Topics Pioneers Insight Method Author
Back

From a 3.9% Share of User Time to “ByteDance Time”: How Did We Get Here?

·屠龙之术

Rendering the long image…

Topics Commercial Shifts in Mobile Recommendation Algorithms Personal AI Assistants Adoption and Monetization Crisis PR for Tech and Finance Giants
Pioneers 嘉雪

Summary

  • “ByteDance Time” publicly names a long-term shift in the competitive landscape, and the fierce pushback only made the label stick faster. Zhuang Minghao recalls that in June 2017, Tencent’s ecosystem accounted for 54% of user time, versus 3.9% for Toutiao’s. The shares kept moving in opposite directions; by a strict time-spent measure, ByteDance may have entered its “time” era three to five years ago. Both guests say the article’s argument was hardly new. What made it sensitive was an authoritative media figure putting a label on it: “Some things cannot be put on the scales.”
  • The second half of the mobile internet was about shifting from acquiring new users to dividing up a finite pool of time, then turning that time into revenue. Wang Xing introduced the idea of a “second half” in 2016, a decade before this discussion. Jiaxue cites data showing that in February, Hongguo users spent an average of 125 minutes a day on the app. Zhuang says two hours is already considered an excellent level for a platform; users’ screen time is near its ceiling. “This has been a zero-sum game for a long time.”
  • ByteDance’s expansion has meant carrying the capabilities proven out in recommendation algorithms and short video into one new content category after another. Zhuang sums up the past 15 years as “short video devouring everything,” then traces the “app factory” model through free novels, Soda Music, Hongguo and TikTok. He relays an industry view that in its first three years, Hongguo’s user and time-spent growth was even faster than Douyin’s had been, making it a “Plus version of the time furnace.” AI mini-dramas are accelerating the trend further.
  • Control over attention gives platforms power, but antitrust enforcement still depends on market definition, specific conduct and due process. Zhuang points to the blocked Douyu-Huya merger: the concentration picture looks entirely different if the market is defined as game streaming versus streaming as a whole. Jiaxue stresses that a market share above 50% does not automatically warrant a penalty. The 18.228 billion yuan fine against Alibaba discussed on the show was for abusing market dominance to impose “choose one of two.”
  • Since 2021, large platforms have shifted from worshipping scale to taking risks seriously, while the familiar startup-to-IPO story has changed. Zhuang says companies now have to address issues that matter increasingly alongside growth, users and revenue; the divide between “two worlds” shows up in both public messaging and business decisions. Xiaoyuzhou is his case in point: investors, the platform and its users no longer share the expectation that it will grow from 10 million DAU to 100 million, raise more funding and list in the US. A small, stable business is now an accepted outcome.
  • As of this discussion, AI is doing far more to strengthen incumbents than to weaken them, but the shape of a new traffic gateway remains unresolved. Zhuang says the pattern is especially pronounced in China: the strong keep getting stronger. ByteDance’s second AI growth curve could even become larger than its already substantial first curve, and grow faster. At the same time, whether a personal agent can rebuild the traffic gateway remains entangled with habits inherited from the mobile internet era. “Today, you can only choose what you believe.”
  • The key job of PR is to contain risk; jumping into a public fight can turn routine criticism into a major story. Jiaxue acknowledges the clear comparative slant in Qin Shuo’s article, but still considers the “either stupid or malicious” response inappropriate. Zhuang’s advice is to let the piece fade on its own: “Leave it there, and two days later nobody will remember it.” A platform can reject a label imposed from outside, Jiaxue warns, but that does not make its real impact on people’s lives any smaller.

Deep dive

1. The sensitivity around “ByteDance Time” lies in putting an official stamp on a shift that has already happened

  • Jiaxue recaps that on October 26, 2026, Qin Shuo published “The Fearsome Hongguo and ByteDance Time in China’s Internet,” citing a September Nomura report and using Hongguo’s explosive growth to discuss the changing distribution of attention. Li Liang’s response escalated from “pure nonsense” to “either stupid or malicious,” taking the dispute from data definitions to personal attacks.
  • Jiaxue notes that Li Liang had also wrongly attributed someone else’s article to Qin Shuo. On November 2, Qin said he had submitted materials for a defamation lawsuit; Li Liang then said he would respond in court. The boundaries of platform power and its public responsibilities consequently became part of the debate.
  • Both guests say the article offered little that was new. Zhuang Minghao’s read is that companies do not want an established fact “stamped” by an authoritative figure, but their fierce response drew more media attention and may have backfired: “Some things cannot be put on the scales.”

