In the first three parts of this “global by nature” series, we went in-depth into the three forces that will power the technology-driven Globalization 2.0: developers (people), APIs (product), open source (paradigm).

This developers-APIs-open source lens is what I use to look at investment and emerging opportunities for the next 10 years and beyond. But there are also opportunities we can evaluate today. In the final post of this series, we will look at three public companies -- Agora, JFrog, MongoDB -- that are illustrative (both good and bad) of this lens and worth watching.

NOTE: nothing in this post should be construed as investment advice. Please do your own research when you make investment decisions.

This lens provides only a qualitative framework and is by no means a checklist of criteria. In fact, you will see in this analysis that some of the companies do not exhibit all three elements of this lens nor do they have to. If you would like to combine this qualitative evaluation with more quantitative metrics, I recommend you check out the Clouded Judgment newsletter by Jamin Ball at Redpoint Ventures. He tracks and updates the common industry metrics of a large basket of software companies, e.g. EV/NTM revenue multiples, dollar net retention, gross/operating margin, etc.

Agora ($API)

Founded in 2012, Agora is an API company that enables developers to build real-time video, audio, and messaging products (so much so that its stock ticker is “API”). Its founder, Tony Zhao, is a veteran technologist of the video and audio space. He was on the same Webex team as Eric Yuan, who founded Zoom. He was the CTO of -- one of the first live streaming platforms to reach scale in China, if not the world. YY is also recently under the scrutiny of the famous short-selling firm, Muddy Waters. (I recommend watching the documentary “People’s Republic of Desire”, if you are curious about YY’s cultural impact.)

Developers: Agora’s developer focus emanates from the top. Zhao left YY to start Agora to build a better network of APIs to enable developers. Instead of building a better Webex -- what Yuan decided to do with Zoom -- Zhao decided to help developers build their own Zoom.

Thus, Agora supports a wide range of platforms (Android, iOS, macOS, Web, Windows, Linux) and frameworks (Electron, Unity, Flutter, React Native, Cocos Creator). This range of support takes a long time to build and a strong commitment to developer empowerment. The range of supported platforms, frameworks, or tools is usually a good signal to see whether a company is truly developer-focused or just paying lip service to a trendy concept.

Some of Agora's supported platforms

Agora’s developer experience is quite good; as an amateur developer, I felt this first hand. Earlier this year, when there was a real possibility that WeChat might be banned in the US, I spent half a day spinning up an Agora instance that can be used for video chatting with family in China. While that option did not end up being necessary, the process of learning about Agora’s APIs and reading its documentation to figure out how to use them on a basic level was easy and straightforward.

APIs: Agora is a quintessential API product company, similar to Twilio, though tackling a different communications vertical (Twilio is more text messaging focused). It also exhibits the two hallmarks of a well-designed API: self-service and pay as you use, which we discussed in-depth in Part II.

Agora gives every new account 10,000 free minutes, so a developer can prototype and demo a working concept without worrying about cost. Its pricing is all based on a “per 1,000 minutes” scheme, and the price varies depending on how heavy and complex the workload is on the Agora network, e.g. live video streaming is more expensive than live audio streaming.

Similar to most APIs, being an abstraction layer over common resources, the Agora API has a broad range of use cases. Its three most prominent use cases are online education, telehealth, and social streaming -- all of which have been propelled forward many years by the COVID-19 pandemic. The network that the Agora API sits on spans more than 200 data centers located in all continents, except for the one that has more penguins than people.

Source: Agora whitepaper “Agora’s Software Defined Real-Time Network; Delivers Real-Time Internet Advantages Over Content Delivery Network”

Open Source: Agora has not open sourced any of its code base. The only way to learn about its internal implementations is through the technical whitepapers it publishes, which are also used as a lead-generation tool just like most software companies. That being said, the whitepapers are quite technical and not fluffy marketing collaterals.

Agora has started to expand its presence on GitHub with a series of open source SDKs aimed at specific developer sub-communities, e.g. Unreal developers for gaming, web developers, React Native developers, etc. This is a necessary, though early, effort to expand Agora’s global reach. Its education-focused iOS and Android SDKs were just uploaded a few days ago. Evidently, Agora is still a young company -- the youngest of the three I’m discussing in this post -- so many of its processes, executions, strategies, and products still have much room to mature.

JFrog ($FROG)

Founded in 2008, JFrog more or less jumpstarted the DevOps trend before “DevOps” was even a word. The DevOps concept is still ill-defined. The way I understand it is a merging of two traditionally separate responsibilities in software engineering: developers (people who build applications) and operations (people who provision and manage the infrastructure resources to enable building applications), thus DevOps. (How much of a DevOps person’s work is “dev” vs “ops” is a constantly debated topic.)

