LinkedIn’s announcement to (mostly) exit China created a lot of headlines this past week.

Why did LinkedIn leave? The one-sentence answer: it's a media company.

In a market like China, where the media environment is tightly controlled, all advertisement-funded social networks are regulated as media companies. Russia has banned LinkedIn since 2016 for more or less the same reason. Countries like Brazil, India, and Nigeria are all working on their own policies based on similar logic to regulate social networks as media companies. The US has yet to modify Section 230 of the Communications Decency Act, so social networks are still not regulated as media properties but neutral platforms.

Interestingly, China did not technically ban LinkedIn. It just made the compliance cost of operating too high to be worth the effort. LinkedIn still plans to keep a small piece of itself in China by launching a recruiting-only app called InJobs, so this is not a wholesale departure.

The larger story worth unpacking here is how different types of networks demand different levels of compliance. The tradeoff between network strengths and compliance cost determines whether a foreign market is worth entering or not.

There are three broad types of networks:

  • Social network
  • Transactional network
  • Collaborative network

LinkedIn occupies social and transactional, but not collaborative. Let’s unpack all three types contextualized in compliance cost to understand why LinkedIn (mostly) left China and, more importantly, what kind of network-based product can still do well in a place like China.

Social Network: Strong Network, High Compliance Cost

Even though Facebook is the ad-funded social network juggernaut, LinkedIn was earlier to this business model. LinkedIn launched its first ad product in 2005, after founding the network in 2003. (Facebook was founded in 2004 and launched its first ad product in 2007.)

A social network has strong network strength, as it mimics, amplifies, and accelerates human-to-human connections digitally. The number and degree of interconnectedness of the users in a network determine its value. This value grows as the network grows over time; this is network effect 101. That’s why the most successful social networks to date are all “free” services, because “free” maximizes user numbers, connections, and longevity on the network. The operation is paid for by distributing media content while selling advertisements alongside that content. The telephone is one of the oldest forms of social network.

LinkedIn may be more known for its Talent Solutions (recruiting) or Sales Navigator (deal transactions) products, but its advertisement product has always been a dependable cash cow -- still commanding one-third of the US’s total B2B display ad revenue.

This media-based business model works beautifully in the American Section 230 world, where media freedom is guaranteed and social networks are not treated as media. (How long this will last is a different topic for another day.) However, in a controlled media environment, this model triggers high compliance costs in the form of content moderation, redundant server infrastructures (so sensitive content is partially removed to comply, but not removed elsewhere to maintain a somewhat consistent network), and investment in legal and government affairs expertise. As the media environment becomes more controlled, this cost increases. Thus, the size and strengths of the social network must also increase to justify the continuous operation.

China’s media environment is quickly becoming more controlled. LinkedIn China’s social media network has not become more valuable nearly as quickly. Thus, as Rui Ma of Tech Buzz China suggested, the rational business decision for LinkedIn is to leave China.

Transactional Network: Weak Network, Low Compliance Cost

However, LinkedIn is not just a social network, it has also built a transactional network. This portion of LinkedIn continues in China, reincarnated as InJobs. Thus, unlike all the headlines, it is more accurate to say LinkedIn mostly left China.

How LinkedIn can continue operating in this limited fashion is worth pondering. A transactional network facilitates discrete, short-term tasks. These tasks may be valuable in isolation, but the overall network strength is weak, because participants only join the network when they need to complete those tasks. (That’s why there’s a strong correlation between the moment a person updates their LinkedIn profile and that person's intention of looking for a new job. As soon as the new job landed, that person tends to disappear from the network.)

Hiring and recruiting is an example of such a transactional network. This task maps directly to LinkedIn’s Talent Solutions product, which surely will serve as the foundation for InJobs. A transactional network may be weak, but so is its associated compliance cost. There is no media to distribute, no content to moderate, and not much data is created, unlike a social network. That’s why LinkedIn’s official announcement clearly stated that InJobs “will not include a social feed or the ability to share posts or articles.”

