Quantitative methods and seed stage investing
The recent premier of Moneyball has caused many to ask the question: Can the same principles be applied to venture capital investing? Dan Frommer interviewed a handful of smart and influential early-stage investors, each of whom was cynical about the applicability of quantitative methods to early-stage investing. Fred Wilson further amplified his thoughts in this post. Given my pre-VC background in quantitative trading, I’ve given a lot of thought to this issue.
For starters, I think it is important to distinguish between security selection and investment selection.
When I speak of security selection, I’m specifically referring to the choice of founders with whom to work. So when Fred, Brad and Albert decided “We’re going to invest in this next-generation media platform called Twitter,” it was a decision based upon spending lots of time with the founders, interacting heavily with the platform and developing a thesis of how the business could scale, transform and disrupt industries. It was the outgrowth of a decision between the founder(s) and the investor(s).
When I speak of investment selection, I’m referring to the choice of investors with whom to invest. It is far less security-specific; a given company matters less than the objective attributes of the investment. This is completely different than security selection.
I was on the quantitative end of Wall Street for more than 15 years, and have been doing seed stage venture investing for more than 7 years. Both experiences have shaped my perceptions concerning the application of data and quant tools to choosing the best companies with which to partner. As it relates to security selection, I’m firmly in the camp of Fred, Chris Dixon and Paul Graham. My own experience has shown that the impact of people and team is far greater than market and product, the last two of which better subject themselves to quantitative assessment than the former. The degree of variability concerning the range of possible outcomes is much more heavily impacted by how a team interacts, innovates and executes than the perceived market opportunity. Modeling markets is often a worthwhile endeavor; modeling human dynamics in seed stage investing is generally not. The notion of “pattern recognition” - of accumulation of subjective experience - is in my opinion far more constructive than any heuristic.
Investment selection is another matter entirely. In fact, a wide swath of the angel and Micro VC segment are based upon the efficacy of “social proof.” This speaks more to the success and credibility of those also investing in a “security” than a deep understanding of the security itself. My friend Bryan Birsic made the comment that Correlation Ventures (CV) “takes quantitative approach to VC.” Given the framework above, I’d say that CV take a quantitative approach to investment selection - not security selection. This is a very important distinction. The CV model does not allow them to lead rounds, because they are not claiming that their algorithms work on a stand-alone basis; they require the quantitative version of “social proof” as key inputs into their model for whether or not to invest in a company. USV doesn’t rely on other investors as ballast for their investment decisions, and neither does IA Ventures. CV has a taxonomy of investors (read: top VCs and potentially some quasi-institutional angels) for whom they’ve crunched numbers and determined “These are the firms we’d like to co-invest with.” Then they probably look at a template of company-specific metrics to say “This company passes enough of our metrics-driven tests such that as long as their round is being led by one of our "approved” firms, we’re enthusiastic co-investors.“ But this is not dependent upon their getting to know management or to have the kind of connection that seed stage investors such as Fred, Chris or I have with our founder partners.
In short, I do think that quantitative methods such as those used by CV are likely to be effective as they are systematizing and institutionalizing the notion of "social proof.” But it is not a substitute for security selection; that needs to be done by firms relying on keen assessments of people and their ability to execute a vision, adapt in the face of change and persevere against all odds. At least at this point in time, you can’t model that.