This merging of responsibilities is motivated by the need to build, deploy, and maintain software more quickly than ever before. With more of everything getting “technologized”, a big theme we discussed in Part I, speed and security of building software is key. That’s the business JFrog is in, hence its catchy vision of a “liquid software” future.

Developers: Artifactory, the open source project that underpinnings JFrog’s commercial products, is a package manager, which is a foundational element of a DevOps’s workflow. In plain language, what it does is it manages and secures the binaries (literally 1’s and 0’s) that are running on computers, which are often also called “artifacts.” Thus, this project is a factory for managing artifacts, hence “Artifactory.”

There are many package types in the software ecosystem, and JFrog’s Artifactory supports pretty much all of them -- from the ubiquitous npm, to the obscure Bower.

JFrog's officially supported pacakges

This breath of support makes JFrog a classic developer-focused company. Supporting the obscure options also has some added benefits of smoothening enterprise adoption that are not well understood. A large, old enterprise tends to have aging technologies in various corners of its IT operation, which it prefers to be supported even when adopting new technologies. By supporting all package types, regardless of their current popularity, JFrog gives potential enterprise customers one more reason to say “yes” to its offerings versus other vendors that support less options, like GitHub. One less hurdle can make a world of difference in enterprise sales.

JFrog is meeting the developers where they are. That’s why the size of its developer community is in the millions.

APIs: JFrog is not an API company. Given the specific needs and challenges of DevOps, an API abstraction layer may or may not even make sense.

Its commercial product is a suite of tools around Artifactory that helps with security, continuous integration/continuous delivery (or CI/CD), distributing releases, and other parts of the software delivery workflow. It charges customers either an annual licensing fee (the traditional way) or a monthly cloud-based subscription fee (the SaaS way), which is pay as you use, priced on a combination of storage size, network transfer bandwidth, and CI/CD minutes usage per month (e.g. 10,000 minutes / month).

This duo license-plus-SaaS pricing method is common in software companies, though usually as a transition phase to lessen licensing revenue and increase SaaS revenue. As the world moves to the cloud, the way software products are consumed is increasingly via subscription services that are constantly updated and maintained, not once or twice a year based on a licensing agreement. (Also investors give SaaS revenue a much higher valuation multiple due to its recurring nature and predictability.) Similar to Agora’s 10,000 free minutes and other companies’ free tier options, JFrog recently introduced a free cloud tier, to grow and strengthen its developer community.

Open Source: JFrog is a great example of a successful company built on open source. Artifactory, in this case, is already widely used among the DevOps teams inside tech companies, and will likely be one of the first options to consider for non-tech companies trying to hire DevOps to digitally transform themselves.

The “global by nature” open source paradigm is so powerful in JFrog’s case that the company grew from nothing to IPO without any outbound sales effort. All of its customers come from inbound requests, meaning they are already using open source Artifactory for free and want to buy more features and products from JFrog to continue using it better. The open source global reach has also pushed JFrog to expand its presence to 10 countries and counting.

Only during JFrog’s first earnings call as a public company did we learn that the company is building a “strategic team” inside its sales structure to pro-actively expand business with its top 100 enterprise customers. Even so, the effort looks to be an extension of its inbound sales motion. With an open source-powered developer flywheel turning, outbound sales, which are more expensive and less efficient, may simply be unnecessary for JFrog.

MongoDB ($MDB)

Founded in 2007, MongoDB (formerly known as 10gen) is one of the most popular NoSQL, document-based databases in the industry. The company is built on an open source version of MongoDB, similar to JFrog’s relationship with Artifactory.

Developers: MongoDB has built one of the largest developer communities in the software industry. Its Mongo World 2019 conference had more than 2,000 in-person attendees -- impressive for a single product conference.

Being an easy-to-use, scalable database designed for building web applications, Mongo’s developer mindshare grew in part by popularizing the MEAN (MongoDB, Express, Angular, Node) stack pattern and its close cousin MERN (Mongo, Express, React, Node). The MEAN acronym was first coined in 2013, not so coincidentally by a MongoDB kernel engineer in a blog post.

These acronyms may feel gimmicky, but can be a powerful tool to increase developer awareness, education, and productivity -- a trusted stack or pattern can be very useful. The OG of successful stack acronyms is the LAMP stack (Linux, Apache HTTP server, MySQL, PHP), which was what Facebook (among many, many other companies) used to get started. Facebook is still a heavy user of MySQL today. I personally learned web development on the MEAN stack, thus MongoDB was the first database I used, just like many other new developers.