Whether InJobs will succeed or not is anyone’s guess. But what’s clear is that these weak, transactional networks are allowed to exist in China, even ones built by a foreign company. Their survival will be left, ironically, to the free market.

A relevant precedent is Uber China. Ridesharing is also a transactional network -- getting a ride from point A to point B is a discrete, short-term task, supported by a network of drivers. Uber China lost to Didi, but it lost from a brutally capital-intensive, Darwinian, free market competition. And even after it lost, Uber was allowed to own a piece of the winner in Didi -- 17.7% to be exact.

Collaborative Network: Strong Network, Medium Compliance Cost

The one type of network that LinkedIn does not have is collaborative network. This type belongs to products like Slack, Microsoft Teams, Zoom, Atlassian’s Jira, GitHub, and many other enterprise-focused SaaS products. Some of these have been available in China for years. Others can still enter the market and do well, if they understand the fundamental difference between themselves and LinkedIn, and not get spooked by the headlines. Meanwhile, Chinese tech companies are creating their own collaborative network products to expand beyond China, like DingTalk (from Alibaba), WeChat Enterprise (Tencent), and Lark (ByteDance).

In this day and age, collaborative network products actually have a better chance of becoming successful globally than social network products, including in a highly-controlled media environment, because they are not media. As soon as you get into “media” or are perceived as such, that triggers harsher scrutiny. Content generated from a collaborative network is internal to the users who are collaborating; they are by definition not public and thus not media.

Therefore, a collaborative network’s compliance cost is somewhere in between a social network and a transactional network. It generates much less content and data than a social network, but usually more than a transactional network. The content type is also less varied than an open social network. Even if content moderation is required for compliance purposes, the workload is done in a smaller environment (typically a company’s intranet) with less variety, and thus less costly. Unlike a transactional network, a collaborative network is ongoing, not task-based, so its size, complexity, and value grow over time, more akin to a social network. All three types of networks have to abide by local data governance laws, so the compliance cost difference on this dimension is only predicated on the volume and type of data generated.

Effective collaboration requires common context. Thus, customers of collaboration products typically like to standardize on one solution. Companies that don’t tend to be less effective, less collaborative, and ultimately less successful. A startup that uses WeChat Enterprise for its China team and Slack for its US team, which is not unheard of, would need to deal with more context-switching and communications costs than a startup standardized on just Slack, Teams, or something else.

This “standardization efficiency gain” can be extended to the cloud infrastructure level too. A German car maker who uses Microsoft Azure in its home market would prefer to also use Azure for its China-based, US-based, or Russia-based teams, instead of a local cloud solution. With clear business value, strong network effect, and medium-level compliance cost, the tradeoff to continue operation in less hospitable markets may still be worthwhile.

So should LinkedIn’s (mostly) wholesale departure from China concern all foreign tech companies? Maybe a little, but LinkedIn destiny was written the moment it decided to sell ads alongside media content generated in a social network.

Collaborative network products on the other hand, foreign or domestic made, still have wide lanes to expand into in China’s fledgling enterprise software market. The trick is: don’t accidentally become a media company or be seen as such!

If you like what you've read, please SUBSCRIBE to the Interconnected email list. To read all previous posts, please check out the Archive section. New content will be delivered to your inbox once a week. Follow and interact with me on: Twitter, LinkedIn, Clubhouse (@kevinsxu).




