Welcome to all the new subscribers! I write a longform analysis about tech in China from the perspective of a 🐢 VC, sent out every Wednesday. If you like the content, please share with friends!

A former colleague of mine, W, was a very sharp-eyed and blunt Eastern European. W can spot the single error on a packed PowerPoint slide at a glance, and will tell you that you fucked up in the next breath. A stickler for details and facts, W would be on the verge of pulling their hair out every time we had to put together some venture market overview slides. Why? Because Pitchbook's data would miss chunks of deals, miscategorise companies, be inconsistent with their own data 6 months prior, etc. etc. The list goes on. These errors stabbed at W, and what usually happened was days spent on cleansing and augmenting data by some sorry team of analysts and associates until it made the W cut.

I tell you this because what I'm about to present to you will not pass W's scrutiny. But really, I think that in itself is interesting. So let's begin.

In our long-awaited second instalment of 🇨🇳 SaaS Series, I want to understand what sectors of SaaS (or failing complete SaaS, at least enterprise software) are gaining traction in the Chinese VC space. Since there's no current Chinese SaaS giant, knowing what's getting funded can give us insight into the future unicorns. However, getting clear data on Chinese VC landscape has three setbacks. The first is that of coverage, second that of completeness and the third is that of credibility. These issues afflict all VC investment databases (and dare I say it, almost all tabulated databases), and especially so for China.

In terms of coverage, established western venture data sites like Crunchbase and Pitchbook are by their own admission, patchy. Crunchbase relies on self-reporting and moderators who are biased towards English news sources. Pitchbook uses crawlers and more moderators, but is very US-centric in their data coverage. My experience was that Pitchbook didn't pick up the smaller deals in Europe, and no doubt that goes for other regions too. So I went with a China-focused venture data site called ITJuzi hoping it has more comprehensive coverage. Similar to Crunchbase and Pitchbook, ITJuzi uses a combination of web-crawling, user submissions, 3rd party data channels like recruitment websites to gather data.

The issue of completeness occurs when fundraise news releases leave out information. I'm not talking about a lack of top-line valuation (which is quite common) but rather crucial data points like the round size or the investors involved. Leading to entries in the database like these:

Screenshot from ITJuzi data - each line is a new deal announcement, the red circles mean undisclosed amounts.

One of the key themes of this newsletter is that I believe China's difference to the West is almost always one of the degrees rather than one of the absolutes. There are also investment announcements in the West that omit information, though glancing through the entries I feel that China has more of these (though mostly for Seed and Angel rounds). My interpretation is that selective information withhold might make more strategic sense in a hyper-competitive context. Why let the others know how big your war chest is? Or what strategic alliances you made through this fundraising? Only the paranoid survive.

Credible data is somewhat optional in an industry based on the manifestation of magical thinking; just look at your standard hockey stick startup business plans. An open secret in startup land is that sometimes the announcements are on the stretchy side of truth. The $10m Series A may include a $2.5m convertible bridge loan, and a pre-A round by an undisclosed strategic investor. Or sometimes the deal is tranched (meaning that the funds are released in blocks when certain conditions, such as growth, are met) and so not all of the $10m will actually be given to the startup. I don't think these actions are always driven by malice or vanity, mostly just busy founders working with stretched resources, faking it until they made it. In terms of fraud, I'm sure it happens in China just like it happens in the West. Founders and VCs who have skin in the game are also working on minimising this (GGV capital famously hires private investigators to profile founders they are closing a deal with). So I think there is an inflation factor from fraud, but its size is unclear.

The combination of lack of coverage, completeness and credibility compounds to mean that it's hard to get a perfect picture of China's VC investment market. The question is then, how good of a picture can we get? In a sense check comparison, there are significant divergences between Pitchbook (above from the Economist) and ITJuzi (below) on overall levels of capital deployed. The W voice in my head is freaking out about the gap. Because if Pitchbook is this wrong about China, what else could it be wrong about? Or it is ITJuzi who's completely wrong with their overinflated numbers? Is there no single source of truth?

