Gambling Is Not A Thesis: Early-Stage VC Investment Strategy
By Avraham Shisgal
Venture Partner @ NYVC
January 2021
Imagine the following scenario: A fifteen-year follow-up study of 100,000 high-school grads from Duckburg, a quaint town in Calisota, found that 15% turned out to be very successful and wealthy, 15% achieved moderate success and wealth, 35% were hard-working middle-income employees with very little wealth, and the remaining 35% were unable to keep a job and had a negative net worth, with many of the latter turning to crime. The Duckburg study included subjects from many families and high-schools, but the data only provided the level of success achieved by individual unidentified subjects.
In an effort to arrive at an ideal number of children per family, a Monte Carlo Simulation of the data from the Duckburg study was used to slice the larger dataset into smaller groups of various sizes. The results of the Monte Carlo Simulation suggested that, if a family desires to produce at least one highly successful child, such family must raise at least ten children and preferably twenty or more children. The Monte Carlo Simulation found that a family’s chances of raising a highly successful child diminish with every child less than a total of ten and increases the more children the family raises.
A Monte Carlo Simulation is, in this case, a number-crunching exercise that takes the large aggregate dataset, calculates the random probability of different outcomes based on the larger population’s statistics, and applies the random effect probability to various size randomly selected groups taken from the larger dataset. It does not take much acumen to realize that the suggestion proposed by the Monte Carlo Simulation is, in fact, counterproductive. This is because a Monte Carlo Simulation uses a random sampling of independent random outcomes, whereas the science and art of raising successful children is, in most cases, a deterministic system.
A deterministic system is such that it will produce similar outputs from similar starting conditions, provided that certain controls or factors have been applied to the system. In the context of the subject matter of the Duckburg study, the formula for raising successful children is a nuanced multivariate mixed system in which multiple fixed or deterministic factors that are within our control (personal role modeling, quality parenting and education, safe and healthy upbringing, extracurricular opportunities, access to study aides, etc.) combined with, but to an often significantly lesser extent, multiple random-effect factors that are, to varying degrees, out of our control (genes, personality, external influences and circumstances, illnesses, etc.) affect the outcome of our childrearing efforts.
In a completely random stochastic system, such as, for example, a statistical analysis (rather than a calculation of mathematical odds) of the results of many spins of a roulette wheel, a Monte Carlo Simulation should accurately identify a valid strategy for increasing the probability or certainty of landing on a desired number. But in matters such as raising children and early-stage investing in venture capital portfolio companies, deterministic systems in which outcomes are affected by the investment of time and capital and by astute decision making before and throughout the process, spreading oneself too thin by raising too many children or randomly investing in too many startups — would in fact result not in an increase but in a decrease in one’s likelihood of achieving a successful outcome. In venture capital investing, the individual variation in the likelihood of success of any given portfolio company is affected by deterministic effects (quality, strategy, potential, execution, etc.) and by random effects (market, competition, platform, geopolitical, luck, etc.). Treating childrearing or venture capital investing as pure gambling would thus in fact yield less than desirable outcomes.
As I’ve mentioned in “‘Venture Capitalism’ Makes the World a Better Place”, because a third of early-stage VC investments, unfortunately and unintentionally, become worthless and are written off and another third, unfortunately as well, yield low returns that do not make up for losses on other port-co’s, VC strategy mandates the pursuit of mega outlier returns. I emphasize ‘unfortunately and unintentionally’ to clarify that a VC fund manager does not have the luxury of investing in startups that are anything less than likely to achieve disproportionately outsized returns. Instead, prudent early and follow-on VC strategy requires investments to be made exclusively in startups that present a realistic likelihood of achieving at least a $500 million exit. Considering that, even with this lofty aim, 80% of portfolio investments will fail, in part or in full, the overall success of an early-stage VC fund rests on the success of its top quintile port-cos. This strategy is a variation on the classic “shoot for the moon” adage. In early-stage VC investing you must aim for the stars with every investment, not because a lesser outcome will suffice but because we know that eight out of ten portfolio companies will fall short of yielding an acceptable return on investment and many will fail to even preserve the capital invested in them. With this strategy, and with an adequate follow-on investment of capital, time, and resources in the most promising port-co’s, optimistically, 20% of port-cos will achieve outsized returns sufficient to make up for other port-co losses, yielding a good return on the aggregate investments of the fund. On the other hand, randomly scattering investments with no strategy other than aimlessly squandering funds on hundreds of startups, many with no potential for runaway success, and without the ability to research or guide these startups or to make significant follow-on investments in port-cos that have started to succeed — is as counterproductive as raising twenty children with limited resources in an effort to increase the likelihood of, randomly, producing one or two highly successful children.