Being able to coin a catchy acronym isn’t enough. MongoDB has one of the most meticulous and sophisticated developer marketing and relations operations to push MEAN, MERN, and other ways to get developers to build applications on MongoDB. Developer community building takes dedicated resources and organization, and requires years of patient effort. I learned this the hard way. When I was the GM at PingCAP, I also coined a stack acronym, KOST (Kubernetes, Operator, Spark, TiDB), as an open source, cloud-native hybrid database pattern. The use case certainly existed (and still does), but the acronym never took off.

MongoDB also officially supports a wide range of developer tools in the form of drivers (common for databases), though it’s not as wide as I’d expected. Fairly popular languages, like Erlang and Lua, are missing. Given MongoDB’s new direction to become the go-to database for mobile development, to replicate its success in web application development, its attention is likely focused on Java (for Android) and Swift (for iOS). It relies on its large developer community to support other drivers.

Officially supported MongoDB drivers. Source:

APIs: The MongoDB API is powerful. I personally underestimated its power, because of the many alternative NoSQL, document-based databases in the market -- some from the cloud giants themselves, others from startups. As we discussed in Part II on APIs, one way to spot a winner in the “API economy” is to look at where the trend of standardization is happening toward or the “compatibility targets”.

MongoDB is definitely a compatibility target. Its two largest competing products in the cloud -- AWS’s DynamoDB and Azure’s CosmosDB -- both have MongoDB compatibility. In October 2019, Alibaba Cloud formed an official partnership to offer MongoDB as part of its AsparaDB database product line. GCP simply decided to make MongoDB a first-class product to directly distribute, support, and sell, instead of building a competing product to be compatible with MongoDB. This is part of GCP’s broader strategy to co-exist harmoniously with open source based commercial vendors, rather than competing with them head on, like AWS.

Open Source: Depending on who you ask, it’s debatable whether MongoDB should still be considered open source anymore, after its licensing change in 2018. In a nutshell, the new license, Server Side Public License (SSPL), restricts other companies from providing MongoDB as a cloud service. It stirred up a lot of industry-wide controversy, and MongoDB eventually withdrew its request for the license to be approved by the Open Source Institute (OSI). The OSI is the organization that has the power to approve whether a license is considered “open source” or not. Again, depending on who you ask, it’s debatable whether OSI should have that power or not.

We won’t focus on the “open source” merit of the SSPL, but it is important to mention it, because it’s a tangible manifestation of the tradeoffs associated with the “open source paradigm.” MongoDB clearly benefited from the paradigm to drive its software’s adoption and developer community growth. The same openness also allowed cloud companies to commercialize MongoDB, the open source software, without paying MongoDB, the company, a dime. The SSPL was designed as a response to this market behavior, not just to AWS, but also to many companies in Asia, who were doing the same. As MongoDB CEO Dev Ittycheria revealed in the partnership announcement with Alibaba Cloud, China has been the biggest source of downloads for MongoDB in the last four years. It’s most likely that only a small fraction of the revenue generated from those downloads went to MongoDB, the company.

MongoDB’s experience with the open source paradigm illustrates that openness can be a double-edged sword. How to navigate the pros and cons of openness effectively will determine who the winners will be in Globalization 2.0 -- a world driven by the open source paradigm.

Many Other Opportunities

Full disclosure: I own shares in all three companies discussed. Some of you may view this as a conflict of interest, which I fully understand and appreciate, but I also believe strongly in having “skin in the game.” If I think a public company is worth watching and I’m writing about it, I should invest my own money in them and be transparent to my readers. Otherwise, I’m being hypocritical -- principal-agent problems abound. (As the saying goes: no conflict, no interest.)

There are quite a few other public companies that can be analyzed in the “global by nature” lens, like: Twilio (I own shares), Cloudflare (I own shares), Fastly, Elastic, PagerDuty, Unity, and Datadog. Stripe and GitLab would also be great companies to consider, when they go public. If readers are interested, I will write more analysis on these other companies in the future.

There is also a huge crop of promising startups -- too many to name -- where applying the developers-APIs-open source lens may be useful. Who among these younger companies will emerge as winners in the next 10 years will be the most intellectually challenging and potentially profitable question to think about.

ICYMI: please give the other three parts of the "Global by Nature" series a read: developers, APIs, open source.

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 (twice per week). Follow and interact with me on: Twitter, LinkedIn.


