Quant Approaches to VC are Overrated

A popular meme in startup investing right now is on the increasingly data-driven nature of the industry.  More and more firms are employing developers and data-scientists internally to mine the trove of publicly available data to provide signals for companies that are exhibiting attractive momentum.  Here are a couple recent articles, and we are also seeing companies like Mattermark popping up specifically to assist both VCs and angels in this endeavor.

Call me old school, but I think the impact of quantitative approaches on early-stage investing is pretty over-rated and misunderstood.  Data is important and helpful (more on that later) but will not be core to what makes VCs successful, especially when it comes to identifying and sourcing the best early stage investment opportunities.

Here’s part of what informs my thinking.  My prior firm, Spark Capital, has had a number of pretty terrific exits recently (congrats guys!).  Namely:

  • Tumblr: Acquired by Yahoo for $1B
  • Adap.TV: Acquired by AOL for $400M
  • Admeld: Acquired by Google for $400M
  • OMGPOP: Acquired by Zynga for $180M

I often ask myself “how obvious was it at the time that these companies would be successful?  How attractive would these companies have been at the series A (and in some cases, the series B) stage based on trackable measures of rapid growth or momentum?”

Honestly, I think the answer is that while these companies had a lot of good things going for them, many data-miners would have been relatively unimpressed early on. These companies did have some solid metrics, but they did not see hockey-stick like momentum and there were still a hundred reasons why they could fail.

At the time, we also saw a number of other companies that seemed to have surprisingly remarkable surges in growth and usage.  These are the companies that quants would have identified as high-momentum opportunities. We didn’t invest in any of them. I don’t think many of them are still in business today.

I think ultimately, this is because early-stage investing is so much more about people, markets, and judgement than cold, hard, data.  These great investments were much more a bet on a team, a vision of the future, and an early product that seemed right, even if the quantitative evidence was still pretty slim. Maybe at the later stages, one can take a more quantitative approach to picking the best companies, but I wonder if that will really be true, especially since pricing at the later stages has been so astronomical in recent years.

Now, I don’t think quantitative approaches to VC are useless.  It’s an ever-changing ball game, and we at NextView employ data mining software for our purposes as well. But I think quantitative analysis and data mining is much less about finding the best companies or identifying hidden gems, and much more about understanding trends and market shifts.  I think VCs that use data well are much less likely to “source” their next hot company directly from a data signal, but are much more likely to make smarter decisions about investments and better help their portfolio companies because of the data they are tracking and analyzing across thousands of companies.

Rob Go

Thanks for reading! Here’s a quick background on who I am:
1. My name is Rob, I live in Lexington, MA
2. I’m married and have two young daughters. My wife and I met in college at Duke University – Go Blue Devils!
3. We really love our church in Arlington, MA. It’s called Highrock and it’s a wonderful and vibrant community.  Email me if you want to visit!
4. I grew up in the Philippines (ages 0-9) and Hong Kong (ages 9-17).
5. I am a cofounder of NextView Ventures, a seed stage investment firm focused on internet enabled innovation. I try to spend as much time as possible working with entrepreneurs and investing in businesses that are trying to solve important problems for everyday people.  
6. The best way to reach me is by email: rob at nextviewventures dot com

    • Well said Rob.

      • robchogo

        Thanks Brad

    • T .S Vineet Devaiah

      I agree 100%. I would generally be encouraging of these services especially Mattermark but I was just sick looking at their methodology. They still use Alexa ranking ( really !! ) . Anyways I do not have too much experience in picking startups but atleast I can rational numbers when I see them and those scores lack basic regressive analysis – Not to mention the heavy usage of the term “BIG data” .

      I would encourage these companies to atleast adopt better ways of data mining – whether or not they are relevant is based on their customers i.e VC’s

      • robchogo

        If you torture the data, it will confess

    • Mathieu Gosselin

      I would totally agree if i was a VC… Ok i do agree if my 2 cents matters.

      I see this linked to a long-term versus short-term vision of the game. Either you’re jumping into a hype that might crash as fast as it spread or you’re looking for something that has the potential to grow on solid grounds. Which sounds actually like a safer bet. That’s what common sense investment should be about in my opinion. But that’s true that data cannot tell you that. And what we can’t measure, don’t exist… if we follow the rational approach.

      VCs shouldn’t be shy to trust their guts i believe. 90% of decision making is unconscious. If we can’t put words or quantify it doesn’t mean it’s wrong. Actually it’s just we can’t express those into facts or words, but our brain figured out the right answer. It’s just a tad more difficult to convince a board or a committee based on those feelings you can’t express 😉

      • robchogo

        It’s both an art and a science. But will always be more art than science, IMHO

    • The RydeMyPony Team

      All valid points about the situation right now. But your sentence “These are the companies that quants would have identified as high-momentum opportunities.” tells me one thing: The quants aren´t looking at the right data yet! I think the job at hand right now is finding the criteria that successful startups have in common at an earlier stage. They might not exist, they might be “founder hasn´t shaved in weeks”, but my gut feeling is: we will become a lot smarter about this in the next 3-5 years.

      • robchogo

        Possibly. I tend to think that the market changes so much that historical correlations don’t match current environments, typically. The feedback loop is too long.

        • The RydeMyPony Team

          Valid point. So if “quant VC” is going to work, the art is going to be to account for those market changes in real time. Hey, I never said it was gonna be easy! 😉

    • amen brother

      • robchogo


    • I don’t have a horse in this race, as I’m not a VC and largely just long my own efforts.

      BUT…This is basically the script articulated by every business that was just about to be turned upside down by the use of stats & numbers:

      –“A computer can’t pick great baseball players. There’s too much nuance in knowing how young players develop.”
      –“A computer can’t pick stocks. There’s too much built into the unique story of each company that can’t be quantified.”
      –“A computer can’t drive cars. There are too many crazy variables coming from every direction.”
      —etc, etc, etc

      Sure will be fun to see what happens. 😉

      • robchogo

        Good point. But in your first two examples, the winning strategy is a combo of both human judgement assisted by data. And as more people utilize data, judgment remains more important.

        • Yup. And I’d think that would be the only working model for early-stage investing. Way too many factors that you can’t (yet) quantify.

    • Great post, but just a couple notes:
      – from what i understand, companies like MatterMark can be used at the sourcing stage rather than the decision-making stage: i find them an interesting way for VCs to discover companies gaining traction outside of the SV bubble.
      – Quant approaches can be employed through other means, for example Correlation VC taking into account area specialization of coinvestors and making bets based on that. Piggy-backing off of the individuals.
      – For every VC, even the human ones, looking at data and employing quantitative analysis at some level is imho essential. For sourcing and due diligence, but also for tracking your own team’s performance. The deals you missed, the ones you passed on, etc etc. It’s essential.

      The discussion is very similar to the one we have in product circles regarding “data driven decision making”. Blindly following analytics makes you into an adult site, but ignoring them is nonsensical. The answer is somewhere in the middle.