Correlation Ventures released data indicating the MOIC of 21,640 ‘financings’ it invested in portfolio companies that exited or were written-off between 2004 and 2013. The Correlation Ventures data shows that 65% of its ‘financings’ yielded a return of less than 1x MOIC, 25% an MOIC of 1–5x, 6% an MOIC of 5–10x, 2.5% an MOIC of 10–20x, 1% an MOIC of 20–50x, and 0.4% of ‘financings’ yielded an MOIC greater than 50x. An opinion piece in Institutional Investor uses a Monte Carlo Simulation of Correlation Ventures’ data and concludes that funds should have “a portfolio of at least 500 startups” and that “anything less risks having a portfolio without any mega-winners”. The Institutional Investor opinion piece suggests that a much greater portfolio diversification would result in a smaller disparity between the top performing quartile of VC funds, which yield a Net-to-LP IRR of 21%-43%, and the bottom quartile, with an IRR ranging from a 4% to a -9%. But portfolio diversification works only with passive investments where the bulk of the portfolio is expected to yield reasonable returns and only a few investments, if any, might underperform and, even then, in most cases only slightly. When you randomly buy shares in thirty or two hundred or five hundred S&P 500 companies, you do not expect to have to write off half your portfolio within two years. In the VC scene, on the other hand, as a matter of strategy most investments will unfortunately fail, sometimes completely, and the overall success of the VC fund relies on its managers’ ability not only to identify good prospects but also to actively help them develop.
Dave McClure, formerly of 500 Startups, similarly concluded that “if unicorns happen only 1–2% of the time, it logically follows that portfolio size should include a minimum of 50 to 100+ companies in order to have a reasonable shot at capturing these elusive and mythical creatures”. Seth Levine, a partner at Foundry Group, in a three-article series ( here, here, and here), and Alex Graham, a leading UK based VC-finance expert, in this article, relying on the same Correlation Ventures dataset, also argue that early-stage VC funds need to invest in a large number of portfolio companies in order to achieve a reasonable likelihood of having a minimal number of port-cos yielding outsized 50x+ MOICs. But Seth Levine and Alex Graham proposed far more reasonable numbers than the other scattershot recommendations and, as Alex Graham states, “there is a quality <> quantity trade-off in venture investing”.
Alex Graham concludes that “each investment made needs to have the potential for outsized returns” and, in a writeup on VC strategy, Fred Wilson, a partner at USV, a successful firm showing solid IRRs on multiple funds, explains that “We do not take a shotgun approach. We do not view seed investments as “options”. We only make a seed investment if we have as much conviction on the team and the opportunity as we would at the Series A round. We are as committed to our seed investments, both in terms of the time we spend with them and the willingness to follow-on in them”.
Additionally, Correlation Ventures’ data does not provide a good foundation for analysis of VC strategy. For starters, the data is unclear in that it considers each round as a ‘financing’ and does not include portfolio-company performance or ROI data. Such data could be meaningless since, for example, if a single VC fund made a $1 million investment in a port-co’s Series-A round, followed by a $5M in its Series-B round, $10M in its Series-C round, and $30M in its Series-D round, then received $70M from an exit, say a week after the Series-D round, any method used to split the $24M profit between the four ‘financing’ for the purpose of allocating and calculating an individual MOIC for each ‘financing’ — would be arbitrary.
Correlation Ventures prides itself in making rapid investment decisions using an algorithm that assesses each prospective port-co’s investment-potential, but its algorithm might stand to benefit from an upgrade to a multivariate AI analysis system. Correlation Ventures also typically subscribes to rounds led by other VC syndicators and does not syndicate and lead rounds itself. This means that a syndication leader, which has significantly more time to analyze the port-co and is typically an early port-co investor with an interest in securing additional funding from new investors even if the port-co may be just treading water, or worse, might occasionally entice Correlation Ventures into investments that are less than ideal in terms of risk v return.
One could analyze VC strategy to death only to end up on a wild goose chase after deficient or low-relevance data. This could lead an investor off a cliff without resulting in improved investment decisions or could even contribute to the adoption of a counterproductive investment strategy. Bad VC investment strategy can lead struggling investors to wrongly conclude that VC investing is a game-of-chance rather than an investment strategy. Investors unable to comprehend or appropriately weigh fundamental VC factors are likely to mistakenly conclude and argue that any two or twenty or two-hundred portfolio companies have the same probability of becoming successful unicorns.
It is important to have a solid grasp of VC strategy analysis, but it is also imperative to develop guiding principles that summarize this analysis into effective criteria that lets one efficiently make successful decisions. The difference between using principles and full-scale algorithmic analysis of all the factors in play is that principles are assigned more weight than other factors. Ray Dalio’s excellent book, “Principles”, is therefore relevant to VC strategy as it is to other investment strategies. Ray ran a hedge fund which succeeded at first but then failed. After learning from his mistakes, he restarted his investment management activities with great success and shared his lessons in his book.