2. Since “54 to 3.9,” users’ time has kept changing hands

  • Zhuang traces the data to QuestMobile, which supplied the figures cited by Nomura. A time-spent chart first appeared in QuestMobile’s Q2 2018 report and looked back to June 2017: Tencent’s ecosystem had 54%, while ByteDance, then still called “Toutiao’s ecosystem,” had 3.9%. “54 to 3.9” is where he remembers the story starting.
  • The trend since then can be summed up as “Tencent kept declining, ByteDance kept gaining.” He thinks ByteDance may have overtaken Tencent three to five years ago under a strict user-time definition; public recognition of “ByteDance Time” came later.
  • He assesses platforms on three measures: “the number of users, the time users spend, and users’ money.” ByteDance may have surpassed Alibaba and Tencent in revenue years ago. Among Chinese internet companies, his view is that none matches its talent density, access to capital or scale of AI investment.

3. A decade into the second half, competition has shifted from user counts to time and money

  • Zhuang divides the past 15 years into three periods: 2011–2016, 2016–2021 and 2021–2026. The first was the mobile internet boom, with startups and VC investment springing up everywhere and companies “riding hard and fast.” Meituan, Didi and ByteDance—the TMD trio—became symbols of the next generation taking over from the prior wave of giants.
  • In his recollection, China’s internet penetration had probably passed 50% by 2015, making another doubling of the user base impossible. Wang Xing’s “second half” comment in 2016 was both a guide to Meituan’s business and a marker of changing competitive metrics: from growing user numbers to growing time spent.
  • Meituan had heavy offline fulfillment costs and needed to consider ROI. Livestreaming and short video, meanwhile, became important ways to capture users’ time. The next question for platforms was: “They spend an hour, two hours a day with me. What then?” Monetization through ads and livestream commerce moved to the fore.
  • The pandemic provided another extreme test: the infrastructure was already in place, users could not go out, and platforms competed directly for time at home. Zhuang says many of today’s market positions took shape after this “very, very rigorous test of competition for users’ time.”

4. After “short video devours everything,” ByteDance keeps growing new branches

  • Zhuang says Toutiao first proved the central role of algorithms in a news app, then applied them to short video. If he had to pick one phrase for the past 15 years, it would be “short video devouring everything.” Short video plus algorithms became ByteDance’s “master key” for expansion.
  • He points to free novels and Soda Music as extensions of the “app factory” model. Asked whether Soda has overtaken Tencent Music, he first says “maybe” and then walks it back, but is clear that it has surpassed NetEase Cloud Music. He judges the scale of free novels to be far larger than that of paid novels.
  • ByteDance pursued its overseas business seriously from day one, and TikTok has become huge. At home, the new growth area is mini-dramas. Zhuang relays an industry view that in its first three years, Hongguo’s user and time-spent growth curves were even steeper than Douyin’s had been, making it a “Plus version of the time furnace.”

5. Hongguo’s 125 minutes put it near the upper end of the time competition

  • Jiaxue cites February data showing average daily time spent on Hongguo had reached 125 minutes. Zhuang says a platform needs to hold users for an hour before it can call itself “a regional power”; “two hours is everyone’s target,” and beyond that, it is hard to imagine.
  • The constraint is the finite number of hours in a day. Once sleep, work and conversation are accounted for, the time people can give to screens is limited; it has barely grown over the past five or six years. The pandemic even tested the extreme case of ten or more hours a day. “There isn’t much new time to add. Everyone is fighting over the same pool.”
  • Zhuang says a new QuestMobile report released during the dispute showed Hongguo had overtaken Doubao in users to become ByteDance’s third-largest app. When a product only about three years old reaches that scale and time spent so quickly, it naturally prompts the question, “Why so fast?” AI mini-dramas are still “accelerating, accelerating, and accelerating again.”

