Excuse my limited knowledge of statistics, but I thought the mean is considered an improper metric for the "average" value of asymmetric distributions.
If not mean, nor median, what metric would you suggest to approximate a typical value?
If not mean, nor median, what metric would you suggest to approximate a typical value?
* 30 return 0.5X (i.e. half of the initial investment)
* 30 return 1X
* 25 return 3X
* 10 return 6X
* 4 return 10X
* 1 returns 20X
The fund class overall returns 2.4X, but the median fund is very underwhelming (investors just get their money back). The "average" fund return (2.4x) is also kind of underwhelming because that's so much worse than the top funds. However, if an investor either a) has broad exposure to multiple VC funds or b) has some insight that helps them pick out the top 20% or 40% of fund managers, then the asset class is a pretty good investment. I think a good evaluation metric for VC as an asset class would a combination of expected returns and variance, and how those two quantities compare to other assets like public stocks or bonds.
a) Some bigger LPs diversify across multiple VC firms and smaller LPs may access fund of funds (though very expensive). Either way broad diversification across VC funds is not significant in the industry. One reason is because VC fund returns and broad exposure to multiple VCs is not really the main point here - it is the startups behind the VC layer that generate returns to LPs. Thus VC firms themselves are the conduit by which LPs get diversified exposure to the growth/returns of multiple startups, which is the goal for this asset class. Diversifying on top of your diversification gets expensive and impractical.
I agree with you risk adjusting returns is important. Sharpe ratios would use standard deviation and historical returns. Of course caveat emptor "past performance does not necessarily predict future results." https://en.wikipedia.org/wiki/Sharpe_ratio
There is no official must do approach to looking at VC performance. In practice median and mean are both used, along with some other measures that look at consistency and write off ratios etc...
For those interested one can google search for a professional VC analysis to see what they do. https://www.preqin.com might have something not behind a paywall.
a) I generally agree. I was just trying to illustrate why mean and median aren't great for analyzing asymmetric distributions.
Why not just call it like it is...VC is really just a people business, there's money involved, but trying to quantify the process at all can be misleading. A VC firm is a small group of people (partners) using their best judgement and experience to find another group of people (founders) worth investing in. That's it, there's a ton of key man risk and there's no secret sauce.
Lowercase and Matt Mazzeo are a great example: VCs are basically just like Hollywood talent agents who get to find the next movie stars and help them along a bit - but in VC they are finding the next big tech founders instead of actors. These are people businesses, numbers can't really capture it but they can distract.
I didn't mean to get into a statistics debate on here. The great thing about VC and startups is everyone gets to be right until they're not :)
Sidenote: I hear you guys are running one the best shops in the space. Congrats on the recent close. Best of luck to you and your portfolio co's.
E.g., in coin flipping, if assign 0 to heads and 1 to tails and flip coins for a long time and take the empirical average, then that average coverges to the mean of 0 and 1, that is, 1/2. Of course, here the 1/2 is not even a typical value.
Similarly, to estimate what a venture firm will return in the long run, just take the average of what they have returned so far.
It entirely depends on what you are looking to explain or describe.
Mean is appropriate for investment returns if you can diversify risk over many investments. Median is appropriate for looking at incomes in terms of social policy. 95th or 99th percentile might be appropriate when you are looking to ensure that your web server provides reliable response times.