Killing Toutiao

Author Wang Jianfei / Edited by M.E. Strickland

Here and there around Beijing, one can spot bus stop billboards emblazoned with red and white ads for a news aggregation app called Jinri Toutiao—literally, “Today’s Headlines.” There’s just an air of irony in that, perhaps: such a low tech and traditional form of advertisement for a company that employs more than 800 machine learning algorithm engineers.

By most measures, Toutiao is a breakout success. At just a few years old, it already holds a daunting level of influence in China’s online media. Users spend, on average, upwards of an hour a day in Toutiao’s app, and its core of monthly active users numbers 160 million, or roughly equal to half the total population of the US.

Last year, driven by the power of its AI and machine learning, Toutiao had reached a level of prominence in the content market such that it could no longer be ignored. China’s three largest tech giants, Baidu, Alibaba, and Tencent, have all taken notice, and now, perhaps for the first time ever, they are in agreement on an issue: Toutiao must die.

Perils of Success: Baidu vs. Toutiao

Within the last two years or so, an odd shift has occurred in China’s content market: where big companies had previously been loathe to ever pay content creators, they are now throwing money in every direction, and almost every platform is trying to paint itself as a source of “content” and “knowledge.” News outlets that had for years been progressively marginalized have now come roaring back, with funding surging and new products being launched at a heady pace.

Even Baidu, often rebuked for being slow to catch on, announced this year a new strategic center for content distribution. But this revival of content also signals the end of the web portal era. Behind the new “content distribution” fight is yet another, a fight for the online advertising market.

When Toutiao was born in 2013, it wasn’t the only personalized news service out there. In early 2010, a company called Wumi Web put out a personalized reader app, but with a downward slide in the business, the company decided to close its news service in 2013 and pivot to an anonymous social networking service instead. Yet at virtually the same time, Toutiao managed to grow by 10 fold, from one million users at the start of the year to 10 million by the end of it.

Just three years later, Toutiao had 67 million users, putting it second only to Tencent News, and its per user average usage time reached a formidable 71 minutes a day. It was absorbing a massive amount of users’ time and attention, even surpassing the more familiar “attention blackhole” of video apps and sites.

And since users can’t pull themselves away from Toutiao’s app, this “container” for the news has come to exert a fatal attraction over advertisers.

For Baidu, whose search engine has for years been the central channel through which people have accessed the news, Toutiao poses an almost existential threat.

In the PC era, Baidu’s income came mainly from advertising, but with the rise of the mobile web, the company has been faced with three life-altering questions: how much traffic can it now claim, who should it share that traffic with, and where will its new traffic come from?

Baidu has lost some of its sway. According to the China Internet Network Information Center’s 39th internet development report, Baidu’s search app is among the top five apps used in China, with a usage frequency of 15.3%, putting it higher than Alipay (Alibaba’s payments app), but far below WeChat, QQ, and Taobao. And as mobile data analysis firm TalkingData’s 2016 report shows, Baidu’s app coverage is at 24.68%, below even Alipay. Meanwhile, the statistics for Baidu’s news app are pretty much negligible.

With these anxieties weighing on the company, Baidu has chosen to bet its future on AI. And one of the manifestations of that investment is Baidu’s newsfeed.

In December of last year, Baidu updated its news client with a chatbot in imitation of Quartz. In the Baidu News “dialogue,” the chatbot would work out current news topics the user might be interested in and fit them into the conversation.

According to one of Baidu News’s former developers, the team behind the app had at one point more than a hundred members.

Thus when it comes to AI, Toutiao is unavoidably engaged in a head-to-head fight with one of China’s biggest, and most advantaged, companies. But Baidu’s chatbot suffers from the same lack of sophistication as all others. If you converse with it long enough, its dialogue becomes simply a breakdown of articles. And if you attempt to guide the dialogue yourself, its responses quickly become unsatisfying. It’s hard to see what revolutionary experience this interaction really brings to users.

What Baidu wants is a newsfeed powered by AI, and housed within—at least for the near term—Baidu’s search and news apps, and in the longer term perhaps in some form of smart hardware. It doesn’t just feed information to the user, though, but collects it from them, harvesting data on what users want to read or buy. This is the only way to solve Baidu’s problems of where to obtain traffic and where to deliver it.

From its early days, Toutiao gave itself two key labels: that it was “without editors” and was instead “all engineers.” But its technical reliance is also what has earned it a reputation for “crass content” among many. Its algorithms are a sort of magic mirror, reflecting the most secret, and base, desires.

