In this special feature by Thomas Thurston of Growth Science International, we look at how HPC could be used by venture capitalists to pick winners more consistently.
HPC and Wall Street have a happy marriage. It’s a love-fest – making perfect sense since banks have big enough wallets for HPC and are always looking for a technological edge. The scale of this love-fest is tremendous, for example around 40% of trades on the London Stock exchange were done by robotic intelligence in 2006. United States estimates are closer to 80%. But what about venture capital?
In this world of data and calculation, the job of a venture investor seems an anomaly indeed. While there are some exceptions, most venture investors allocate billions of dollars every year based on little more than gut intuition and subjective experience.
This is not to, in any way, detract from successful investors; the ability to pick winners and wrestle out a deal can require tremendous skill. The point here is merely to pose a question. Given the dollars at stake and lives in the balance, will venture investing inevitably evolve in a more empirical direction? Will there someday be robo-venture capitalists? Perhaps more pointedly, can venture capitalists learn to benefit from HPC?
Once upon a time advertising was a matter of visionary intuition and a “special human touch.” Today marketing MBAs are increasingly being replaced by statisticians and datacenters. Software developers are becoming the new “Mad Men.” If you want to know what makes left-handed, Republican, blonde, female smokers buy pink purses in Nebraska, you’re increasingly better off writing a few lines of Linux rather than hosting a living room focus group.
While intuition-based venture investing works sometimes, it mostly fails. Investors spend lots of time screening deals only to see around 90% fail (on a good day) and industry-wide VC returns in the US over the past 10 years were negative (as of 2010). Not only is venture capital falling short of investor expectations but it – by necessity – excludes most ventures. Less than 1% of startups attract VC investment in any given year, while less than 95% of businesses attract any equity investment at all (either from VCs or angel investors).
A common counter-argument to HPC tools is that venture investing is inherently unquantifiable. Too unpredictable. Too subtle. Forever exempt from the purview of robots. Yet is venture capital really more multivariate than manufacturing, biology, chemistry, quantum mechanics or many of the other realms HPC tackles year after year? What makes venture investors immune? Why are they so special?
The field of Psychology is perhaps a worthy analogy to venture capital. Equally amorphous and intangible, psychology is divided between clinical methodologies (relying on human judgment and subjective analysis) and mechanical methodologies (relying on statistics, algorithms and other more objective tools). More than 136 different studies have tested the relative accuracy of both methods going as far back as the early 1900s, almost invariably concluding that mechanical methods are more consistent, accurate and yield higher quality results.[i]
Even in Blink[ii], a book often held up in defense of intuition, author Malcolm Gladwell goes to lengths to call out the limitations of intuition, such as inaccuracy and cognitive bias. It’s been found that even simple checklists (the most basic of objective tools) can reduce hospital surgical mortality by around half. Half![iii] How’s that for a little data-orientation?
With so much at stake in venture capital, can HPC lend more of a hand? Perhaps it’s time for venture capitalists to follow their Wall Street cousins and further explore the benefits of HPC. Predictive analytics, multivariate simulations, complex adaptive systems and agent based modeling are just a few examples of how HPC may give venture investors a leg up. We’d all like to see money more efficiently find its way to the very best ventures (and vice versa), which can create jobs, technological breakthrough and invigorate entire ecosystems. In that spirit, we may all benefit from a little more robo-venture capital as we look towards the decade ahead.
[i] Grove & Meehl, Comparative Efficiency of Informal (Subjective, Impressionistic) and Formal (Mechanical, Algorithmic) Prediction Procedures: The Clinical-Statistical Controversy, Psychology, Public Policy, and Law (1996).
[ii] Gladwell, Blink: The Power of Thinking Without Thinking, Little, Brown and Company (2005)
[iii] Haynes, Weiser et al, A Surgical Safety Checklist to Reduce Morbitidy and Mortality in a Global Population, The New England Journal of Medicine, Volume 360:491-499 (January 29, 2009)