在 "生来全球化" 系列的前三篇中,我们深入探讨了科技驱动的“全球化2.0”的三股力量:开发者(人)、API(产品)、开源(模式)。



这个视角也仅是一个质性的框架,绝不是一个挑选投资的checklist。您会在本篇分析中看到,有些公司并没有框架内的所有三个要素,也不是必须都具备。如果您想把本篇的质性分析与一些量性指标结合起来,我建议看Redpoint Ventures的Jamin Ball的Clouded Judgment博客。他跟踪并更新一批软件公司的常用行业指标,例如:EV/NTM收入倍数、美元净留存率、毛利率/营业利润率等。

声网 ($API)

声网成立于2012年,是一家API公司,让开发者能够构建实时视频、音频和短信产品(以至于其股票代码都是 "API")。其创始人赵斌是视频和音频的资深技术人士。他与创立Zoom的袁征是同一个Webex团队的成员。他曾是YY的CTO;YY也是在中国,甚至可能是全世界,最早达到规模的直播平台之一。YY最近也受到了著名做空公司 Muddy Waters 的审查。(如果您对YY的大众文化影响力有兴趣,我推荐您看纪录片《People’s Republic of Desire》)。


因此,声网支持多种平台(Android、iOS、macOS、Web、Windows、Linux)和框架(Electron、Unity、Flutter、React Native、Cocos Creator)。这个支持范围的建立需要很长时间,以及对开发者群体的坚定关注和承诺。支持的平台、框架或工具的范围通常是一个很好的信号来看一家公司是真正以开发者为中心,还是只是口头上说说一条时髦的概念。




声网给每个新账户1万分钟的免费使用时间,因此开发者可以在不担心成本的情况下做出一个原型和demo。产品定价都是基于 "每1000分钟" 的方案,价格根据声网网络上工作负载的繁重和复杂程度而区分,比如:视频直播比音频直播要更贵。


Source: Agora Whitepaper "Agora's Software Defined Real-Time Network; Delivers Real-Time Internet Advantages Over Content Delivery Network”


声网已经开始在GitHub上扩大自己的影响力,推出了一系列针对特定开发者群体的开源SDK,例如用Unreal的游戏开发者、Web应用程序开发者、React Native开发者等。这是扩大声网全球影响力的必要一步,虽然还很早期。其专注于在线教育用例的iOS和Android SDK在几天前才刚刚上传。显然,声网还是一家年轻的公司,是我这篇文章中讨论的三家公司中最年轻的一家,所以还有许多流程、具体操作、战略和产品方面有很大的成熟空间。

JFrog ($FROG)

JFrog成立于2008年,在 "DevOps” 这个字眼还不存在之前,就或多或少地开始发动了整个DevOps的趋势。DevOps这个概念在业界还没有明确固定的定义。我对它的理解是将软件工程中两个传统上分开独立的职责合并在一起:开发者(构建应用程序的人)和运营者(提供和管理基础设施资源给开发应用程序的人),因此是DevOps。(做DevOps的具体工作内容有多少是 "开发" 与 "运营",还是个有不断争论的话题。)

这种职责的合并是被各大公司需要更快地构建、部署和维护软件的需求所驱动的。随着越来越多的东西被 "科技化",也是我们在第一篇中讨论的一大主题,构建软件的速度及安全性极其重要。这就是JFrog的业务,因此它也畅想了所谓的 "液体软件"的未来愿景。

开发者:Artifactory是JFrog商业产品的开源基础,是一个代码包管理软件,也是DevOps工作内容不可缺少的环节。通俗点来说,它的作用就是管理和保护在计算机上运行的所有二进制文件(binaries),通常也称为 "工件" (artifact)。因此,它是个管理工件的工厂,所以叫 "Artifactory"。

软件科技生态中有很多代码包类型,JFrog的Artifactory几乎支持所有类型 -- 从无处不在的npm,到默默无闻的Bower。








从JFrog的经历来看,"生来全球化" 的开源模式的力量非常强大,以至于公司在没有任何对外销售投入的情况下,就从零做到上市。它所有的客户源都是自己找上门来的,也就是说他们已经在用免费开源的Artifactory,希望从JFrog购买更多的功能和产品来把它用得更好。开源的全球影响力也推动了JFrog将公司扩展到10个国家,并且还在不断扩大。

在JFrog作为上市公司发表的第一个财报中,我们才了解到公司正在内部销售团队中建立一个 "战略小组",以主动地与最大的100名企业客户扩展业务。即便如此,这看起来也只是其内部销售构架的延伸。在开源驱动的开发者“飞轮”的转动下,成本更高、效率更低的对外销售对JFrog来说可能根本没必要。

MongoDB ($MDB)