在中国这种媒体环境受到严格控制的市场里,所有由广告资助的社交网络都会被当作媒体来监管。俄罗斯自2016年以来封了领英,原因大致相同。巴西、印度和尼日利亚等国家都在根据类似的逻辑制定自己的政策,将社交网络作为媒体公司来监管。美国还没有修改《Communications Decency Act》第230条,所以社交网络仍没有被视为拥有媒体属性,而被当作中立平台来监管。




  • 社交网络
  • 交易网络
  • 协作网络



尽管Facebook是广告资助的社交网络巨头,但领英开始使用这种商业模式的时间更早。在2003年成立之后,领英就于2005年推出了其第一个广告产品 (Facebook成立于2004年,在2007年推出其第一个广告产品)。

一个社交网络具有强大的网络力量,因为它数字化的模仿、放大和加速人与人之间的联系。网络中用户的数量和相互联系的程度决定了它的价值。价值随着网络的发展和持续而增长;这是网络效应的基础概念。这就是为什么迄今为止最成功的社交网络都是 "免费" 的服务,因为 "免费"可以使用户数量、联系和网络本身的寿命最大化,最长化。运营费用则是通过分发媒体内容,同时在这些内容旁边销售广告来资助的。电话是最老的社交网络形式之一。

领英的其他产品,比如Talent Solutions(招聘)或Sales Navigator (销售交易),也许更加出名,但其广告产品一直是一棵可靠的摇钱树 -- 仍然占据着美国B2B显示广告总收入的三分之一。

这种基于媒体的商业模式在美国版的《Communications Decency Act》第230条监管世界里运作得非常好 -- 媒体自由有保障,而社交网络又不被视为媒体。然而,在一个受控的媒体环境中,这种模式会引发昂贵的合规成本,其形式包括:内容监督和删除、冗余的服务器基础设施(敏感内容可以被部分删除以合规,但不在其他地方删除以保持某种程度的网络一致性),以及对法律和政府事务专业知识的投资。随着媒体环境变得更加受控,这种成本也会增加。因此,社交网络的规模和盈利也必须增加,以证明持续运营的合理性。

中国的媒体环境的受控程度正在迅速加强,领英中国的社交媒体网络却并没有迅速变得更有价值。因此,正如Tech Buzz China的马睿所分析的,领英决定离开中国从商业角度来说是很理性的选择。




雇佣和招聘是这种交易性网络的典范。领英对应的产品就是Talent Solutions,它也肯定会成为InJobs的基础。交易网络可能很弱,但它的相关合规成本也很低。与社交网络不同,没有媒体内容需要散发,没有内容需要调节,也不会创造出多少数据。这就是为什么领英的官方公告明确指出,InJobs "将不包括社交或分享帖子或文章的功能。"


一个相关的先例是优步中国。共享打车也是一个交易网络 -- 从A到B的乘车是一个离散的、短期的任务,由一个司机网络支持。优步中国输给了滴滴,但它输的大环境是场大量资本投入、自由市场竞争的血战。即使它输了,优步最终还是拥有了滴滴的一部分,确切地说是17.7%。


领英没有的一种网络类型是协作网络。这种类型属于像Slack、Microsoft Teams、Zoom、Atlassian的Jira、GitHub以及其他企业服务SaaS软件产品。其中一些已经在中国市场存在多年,其他的仍然可以步入这个市场,并且能做得很好,如果它们深刻了解自己和领英之间的根本区别,并且不被头条吓到的话。同时,中国的科技巨头们也在做自己的协作网络产品,以扩展到海外市场,如钉钉、企业微信和飞书等。

在当今时代,协作网络产品实际上比社交网络产品更有机会在全球范围内获得成功,包括在一个高度控制的媒体环境中,因为它们不是媒体。一旦步入 "媒体" 或被视为媒体,就会引发更严厉的审查监管。协作网络中产生的内容是协作用户的内部内容,不是公开的,因此不是媒体。



这种 "标准化效率增益" 也可以扩展到云基础设施层面。一家在本国市场使用微软Azure的德国汽车制造商更愿意为其在中国、美国或俄罗斯的团队都标准化在Azure上,而不再使用另外一款本地的云。有明确的商业价值、强大的网络效应和中等的合规成本,就算在不太友好的市场中继续运营的折中可能仍然是值得的。