically team said the teams hadn't merged yet at the time, it felt like a horse race, and the black and white forces were in confrontation. Zhang Yiming drove the organization and aligned information through documents and meetings, and would deliberately set code words in internal documents to block information flow — White T's code name was 1233, Black T's was 1180.
— Zhang XiaojunTreat your biggest rival as your most honored customer
Once integration was complete, TikTok, at a single command, entered a frenzy of spending, spending, crazy spending, and the most important ad channel was none other than its old foe Facebook. From Q3-Q4 2018 to Q1 2019, TikTok's promotion budget grew over 100% every quarter; Facebook's Greater China revenue was about 5 billion US dollars a year, and TikTok at the time planned to contribute about 1.1 billion US dollars over three years. A Chinese person in charge of Facebook's Asia-Pacific sales was thus honored, becoming the global sales champion; at the end of 2018 the company gave him an excess year-end bonus possibly in the millions, and colleagues in Singapore and Hong Kong all said he had gotten lucky. Facebook did not become vigilant, for a simple reason: TikTok's retention was really too poor — "Is this company stupid? The users they buy basically all run away. Day-one retention 30%."
— Zhang XiaojunByteDance used its rival's data tables to reverse-engineer ad spend
At this point, buying a user in the US cost an exaggerated amount of about 10 US dollars, but ByteDance no longer agonized over how long it would take to recoup the cost, and instead shifted to observing ARPU. The target was not designed out of thin air: they poached a group of management from Facebook with high salaries, and one of them brought out Facebook data, with granularity so fine that some had never been disclosed in financial reports, such as how many DAU Facebook had in each country and how much money it could make; based on this, ByteDance made a table covering over a hundred countries, using Facebook's ARPU as the target to reverse-engineer how much money needed to be spent on ads. In the first half of 2019, after completing a beautiful sneak attack, TikTok shifted its ad battlefield to Google and Snapchat, and Facebook's share of the budget slid month by month from 20% to less than 10%. By the time Facebook woke up, it was mid-2019, and TikTok's global DAU had broken 100 million.
— Zhang XiaojunIn their own words · checked verbatim
Asia and Africa's conditions are too poor. Even if you get a very large volume it's still too small. You can't wait for the market to get bigger.
亚非条件太差了 即使做到很大量也太小 等不起市场变大
Zhang Yiming7:08
If you want to do content or social products overseas, the ceiling is Facebook. It is a giant crocodile.
在海外如果想做内容或者社交产品 天花板是Facebook 他就是一个巨鳄
Zhang Xiaojun13:12
But if it had stopped there, this company would have been just another news client.
但若就此停住 这家公司不过是 又一个新闻客户端罢了
Zhang Xiaojun18:17
Zhang Yiming's offer was lower than Facebook's, only 1 billion US dollars, but it could be done as cash plus stock.
张一鸣的报价比Facebook低 只有十亿美元 但是可以以现金加股票的方式进行收购
Zhang Xiaojun25:24
The TikTok team said focusing on East Asia was purely a strategic consideration, "Yiming wouldn't do such a stupid thing." The Musically team said the teams hadn't merged yet at the time, it felt like a horse race, and the black and white forces were in confrontation.
TikTok团队说 主攻东亚是出于纯粹的战略考量 一鸣不会干这种傻的事情 Musically团队说 当时团队还没融合 两边是赛马的感觉 黑白两股势力处于对抗中
Zhang Xiaojun27:29
TikTok's retention is really too poor. Is this company stupid? The users they buy basically all run away. Day-one retention 30%.
TikTok留存率实在太差 这公司是不是傻 买的用户基本都跑了 次日留存30%
Zhang Xiaojun31:34
Figures
| TikTok global monthly active peak | 1.2 billion | 1:01 |
| Douyin China DAU | over 600 million | 1:01 |
| China news feed market DAU ceiling estimate | total size about 240 million, winner-takes-all about 120 million | 7:08 |
| China phone pre-install per-user economic model | daily monetization 30 cents, 50-day lifecycle 15 yuan, pre-install cost 5 yuan | 13:12 |
| US per-user economic model | acquisition cost 3 US dollars, 30-day lifecycle revenue nearly 1 US dollar | 13:12 |
| Facebook's verbal offer for Musically | all-cash 1.6 billion US dollars | 23:22 |
| ByteDance's offer to acquire Musically | 1 billion US dollars, cash plus stock | 25:24 |
| Facebook Greater China annual revenue | about 5 billion US dollars | 30:34 |
| TikTok day-one retention | 30% | 31:34 |
Glossary
- ARPU
- Average revenue per user: how much revenue a daily active user contributes on average; Zhang Yiming used it to measure the maximum commercial benefit a market could extract.
- Lasso
- A short-video app launched by Facebook in Mexico and the US to counter TikTok.
How to listen
Founders and investors making decisions about going global, ad spend and M&A, plus product and growth leads who want to see ByteDance's real globalization decision chain.
The first two chapters on Zhang Yiming's office, personality and Facebook complex can be fast-forwarded.