开发者:MongoDB建立了软件行业内最大的开发者社区之一。它2019年的Mongo World大会有超过2000名亲临现场的参与者 -- 对于一个单产品的会议来说,还是令人刮目相看的。


这些缩写感觉像是不靠谱的噱头,但其实对提高对开发者的影响力、教育和实际产出很有力 -- 一个可以信赖的构架或模式是非常有用的。最成功的老牌构架缩写就是LAMP(Linux、Apache HTTP server、MySQL、PHP)。Facebook(以及许多许多其他公司)就是用它起步的。Facebook至今仍是MySQL的重度用户。我个人是在MEAN栈上学习Web开发的,因此MongoDB是我用过的第一个数据库,就像许多其他新的开发者一样。

能够创造一个好听的缩写是不够的。MongoDB拥有最细致、最成熟的开发者营销和关系建立团队,从而来推动MEAN、MERN等方式让开发者用MongoDB做项目。开发者社区建设需要专门的资源和组织,以及多年的耐心努力。我个人对这一点深有体会。当我还是PingCAP的GM的时候,我也创造了一个构架缩写,KOST(Kubernetes, Operator,Spark,TiDB),作为一个开源的云原生混合数据库模式。这个用例当时是存在的(现在也存在),但这个缩写一直没有流行起来。



APIs:MongoDB API很强大。我个人一直低估了它的力量,因为市场上有很多NoSQL、文档构架数据库 -- 有些来自云大厂本身,有些来自创业公司。正如我们在第二篇关于API的讨论中所说的,在"API经济”中寻找赢家的一个方法是看标准化的趋势朝什么方向走或找 "兼容目标"。

MongoDB绝对是一个兼容目标。它在云行业里最大的两个竞品 -- AWS DynamoDB和Azure CosmosDB -- 都兼容MongoDB。2019年10月,阿里云与MongoDB达成正式合作关系,提供MongoDB作为其AsparaDB数据库产品线的一部分。GCP甚至都没有自己云的竞品,而是直接把MongoDB作为“一级”产品来直接分发、支持和销售。这是GCP更大的战略方向的一部分,即与基于开源软件的公司和谐共存,而不是像AWS的打法与它们正面竞争。

开源:取决于您问谁,MongoDB在2018年的许可变更后,是否还应该被视为是开源,这是有争议的。简而言之,这套新的许可,即Server Side Public License (SSPL),限制其他公司做基于MongoDB的云服务产品。这个变动激起了全行业的很多争议,MongoDB最终撤回了该许可可否得到开源协会(OSI)批准的申请。OSI是有权批准一个许可证是否应该被认为是 "开源"的组织。同样要看您问谁,OSI到底该不该有这种权力也意见不一。

我们的重点不是要讨论SSPL是不是 "开源” 这个问题, 但需要提到它,因为它的存在是 "开源模式" 优缺点权衡的一个好例子。MongoDB显然是开源推动其软件被广泛应用和发展开发者社区的受益者,但同样的开放性也允许了云厂商将MongoDB这个开源软件商业化,而无需向MongoDB这家公司付一分钱。SSPL的目的就是对这种市场行为的回应,不仅仅针对AWS,亚洲很多的公司也在做同样的事情。正如MongoDB CEO Dev Ittycheria在与阿里云合作的官宣中自己透露的,在过去四年中,中国是下载MongoDB这个软件最多的国家。这么多下载所产生的收入只有很少一部分留到了MongoDB这家公司。

MongoDB在开源模式上的经历说明,开放可能是一把双刃剑。如何有效地运作开放的利与弊,将决定谁将成为全球化2.0的赢家 -- 一个由开源模式驱动的世界。


郑重披露:我持有这三家公司的股份。您可能会认为这是一种利益冲突,我完全理解这个观点,但同时我也坚信另一个道理:“skin in the game”。如果我认为一家上市公司值得关注,而又写文章分析,那我也应该投自己的钱,并对读者保持透明。否则,我的行为是虚伪的 -- 委托代理的问题多多。(俗话说:没冲突,就没利益。)

还有不少几家上市公司也可以用 "生来全球化" 的视角框架来分析,比如:Twilio(我有持股),Cloudflare (我有持股), Fastly,Elastic,PagerDuty,Unity和Datadog。Stripe和GitLab等它们上市后,也会是很好分析对象。如果读者有兴趣,我今后可以写更多关于这些公司的类似分析。

此外,还有一大批极有前途的创业公司 -- 数量太多,无法一一列举 -- 可以用 “开发者+API+开源”的视角框架来分析,会非常有价值。在这些年轻的公司中,谁将在未来10年内成为赢家,将是最挑战智力也是盈利潜力最大的一个问题。