AI-powered news recommendation app Jinri Toutiao is as sticky as apps get. Plus the user data it generates could be far richer than the likes of social data harvested by WeChat. It has made its parent company Bytedance a major player in China’s tech scene. According to Liu Zhen, Bytedance senior vice-president for corporate development, the company is going to continue to grow aggressively at home and abroad, following its recent acquisitions of News Republic and Musical.ly. Speaking at a briefing to journalists during Bytedance’s Global Festival for A.Ideas in Beijing, the first time Bytedance has spoken to a group of international journalists, Liu revealed:

  • Revenue for 2018 is expected to be around RMB 50 billion
  • 50% of time spent on Toutiao is on watching videos
  • 50% of revenue could come from overseas in five years’ time
  • A million content creators produce 20 million pieces of new content per day
  • 90% of content is moderated via AI, the rest by humans

Read more: See our report on Toutiao’s upcoming AI advances

Bytedance was started in 2012, after its founder Zhang Yiming realized when commuting that there were ever fewer news kiosks and that people were spending more time on their phones. Just five years later, Bytedance has over 200 million daily active users (DAU) across its apps and the flagship Jinri Toutiao sees its users spending 74 minutes a day on the app. “That’s probably the longest in terms of time spent on content platforms [in China],” said Liu.

“Traditionally we had a lot of OGC providersorganization generated content–but nowadays we see increasing numbers of PGC and UGC creators [professionally- and user-generated content] and now Toutiao has about a million what we call OGC/PUGC creators, with about 20 million new creations every day and about 90% of those are created by PGC an UGC,” said Liu, adding that users prefer these to longer reads by traditional media.

Distribution and long tail content are core strengths of the app. Its AI allows an efficiency of data handling that makes it easy to push obscure content to users. Content generators are incentivized by the platform, taking a share of advertising revenue. She would not divulge how much creators get for any specific metric.

Advertising is so central to the business model that the company “considers advertising as another form of content” via personalization.

Acquisitions are also vital to the company’s growth, though Liu said 10% of the company’s efforts went into acquisitions and 90% into improving the apps. Bytedance is pushing its apps–Toutiao is known as Topbuzz outside China–into Japan, Korea, Southeast Asia, Brazil and North America. Though in more mature markets they find it “easier to leverage existing platforms” by buying them. Topbuzz is doing particularly well in Japan, Tik Tok is proving a hit in Thailand.

“We realize that in mature markets where you have very high smartphone penetration, the IT infrastructure is already there and you have mature creator communities–there are synergies with those companies which have a very good brand, very good content, very good creators and follower [numbers]. What they’re lacking is a more efficient way to distribute their content to reach the audience. We could use the recommendation engine we have… They have the region coverage we don’t have.”

Bytedance has around a dozen platforms, many of which are video-based content. Internationally, video is vital. “For short video type products, it’s easier to make that a global platform,” said Liu. Speaking about the integration with recently-acquired US short video platform Musical.ly, Liu said,  “We share a vision of building a global video platform” for providing content access to the China market, and giving Chinese users access to overseas influencers and creators. The platforms will probably remain distinct as it is difficult to find success with apps that offer combinations of services such as news, messaging, microblogging, due to cultural differences and even language.

Liu would not be drawn on profitability for the company but stated that the business is “very healthy and capitalized:” “We will continue to aggressively grow. By acquisition or expanding into new markets”

The company has a range of priorities at home and abroad. Domestically, Toutiao will continue to work on advertising efficiency. Another priority there is driving growth for UGC short videos such as Douyin, Huoshan, Duanzi which have over 20 million DAUs and growing. Overseas growth is expected from Musically and Flipagram: “I believe Musical.ly is going to be a very strong brand and will be a strong focus for overseas expansion strategy,” said Liu.

The company reckons the ad market in China is huge and will continue to grow–“In China there’s still lots of potential space for us to continue our growth revenue-wise and user-wise”–yet in five years’ time half of revenue is expected to come from outside China, though argued it is very hard to plan anything beyond six months ahead given the pace of China’s tech scene. When pushed on revenue predictions for 2018, Liu acknowledged that RMB 50 billion is about right, though in future “Mature parts will be profitable, new parts will need more capital”.

Liu was not concerned about the hit app’s structure being copied before they have chance to expand worldwide. She stated they were building up Musical.ly and have 2,000 engineers and product engineers, plus five years’ experience moderating and recommending content. She was confident the company’s technology advantage, skills at monetization, driving growth and user acquisitions will help it grow internationally.

Back at home, competition could be fiercest. “Everybody in China is concerned about Tencent,” said Liu, “We all think more about how we co-exist.” However, one advantage Toutiao has over WeChat is the data it collects, both qualitative and quantitative. Toutiao gets gets 74 minutes a day of user data, on the user’s core interests. Compare that to a social platform. On a platform such as WeChat, user data is divided among different activities: “the data that you’re able to retrieve is less than from the reading data,” said Liu. WeChat knows a lot about its users, but not all of that data is useful in terms of targeting and servicing them better. News articles that friends recommend on WeChat might not be what you’re interested in.

There are dangers that the algorithms that have made the app so successful (and, let’s admit it, addictive) could over amplify certain types of content at the expense of others or even achieve a race to the bottom as clicks are rewarded. Liu said that the company wants to train its algorithm to be more like human beings and be able to push content that goes beyond just reflecting the interests it has got you pinned down to, but offer relevant general interest content to keep users interested.

Given the strict regulatory conditions Toutiao is working in, officialdom also has to be taken into account, but apparently this also has its positives according to Liu:

“We shouldn’t only purely focus on technology, but focus on social responsibilities and regulations and policies and take all those factors to train the algorithm to make the content better. We have data showing that the healthier the content is, the longer people tend to spend more time in the long term.”