But the problems brought on by algorithms will have to be solved by algorithms. Apart from using data and algorithms to automatically filter “vulgar” content, Toutiao is using AI to assist writers, such as in giving prompts to content creators: for instance, flagging headlines that are likely to be marked off as clickbait. In fighting plagiarism, Toutiao checks articles against the entire web, and if there is a copyright complaint and the AI judges an article to be dubious, it will refrain from posting it.

Toutiao’s AI ambitions are only getting larger. In the summer of 2016, Toutiao published its own “Little Ming Bot” (named after the company’s founder, Zhang Yiming) to write Olympics-related news stories faster than any human could.

Despite Baidu’s wealth of AI tech and talent, it has lagged behind Alibaba and Tencent the last few years in its ability to convert tech into functioning products; that, in short, is the source of most the company’s worries. To put it another way, Toutiao is not the source of all of Baidu’s issues, but its rise is exacerbating them.

Burning Cash: Tencent vs. Toutiao

In the past couple years, Tencent, who has been sitting so comfortably as the leader in online content distribution, also started taking note of Toutiao as a potential threat. And in contrast with both Toutiao and Baidu, Tencent came to AI relatively late, but it has protected itself with the inherent advantages of its products and content.

From reported data, even with Toutiao’s continuous, high speed growth of the last three years, its coverage and active user rates still don’t match the combination of Tencent News and Kuaibao; the former is Tencent’s basic news service, the latter its personalized news service.

Of these two services, one surpasses Toutiao in number of daily active users and ranks number one for news apps, while the other is ranked third, just behind Toutiao. Furthermore, Tencent owns WeChat, another “attention blackhole” with a seemingly unbreakable grip on Chinese internet users’ time.

Incidentally, the relationship between Tencent’s pair of news services is not unlike the relationship between its QQ and WeChat apps, in that, while they seem to cover the same functions, their roles are complementary rather than redundant. Users interested in news from more authoritative sources turn to Tencent News, while those looking for more niche and personalized recommendations opt for Kuaibao.

In this race for content and user attention, Tencent has chosen an approach that suits it well: burning through cash.

A given user might hate WeChat yet like QQ, just as they might hate Tencent News but like Kuaibao. And yet so long as they stick with one of the two, and don’t turn to Weibo or Toutiao, they remain Tencent’s “friends.” That logic extends beyond users to creators. Tencent last year spent 200 million yuan on its “seed plan” (mangzhong jihua), and this year has expanded the budget for its “seed plan 2.0” sixfold to 1.2 billion yuan, hoping to draw even more creators into its content ecosystem with the lure of hard cash.

In the past, China’s internet grew along with bootlegging and piracy, and content was simply an efficient means for winning traffic. By turns, content generators pushed the development of forums, internet portals, Weibo, and then WeChat, even though they never really made money directly. And yet content creators helped one company after another grow their fortunes. So when internet companies started to share their profits with content creators, no one could object to this long delayed compensation.

Toutiao has made some efforts in this field itself, promoting a series of projects to incubate grassroots media. One of these supported 1000 content creators with a minimum stipend of 10,000 yuan. Another, “Creation Space,” had higher demands on quality but offered higher compensation as well. But these two programs did not achieve the results that Toutiao was likely hoping for, perhaps because, in the former case, the quality of the output was lacking, while in the latter, creators with more in-demand content were not willing to bet their futures on a single platform.

The scarcity of quality content is perhaps one reason that Tencent has offered higher compensation to creators than Toutiao.

Tencent has shown a predilection for quality content with its investments in Zhihu and other platforms, but this isn’t enough. No amount of subsidies and incubator programs can give a company exclusive control over the best grassroots media.

Meanwhile, the subsidy payments have created their own problem, with some gaming money out of the system. One video creator, who wished to remain anonymous, shared with us one simple cheat, saying he would edit together clips of YouTube videos and then post them through Kuaibao, Toutiao, and the like. The result is low quality content with possible copyright violations, but it plays off the distribution channels’ algorithms and generates payouts for the posters.

The person in question works at an internet finance company, using the videos as a supplement to his income. After three months with his account online, his videos collected six million views, ginning up more than 4000 yuan per month.

Another user reported that posting recycled news and “spam commentary articles” could easily gain upwards of 10,000 yuan a month, with some more vigorous posters making as much as 50,000.

After years of “exploitation” at the hands of big internet companies, content creators are waking up to their own power. As ride-hailing, delivery, and other on-demand services have shown, mere compensation is never a longterm solution. At some point, the people doing the actual on-the-ground work demand something more.

Tencent may therefore be looking for a better approach, one that will keep it at the top of the heap. But does such an approach exist? And where does it lie?

A Matter of Values: UC vs. Toutiao

If the objective for aggregators is “clearing out lowbrow content and improving quality,” then there are two possible strategies: money or technology. Which one a company chooses reflects something deeper about its values.