6. Emphasizing active user choice does not settle the debate over algorithmic responsibility

  • Jiaxue notes that Li Liang’s November 5 response stressed that “user reservations for upcoming shows within the Hongguo app have become the main distribution mechanism.” She understands this as an effort to highlight active choices such as following and reserving shows, and draw attention away from the platform’s recommendation algorithms.
  • Zhuang says whether “algorithms have values” had already been debated repeatedly around 2018. The regulations on algorithmic recommendation management in 2021 and 2022 marked a turning point. But algorithms remain a vast black box; platform accountability and adjustments to their recommendations will continue to rub against outside demands. No one can draw a line that resolves every dispute.
  • He compares it to the Xibei prepared-food controversy: companies cite standards, while the public uses a different yardstick. A content platform will always have both good and bad content. Even if the discussion focuses on a 1% or 1‰ share, the first question is how to define and filter that content—no single standard will convince everyone.

7. The 2021 turn made platforms wary of risks beyond scale

  • Zhuang sees platform-economy regulation, antitrust, Didi’s IPO and the Ant-related events as part of the same turning point: hard tech gradually became the dominant narrative, while the public and regulatory climate for platform companies changed. He repeatedly describes the periods before and after 2021 as “two different worlds.” Companies could no longer focus only on growth, users and revenue.
  • Jiaxue adds the legal timeline: China’s Anti-Monopoly Law was passed on August 30, 2007 and took effect on August 1, 2008; the revised law took effect on August 1, 2022. The February 7, 2021 guidelines on antitrust in the platform economy brought “choose one of two,” big-data price discrimination, killer acquisitions and VIE structures into the regulatory frame.

8. Antitrust disputes have to address both market boundaries and conduct

  • Zhuang points to the blocked Douyu-Huya merger. After Panda shut down, the top 2 platforms might have accounted for 99% of the game-streaming market; across the whole streaming market, he estimates their combined share would have been under 10%. Where regulators draw the relevant market boundary directly changes the concentration analysis.
  • On merger review, he cites revenue thresholds of combined annual revenue above 4 billion yuan for the parties, with each party exceeding 800 million yuan, and mentions acquisitions involving more than 51% control. Those revenue thresholds cover a broad swath of today’s platforms; enforcement still depends on the specific market, transaction and process.
  • Jiaxue confirms that a market share above 50% does not automatically lead to a penalty; the conduct also matters, whether it involves a merger, a concentration of undertakings or abuse of dominance. She cites Alibaba’s 18.228 billion yuan fine on April 10, 2021, for imposing “choose one of two.” Zhuang adds that fines are calculated as a percentage of the prior year’s revenue under a defined process.
  • Attention and traffic exposure are difficult to price directly like physical goods, and can be adjusted dynamically. Zhuang says that makes them inherently hard to assess: the public has an intuitive sense of platform power, but regulators still need a legal basis for action. The two do not automatically coincide.

9. From big platforms listing in the US to accepting a stable life as a small platform

  • Zhuang calls the period around June 2021 the “last glow of the Web2 era,” followed by a different world. The shared narrative of “grow the platform, raise money, go public” has disappeared in his view. Xiaohongshu may be the only remaining glimmer: a large platform that is still private and continues to grow.
  • Xiaoyuzhou is a clearer example. Investors, the authorities and users no longer expect it to grow from 10 million DAU to 100 million and then list in the US, as they once did with Zhihu, Weibo and Bilibili. Some platforms can have small teams, stable businesses and “do very well” within their own communities.

10. The debate over AI as a traffic gateway still carries the inertia of the mobile internet

  • Jiaxue asks whether the AI era will have traffic gateways like search and social media. Zhuang says the previous chatbot wave was thought unlikely to produce a good gateway, but personal agents have brought the same question back this time. There is still no answer.
  • He argues that today’s founders, model developers, investors and users all experienced the mobile internet, leaving related “factors and seeds” embedded in them. This path dependence affects not just product features, but also billing, operations, monetization and user habits.
  • Another view is that AI may not go through another round of gateway and traffic competition. The two views are still colliding: “Today, you can only choose what you believe.” A new normal will emerge through experimentation and adjustment; no one will come back from the future with a completely new set of answers.

