Greed Is Good….

Could VC’s twist from their own petard?

The first time I saw a venture capitalist was in 1974. Florsheim shoes, Brooks Brothers suit, school tie, horn-rimmed, tortoise shell glasses and according to my sources a pedigree that read like the marketing brochure from a trendy Cambridge sperm-bank: yada yada Harvard, yada yada Stanford, yada yada MIT, yada yada Phd, yada yada VC. A gen-u-ine, bon-a-fide MoU. Master of the Universe. The real deal. Not one of these Allbirds sportin’, hoodie wearin’, Sand Hill Road pretenders.

Back in the day, getting rich just wasn’t about being lucky; you had to know your stuff and, of course, a lot of rich people who could afford to lose a lot of money. Some things don’t change. But, if you wanted to play this game, the barrier to entry was pretty high. Well, take heart, Pilgrim, thanks to a bunch of wizards from MIT, the secrets of VC success have been revealed and yes! you, even you, Pilgrim, can now be a Master of the Universe.

Pick up the phone, order now and we’ll throw in a set of Ginsu knives. Operators are standing by.

Winning!

Recently, in a paper entitled “Picking Winners: A Framework For Venture Capital Investment” authors David Scott Hunter and Tauhid Zaman of MIT explicate a model and method for selecting the best candidates for emerging company investments, a group they dub “winners”. For those of you mathematically inclined or plagued by insomnia you can find the full paper here.

It’s well known in the venture community that VC investments adhere to their own, very pesky power law. And lately it seems that power laws keep popping up no matter where you look. We found them lurking in innovation (see Innovation: Power Laws – March 2017), and in social media (see Social Subsidization and Diminishing Returns – March 2015) and in micro-blogger content (see Big Bets and Long Tails – June 2013). Turns out, when it comes to venture capital, for every single company that yields a large return there are dozens if not hundreds that simply and assuredly go bust and become the statistics that make up the long, flat tail of the VC power law. But what if you could “scan” all the emerging opportunities and using a set of proven criteria put all your money on the winners from the get go?

According to Hunter and Zaman their framework does just that. Using hypothetical portfolios they believe they can achieve exit rates or positive liquidation events of 60% or greater, nearly double the yield of established venture capital firms. If true, the return to limited partners of venture funds would sky rocket. But, assuming that’s the case, why would you need a venture capital fund in the first place?

It pays to be greedy

A significant portion of the Hunter-Zaman framework relies on the use of a specific type of computational method found in covering problems. A covering problem is one where a combinatorial structure subsumes and creates a proxy for the influence of other factors. They further parse the problem of picking “winners” into a sub-modular optimization that employs a “greedy” algorithmic paradigm.

Stay with me, we might even get to have a little fun with this.

A “greedy” algorithmic paradigm is one where for each set of criteria or circumstance considered, an optimal outcome is established and employed in the selection of outcomes considered as part of a comprehensive solution to a multi-part problem set. The graphic at the top of this post is, in fact, the plot of a greedy algorithm. Go back and have a look at it. It is a “winner take all” tokenization of a specific set of variables, factors or circumstances where values are selected in order to achieve the largest marginal increase to a specific objective function. This approach is only suited to particular types of problems and can have adverse results depending upon the type of problem being solved.

Turns out, when exercising their framework, the MIT guys focused on a very few criteria including sector, investor and leadership employing a limited number mostly public data sources, Crunchbase, Pitchbook and LinkedIn and achieved some startling results.

One of, if not the key finding of this framework resulted from the analysis of the Brownian motion that results from plotting the various funding rounds from the start-ups that were included in the Hunter-Zaman analysis.

We are all familiar with venture based funding rounds from the initial or seed funding, through the subsequent rounds usually lettered A-F, to the last or exit round characterized by some form of liquidity event whether that be the sale of the start-up to another entity or its re-capitalization through an initial public offering.

Similarly, most of us are at least somewhat familiar with Brownian motion, the underlying mechanism that imparts the observed transit of phenomenon through specific intervals of time, where the motion of such phenomenon is random and the popular basis for lots of vintage screen savers. For those of you who would like a refresher here’s a look.

 

 

After plotting the drift and volatility, which one might liken to distance v direction per interval of funding events, although this is not an entirely faithful simile, the Hunter-Zaman framework produced an interesting observation. The Brownian plots for entities that successfully achieved an exit, the ones that produce the “winning” returns craved by their financial backers, exhibited a moderate amount of drift, distance traveled per interval but higher than average rates of volatility, changes in vector due to subsequent funding rounds.