And when neither algorithms nor cash can quickly shift the quality of content, some companies choose to go back to the more traditional model for generating quality content. Last year, after a bruising fight with Toutiao over claims about who was first in market share, another news media startup, Yidian News, brought on Sina’s chief editor Chen Tong as their own. Having come up from the first generation of internet companies, Chen Tong worked to inject a sense of “editorial discretion” into the recommendation algorithms, i.e. a human touch to compensate for the foibles of machine learning. When Yidian News’s CEO, Li Ya, was interviewed by Sina, he didn’t mince words: he brought in Chen Tong, he said, because machine algorithms are flawed.

Although Toutiao has human editors, it’s moving in the opposite direction, with CEO Zhang Yiming arguing that “algorithms don’t have values.” (This, however, is arguably itself just another value perspective, laying idealistic expectations on technology). With enough filtering and machine learning, there is every reason to believe that someday Toutiao’s algorithms may surpass human editors. But in the short term, that idealism isn’t bearing results.

Toutiao isn’t the only company with an unabashed faith in algorithms, though. UC, better known for its browser, has also started a “UC Toutiao” news service (the similarity in names is a bit unfortunate). The app itself shows off a level of slick interaction design such as few Chinese companies even attempt. But below the interface, the content it displays is, sadly, as lacking as one might fear, riddled with clickbait headlines and undeserved exclamation points.

UC Toutiao is an entry point for UC’s other services, and UC, as it happens, is a content distribution division owned by Alibaba. UC Toutiao only became an independent app separate from UC’s browser last year, and so its relationship with parent Alibaba is paradoxically at once both close and distant. UC’s marketing manager Zhou Mo says that the reason the app was so well designed, from a user experience standpoint, is that UC treated it like a startup project, contributing the company’s brightest young workers to its team.

As to the clickbait problem, UC Toutiao has been trying from the start to find technical means of dealing with it. The team were a bit startled when they first encountered some mockery online of what users called the “UC Shock Department,” deriding it for the abundance of overwrought headlines.

“Of course, no matter the reason, this is definitely the wrong way to go about making content,” Zhou Mo says.

UC had previously introduced a higher-level commenting feature to help with flagging lower quality content and, while not removing it completely, keep it “below the fold” to avoid misleading readers.

It’s worth noting that this effort to combat clickbait and other low quality content is yet another case where social values are making themselves felt in business strategies. But UC’s approach here is different from Toutiao’s reliance on algorithms and Yidian News’s use of human editors, as it has been trying to understand the causes behind clickbait and sift it for whatever parts might have value.

The relative leniency with which people have treated UC’s “Shock Department” is in contrast to the harsh judgement held out for Toutiao’s own “vulgar” content. But this is a consequence of the way that users respond to a company’s purported values. For Toutiao, with its emphasis on technical brilliance, expectations are more exacting, and users’ responses to Toutiao’s performance in this regard shows something of the suspicion around machine algorithms. On the other hand, users have been more forgiving of UC, perhaps, because of preexisting positive feelings for the brand.

The UC team’s history reflects the changes in mobile web browsing, from direct website visits to the rise of search to personalized reading. UC has been a victor in the first two periods, but in the latest stage it has faced stiff competition; hence, the develoment of UC Toutiao.

Users for UC Toutiao, as opposed to the rest of UC’s services, skew younger, with a more even gender ratio. They also shop more online, are more likely to interact with others, and more focused on online, digital, and (strangely) military and economic news. Those are the traits that UC hopes to leverage for conversion. For parent company Alibaba, it’s a bonus.

Unlike Baidu and Tencent, Alibaba has less to worry about, as Toutiao isn’t a competitor with its core ecommerce business, nor is content distribution likely to affect its growth one way or another. Generic content distribution is something that Alibaba has long been attempting to wade into, but it has never truly done so. For Alibaba, then, Toutiao’s clashes with Baidu and Tencent seem far removed.

After years of failing to enter the social media space, Alibaba isn’t likely to casually try again to enter content distribution, but fortunately it has UC and Sina Weibo to rely on.

Toutiao’s Break Out

Against the apprehensions and reactive maneuverings of the Big Three, Toutiao gives off an air of confidence.

Whether it’s in Zhang Yiming’s statements to the media or in private talks with the company’s employees, Toutiao seems upbeat. And while there is internal concern about being ganged up on by the Big Three, it’s not as much as outsiders might imagine.

Toutiao’s confidence shows through in two places: first, it’s growth has been excellent, and second, the data shows that only Tencent is in real competition with it. Toutiao is second only to Tencent in the number of active users, with more than 600 million as of last year (some data would put it upwards of a billion), so the data would seem to vindicate its chosen path.