11. AI is currently more of an amplifier for giants, especially ByteDance

  • Zhuang’s interim judgment is clear: over the past three or four years, AI has done far more to help existing giants than hurt them. The pattern is especially pronounced in China, where the strong keep getting stronger. The US has seen several new giants emerge, while China’s existing giants have clearly become stronger.
  • His counterfactual is that without AI, ByteDance might have remained strong in Douyin, TikTok, ecommerce, livestreaming and advertising. AI has let the tree “grow new fruit”: the early signs are already visible, and the second growth curve could become larger than the massive first one and grow faster.
  • At the same time, he says governments around the world responded to AI early and quickly. But the technology keeps posing new challenges that call both for rapid responses and more room to maneuver. Those demands are inherently in tension with how law and regulation operate.

12. When faced with a critical article, the most effective move may be to let it cool off

  • Jiaxue does not sidestep the article’s slant. In her experience as an editor, the article raised questions about ByteDance while giving another company a “positive assessment at least 95% of the time.” She understands why that comparison might have provoked Li Liang, but still considers “either stupid or malicious” clearly inappropriate.
  • Zhuang observes that companies used to default to private conversations about concerns when problems arose; today, many take the fight straight into public view. His first reaction to this response was “unnecessary,” though outsiders cannot know whether it came from an individual’s emotions, an organizational decision or something else.
  • Jiaxue asks again, “If you were the head of PR?” Only then does he give a direct answer: let the article run and leave it there. There are too many newer, bigger topics competing for attention in the Chinese media cycle. “Leave it there, and two days later nobody will remember the article.” Management may not even question the PR team over it.
  • Li Liang’s later article said his language had been too harsh and that he had been criticized internally, then pivoted to product promotion. The two guests sum it up as “one sentence of apology, three paragraphs of advertising.” Zhuang says that after going on the attack and then having to walk it back, the response only reinforced the impression that the article had been stamped as true.

13. Corporate temperament shapes crisis response; PR has to smooth out the swings

  • Zhuang increasingly believes companies have personalities just like people do, and forceful founders are especially likely to shape how an organization speaks. Traffic can lift an entrepreneur up, then turn back on them. PR’s job is to make the peaks “a little flatter” and reduce the sharp swings.
  • Jiaxue offers one explanation: big companies are used to hitting KPIs and solving problems immediately, and may carry that internal impatience into public messaging. When emotions run high, “actions get distorted”; people meant to defuse a crisis can end up setting it off.
  • On whether negative coverage is a competitive attack, Zhuang admits, “It’s too hard to tell.” Some people “come with a knife,” some join in for the spectacle, and some are sincerely trying to help. Add algorithms and online emotions, and it is like “milk mixed with water.” Even when you know someone or have worked with them, it can be hard to tell friend from foe.

14. Business decisions in the AI era now intersect with regulation, supply chains and geopolitics

  • Asked how to build fair-competition constraints into everyday decisions, Zhuang starts with the changing environment: since 2024, there may have been some easing—“of course, not much, but some.” Companies still have to run their businesses, earn revenue, preserve jobs and pay taxes, and all sides need to keep adjusting.
  • He cites ByteDance’s “reportedly several hundred billion yuan” in compute investment to illustrate the complexity, including chip purchases. Policy favors domestic chips, but a company cannot use only domestic chips; overseas chips require approval, and coordination may span the Cyberspace Administration, MIIT and NDRC before becoming part of US-China competition.
  • Companies have to comply with laws and policies while keeping up in the competition, especially in this AI cycle. Zhuang says the pressures and problems facing CEOs today—whether at large platforms or startups in hot sectors—are far more complex than they were a decade ago.

15. Opportunities for new companies remain unevenly distributed; bubble calls are for the history books

  • Asked which types of AI companies have the best prospects, Zhuang’s first answer is, “Who knows?” The model landscape has taken shape for now, with a few startups and giants. Next come video models, world models, robotics and embodied AI, as well as commercial space and semiconductors—but each is at a different stage.
  • He cites Jensen Huang’s “five-layer cake”: every layer is needed, but they differ in development sequence and gross margin. The chaos will gradually sort into layers and produce competitive companies. As for whether today’s market is a bubble, he refuses to give a firm answer: “Bubbles are conclusions drawn by history books.” Nobody knows while they are living through one.