AGILEVC My idle thoughts on tech startups

September 15, 2009

We’ve all heard the phrase “Past performance is not necessarily indicative of future results” in some context or other. It’s a common disclosure in the investment world that’s either mandated by regulators or simply made as a prudent footnote when presenting historical investment performance.

In the world of venture capital though, conventional wisdom is that the opposite is actually true… that past performance is one of the best, if not the chief indicator that a particular VC firm will be successful at delivering strong returns in the future. Conventional wisdom in this industry is that the funds most likely to deliver top quartile returns are the ones raised by firms that had prior funds with top quartile returns.
Top quartile performance is important no matter how you slice it. Various empirical studies of large numbers of venture funds from different periods in time have shown that for all the funds formed in a particular year (“vintage” year in the VC vernacular), the overwhelming bulk of returns for the group can be attributed to the top 25% of funds. This isn’t surprising really… yet another substantiation of Pareto’s 80 / 20 rule.
But there’s an interesting white paper from Landmark Partners, a fund of funds active in the PE secondary market, based on a comprehensive study of VC funds. They outline their findings here, but have elected not to publish the paper publicly. The paper covers a range of topics but what struck me most in reading it was their conclusions about performance persistence in venture capital… this notion that the best indicator that Firm X will deliver top quartile returns in the future is prior funds with top quartile performance.
Because it takes a long time for a venture capital fund to “realize” the returns from all the underlying startups it invests in, a potential LP has to examine not just a VC firm’s most recent fund but perhaps several funds prior. If the current fund being raised is Fund N, the immediately prior is N-1, the one preceding that N-2, and so on. There’s of course plenty of variance across firms and funds, but in general VCs invest their funds within the first 3-4 years of raising them so fund N-1 is probably 3-4yrs old, N-2 is probably 6-8yrs old, N-3 probably 9-12yrs old. At that age, a potential LP can have near certainty on what Fund N-3′s ultimate returns will be and a high degree of certainty about N-2, though only modest certainty about what the outcome of N-1 will be.
So again, conventional wisdom holds that if VC Firm X’s Fund N-2 or Fund N-3 are top quartile then Fund N has a very good chance of being top quartile. Without divulging the full content of their paper, Landmark basically concluded that the peformance of N-2 and N-3 were not terribly indicative of future performance. If VC Firm X’s old Fund N-2 was in the top half (not even top quartile) for its vintage year, the odds that Fund N would ultimately perform in the top half of its vintage are about 55%. If N-3 was top half the odds that N would be top half fell to 52%. In other words… looking at the performance of these prior funds is about as useful a factor in predicting Fund N’s performance as flipping a coin.

To be clear, I’m not suggesting that past performance is meaningless or that if VC Firm X has had terrible prior returns it isn’t a signal of the investment teams’ underlying ability. Also there’s a pretty clear echelon of VC investors (call it the top 2-3%) who seem to be able to deliver exceptional returns over several decades. But it was startling to me to see how little past performance was indiciative of future results in the VC industry, that the top performers of tomorrow won’t necessarily be the top performers of yesteryear. Has to make you wonder a bit about that conventional wisdom…