With all these caveats and freak-outs aside, let's look at the data. First of all, China's pace of deployment has slowed down since 2018. Even though enterprise software is increasing as a proportion of VC investment, it still only accounts for 15% of the total in 2019. Since we're 78% through the year, I've calculated an implied run-rate for the rest of 2020 based on the investment pace so far. Enterprise software investments look to be decreasing in 2020, even more so than other sectors. I'm surprised since I expected Covid-19 to be an enterprise software accelerant.

The final issue which has popped up as we go deeper into the data is around data labelling and classification. The decision for what constitutes a distinct sector and into which sector bucket a startup goes becomes a Foucauldian exercise the more one dwells on it. Should AI be a sector in itself? (Why did it become a sector anyway? Was it because it allowed a lot of VCs to raise sexy AI funds?) Why do we create sectors around functions (e.g. sales and marketing) and the user base (e.g. developers)? And, to tie it together, should an NLP API tool for CRM entries be put in the sector for AI or Developer Tools or Sales and Marketing? Or all of these? Since I only have access to ITJuzi's aggregated sector figures, I have no idea how they've made the call. This issue could be significant as the Chinese enterprise software style is to be all things to all users, I'm sure there's a bunch of ERP CRM AI collaboration tools being misclassified to no end. (If you are interested, contribute to the $100 cost of getting granular ITJuzi data and I'll get to the bottom of it).

The sub-sectors that have been gaining the most investment attention are BI & data analysis, AI & frontier tech, and Infrastructure & developer tools in recent years. On one hand, this makes sense since most Chinese companies are not digitised. On the other hand, it is again surprising that China has not followed in the footsteps of the US, whose first SaaS giants were sales and marketing offerings like Salesforce or ServiceNow. I suspect the lack of a standardised sales process for Chinese enterprises is affecting the wide adoption of CRMs and other sales and marketing related tech. Since the majority of China's communication happens on WeChat, it also calls for a rethink of the entire CRM structure (this new segment is called Social CRMs).

Suppose we want to take the data at face value. In that case, it's interesting to contemplate how China’s enterprise software journey could be starting from the infrastructure layer rather than the application layer. Maybe China is leapfrogging straight into the meta-enterprise companies (which I defined as companies who exist to serve the needs of other tech companies or tech departments within companies), bypassing the phase of standardised applications altogether. Maybe it’s following the traditional route of software (albeit in the cloud) with licenses and customisations from an army of IT consultants before moving to SaaS. Maybe W will chill out to the unknowableness of the universe. I guess we'll find out.

The nature of data in professional services or How I Learned to Stop Worrying and Love the Figures 

When I was a management consultant, I used to make up numbers. Not all the time. That would be committing fraud. But how do you think management consultants come up with market growth estimates for industries that had no market growth estimates?

The formal name for this methodology is called qualitative research. We'd call up 20 - 30 market participants and ask them what they think their industry growth rate was. A bunch of highly educated 20-something sat in tiny grey booths, calling for 12 hours a day and reading from pre-written questionnaires about market drivers, competitor qualities and growth forecasts. We'd find these market experts, either through cold calling or paid knowledge networks and ask them patiently for the 5th time that day, how quickly do they think their industry was growing?

Oh, you don't know? (But I need a number for my client.)

Do you think it's less than 5%? Or like 5%? More than 5%?

5% sounds about right, doesn't it? (It's a nice round number, let's go with that.)

And just like that, as silent as a conspiracy, a new growth rate was created.

To be clear, about 50% of the time, there were widely respected market growth data from public databases like the World Bank or the US government. 40% of the time, there would be industry reports, or the experts were readily chirping out a number, or an intrinctly built bottom-up model. But there’s 10% of the time when the client is paying the firm the big bucks to find out the figures and none of the market participants has a clue (because the number doesn't exist yet).

The experiences picked up during that 10% of cases incept the idea that all those other figures are made up. Because did I believe that at the end of all those other figures, was not some tired 20-something analyst trying to get to a number so they can finish for the day?

For our next week's edition - I'm running a Ask Me Anything post. So please comment or reply to this email with your questions! I'll answer all of them as best as I can. 

Reply to this email to let me know what you think, I read and reply to everything. Or find me @lillianmli where I write about what didn't make it into the post throughout the week