From a micro perspective, Toutiao doesn’t have Baidu’s structural pressures. In conversation, Toutiao’s employees’ personal stresses don’t come from doubts about the company’s future, but about specific operations: is a given channel not reaching its expected readership goal? Is a technical problem making a feature temporarily unavailable? These are the ordinary anxieties of a startup, and not existential worries about a company’s future.

But it should be noted that the seemingly “carefree” Toutiao has slowly begun to change, taking actions it wouldn’t have before, such as:

-Inviting on Uber’s former VP Liu Zhen, buying Flipagram, investing in DailyHunt, releasing Topbuzz, and starting overseas services;

-Poaching Zheng Wei, VP at iQiyi, buying video rights, and partnering with MangoTV;

-Partnering with the event “The Singer of China,” releasing a plan to support independent musicians, and generally working to increase its coverage of the music industry.

On the surface it seems that this company, which has pinned everything on its tech, is finally starting to shore up the shortcomings in its business model. But behind this is concern around Toutiao’s eggs all being in one basket: one model, one set of operations, one product. Thus far, Toutiao has not diversified to anything beyond being a personalized newsfeed service, and is far from having the kind of fleshed out ecosystems of the Big Three, leaving it less secure than its peers Meituan and Didi.

In the last 20 years, internet traffic has flowed where the content is, and no one can guarantee personalized news readers will be able to hold people’s attention for long. One day, a new product format will appear, and users will move. But Toutiao doesn’t have the resources to spare to make any kind of transition.

And with its reliance on an advertising sales model, Toutiao is faced with another cause for concern: it’s a rapidly rising company “without any friends.”

In the current mobile web environment, Toutiao is virutally the only major traffic channel that stands independent. It’s not in the camp of any of the Big Three companies, and in theory could work with anyone. But in actuality, this independence hasn’t brought Toutiao partners, only enemies.

Exploring other businesses has only brought forward more potential rivals. If Toutiao adds local and O2O services, it would be heading into a fight with Meituan; if it adds finanical services, it would give Alibaba more reason to push against it. And companies that lived through the last tech bubble, and who still hold to their proven business models, are more or less all in fixed relationships with Baidu, Alibaba, Tencent, Meituan, or Didi. Under these circumstances, Toutiao can only spread its wings cautiously, lest it provoke its Goliathan rivals even further.

Yet it’s already had a run in with Sina Weibo on the latter’s home turf of social media. Weibo was an early investor in Toutiao, but when the latter made moves into social media features, Weibo announced a partnership with Yidian News and Phoenix Media to counter it.

This should be no surprise, though: the two companies are too similar. In 2014 when Toutiao accepted Weibo’s investment, Zhang Yiming cited the “matching tone” of the two companies. But after two years, the honeymoon was over, and that friendship was broken by business concerns.

Being isolated to such a degree is the second source of concern for Toutiao.

These dual anxieties can’t be solved by AI. To get a grip on its growth, Toutiao will have to turn to non-technical solutions and confront the pressures that come from independence.

Despite the fact that Toutiao has long been touting the percentage of engineers that make up the company, in 2017 its hiring became more “traditional.” Based on data from hiring platform, in two recent months Toutiao was hiring for 501 positions, of which technical positions counted for less than 90, while there were up to 199 listings for marketing and sales. With a combination of more sales personnel and the company’s single profit model, we should expect more advertisement coverage to come. Whether or not Toutiao can balance ads with user experience is yet to be seen.

And there is a third anxiety for Toutiao: survival and conversion. Between the three leaders of next generation tech companies—Didi, Meituan, and Toutiao—the latter is also the farthest from making money, and yet is the one who most needs to do so.

A strategy of independence is the greatest obstacle for Toutiao, as well as its greatest opportunity. Perhaps every founder prefers for their company to succeed on its own rather than to fold themselves into a larger business. And from a commercial logic standpoint, every company hopes to become the next Baidu, Alibaba, or Tencent—not an appendage of one of them.

Therefore, Toutiao is changing its loner ways, parterning with others to do films, local services, weather, and special sales. It’s even partnered with Jingdong to enrich its feed, and with Taobao for sales. In the latter case, it seems as though Alibaba, who has backed out of social media, isn’t willing to go to all-out battle with Toutiao.

Toutiao is attempting to pull in every company it doesn’t currently have a profit conflict with, turning itself into more of an independent intermediary: users consume content on the platform, spend time there, and then Toutiao takes their measure and directs them towards partner’s products. This was always Baidu’s ambition, but Toutiao now seems even closer to realizing it.

Tencent, Baidu, Alibaba, and plenty of others may yet hope to kill Toutiao, but in the end it may also prove that what doesn’t kill it will only make it stronger.