Now it would stand to reason that exit bound start-ups would exhibit this kind of pattern. After all, start-ups that go from Seed to Series A to Kaput aren’t really around long enough to exhibit a whole lot of Brownian motion.

Follow the Money

Several years ago, I suggested to the head of a rather large law firm that technology could and would ultimately replace him and his business model to which he replied, “jerks like you having been trying to screw this up for years but nothing ever happens” (see Platform Strategies in the Age of AI – August 2016). The same could be said for venture capital. There have been numerous attempts at reducing the VC industry to a set of reproducible formulas, incubators, institutes and fool-proof proprietary models and yet nothing has really changed.

Many still argue that the “value” of venture firms will forever remain intangible, resting in the institutional knowledge that accrues through understanding the subtleties of capital allocation, company formation, industry lifecycles, the lessons learned through deal making and team building, comprehending the mind and the dreams of entrepreneurs and spending quality time in the flat, failing doldrums of the power curve, the 80% or more of VC investments that never pay off.

However, if you look at the legendary unicorns that fuel these venture dream machines, firms like Google and Facebook, Uber and Pinterest, none of this mythology seems to factor in. The industries were new. Prior to their entry they didn’t really exist. The teams were new. Prior to their entry they had no track record to speak of. So was it the “wisdom” accrued by their financial backers that made them successful or was it simply a matter of numbers? Had the slot machine been played long enough to finally hit? Power laws would seem to suggest it’s more the latter than the former.

Similarly, the algorithmic paradigm employed in the Hunter-Zaman solution seems to suggest that the greatest value contributed by traditional VC firms is to “scrub” opportunities based on the risk attributes of well known factors, what one might consider to be a form of institutionalized confirmation bias, codified by subsequent funding rounds and, consequentially, built into the MIT model as well. In the venture world success begets success. So start-ups whose prospects brighten will attract more investment and enjoy more funding especially as the probability of a successful exit grows larger. Big company management teams are also notorious for employing this same technique. However, larger entities usually employ confirmation bias when eschewing promising unknown opportunities in favor of pursuing familiar well-known adjacencies; an all too human phenomenon that many are starting to believe will become the achilles heel of AI and not just the predilection of venture firms and large technology companies.

Interestingly, the venture capital industry has largely been a hold out when it comes to the use of modern techniques to improve the performance of their funds. And who could blame them. Once you pull back the curtain, it might be easy to see that the wizard isn’t wearing any pants and for all too familiar reasons.

Recently, the World Economic Forum published a glimpse of exactly what might lie behind the curtain. Back in 2009, Businessweek asked the Quid AI group and its CEO Bob Goodson to pick 50 start-up candidates that would emerge at the top of their class. Surprisingly, 20 percent of the fifty candidates they selected have gone on to achieve billion dollar valuations. To put that in perspective most VC’s have an average limited partner IRR of 20% and only a negligible percentage of most VC portfolios produce anywhere near a 20% return let alone a billion dollar valuation. Curiously enough, the Quad AI selection criteria were very similar to the criteria employed by the MIT effort where rounds in, days between rounds, and founders networks were key in selecting candidates. So to suggest that these methods, sophisticated as they might be, are betting on unknown entities, which is where the venture funnel starts, would be an inaccurate characterization.

The Secret Sauce

There are several factors that suggest that the MIT and Quad AI work is itself part of an emerging trend, one that could soon be a juggernaut. While things like the JOBS Act has served to change capital formation, the maturation of public and private data sets such as S&P Capital, Crunchbase, ThompsonOne, Pitchbook, DowJones Venture Source, LinkedIn, Cambridge Associates among others has finally reached a tipping point, one where predictive analytics proves practical as well. Imagine that. All those VC investments in predictive analytics finally pay off (see Sis, Boom, Bah! – December 2015)

Given these circumstances, why would an endowment or pension fund rely on a costly, antiquated venture model when they could make direct investments in crowd-funded portfolios or tap directly into private or corporate portfolios? Similarly, corporate venture players who are currently staging a bit of a comeback could employ these models to increase their odds of success when investing in unproven opportunities. Likewise hedge funds could remove the hocus-pocus from early stage investments altogether and simply promise a straight forward IRR from a transparent, managed portfolio of emerging companies.

It could well be that disruption, a concept well-known to the well-healed Masters of the venture moneyball universe, has finally gotten around to paying their own industry a long over due visit.

But then again, some jerks have been trying to screw this up for years but nothing ever happens.

 

Graphic, a “greedy” plot, courtesy of Wikipedia