Warren Buffett, one of the world’s most impressive investors, is famous for using simple principles, or rules. When it comes to classic discounted cashflow investing, Buffett argues that he has two rules; rule # 1 being “never lose money” and rule # 2 being “see rule # 1”. VC investing, though, is guided by a different set of rules. A portfolio of VC investments can and will produce many small losses, which are, however, expected to be more than offset by lucrative investment returns from a small number of outsized exits. The key rules in early-stage VC investing are 1.) EVERY prospective port-co must be a viable candidate for achieving a unicorn exit, and 2.) investors should invest only in what they know. Warren Buffett does not bother to consider investing in business he does not understand (including early-stage startups) but he does not shy away from making bold and determined investments in businesses with a business model he is comfortable with. As VCs, we too should only invest in startups operating in a familiar field and with business models we can understand, but we can broaden our core competencies by bringing on board VC partners with different sets of skills and experiences.
From a VC investor’s perspective, a promising baby-unicorn (an early-stage startup with a potential of growing into a unicorn) statistically has a 33% chance of failing and a 20% chance of maturing into a successful unicorn. But a prospective portfolio company that is unlikely to develop into a unicorn is significantly more likely to fail (primarily because of its lesser ability to raise additional funds in the future) and has an almost 0% likelihood of achieving a lucrative exit. Early stage investing in baby-unicorns involves investing relatively small amounts in high-risk startups, knowing that some will fail while having an informed basis to believe that some will achieve stellar success. On the other hand, random scattered investing in startups, without predetermining each startup’s likelihood of achieving a lucrative exit, is a higher-risk low-return strategy with worse odds than those offered to gamblers at a casino.
Good venture capitalists, like good team-sports athletes or port-co founders, have an innate or acquired skill-set that allows them to identify critical factors, ignore the noise, i.e., the plethora of non-critical factors and half-baked opinions, and execute with a purpose at times of uncertainty despite being under pressure. It is important to focus in VC strategy on fundamental factors critical to success and to properly assign less weight to less critical factors, considering the latter but not letting them distract us from the core principles of our strategy and vision.
VC investing in a single startup is risky business; it comes with a very high degree of risk and a realistic opportunity for achieving incredible returns. However, VCs will not succeed merely by investing in more startups but rather by investing in better high-risk / high-reward startups. In fact, the more successful VC firms are very selective even when making seed investments. IVP, with $7B in commitments, for 39 years since its founding, has provided investors with an IRR exceeding 43%. IVP has only invested in approximately 400 companies, and 115 of them have gone public (see www.IVP.com). IVP, however, is currently focusing on late-stage growth-equity VC strategy, investing in already successful low-risk startups. But successful early-stage (seed, Series-A, B, and follow-on) firms also exercise discipline when selecting portfolio investments. Union Square Ventures (USV) is one such firm. USV probably invested in a few hundred startups since its 2003 founding but, with funds ranging from $125 to $175 million and with their selective approach to seed investing, USV clearly did not “take a shotgun approach”, as Fred Wilson explained. According to a 2019 article by Zoe Bernard, three USV funds (raised in 2012, 2014, and 2016) had achieved net IRRs of 32%, 21%, and 32%, respectively. According to Pitchbook, USV’s 2010 Opportunity Fund, a fund used by USV for making follow-on investments in portfolio companies of its other funds, had a calculated IRR of 60.59%.
Correlation Ventures’ data looks at the performance distribution of investments made by a single, relatively small, VC fund — Correlation Ventures. But another dataset, from a study by Horsley Bridge, a fund of funds, shows the performance distribution by port-cc within multiple VC funds. The most important takeaway from the Horsley Bridge data is that all VC firms have a similar percentage of failing investments. In fact, outperforming VC funds had a slightly higher percentage of failed investments than the underperforming VC funds had. What distinguished successful VC funds from failed VC funds was not having investing in more portfolio companies nor having made fewer “bad investments” but rather their disciplined focus on finding baby-unicorns with outsized upside potential and helping these baby-unicorns develop and grow into mature ones.
Some would argue that better performing funds attract better prospects but that is not necessarily the case. Many unicorns received early funding from lesser-known VC funds and some maiden funds have invested in unicorns. Additionally, some smaller funds have admittedly turned away startups that eventually became unicorn.
In summary, effective early-stage VC investing is based on portfolio science and disciplined strategy, not on guessing and speculation. Properly executed, early-stage VC funds can and often do consistently outperform all other asset classes typically relied upon by institutional investors for achieving their investment performance objectives. Well-performing early-stage VC funds tend to even outperform late-stage growth equity VC funds (Figure 3: Top-Performing VC Funds, Cambridge Associates, and Table 3: First-Time VC Funds, PREQIN). Top-performing institutional investors, whose investment teams are not intimidated by VC strategy, have been increasing their VC asset class allocations at rates higher than those pursued by underperforming institutional investors (Figure 2: VC Allocation, Cambridge Associates).
