Why machine learning falls short in early stage venture capital

There’s asymmetrical access to data in venture capital

Quant hedge funds have access to the same market data for publicly traded companies through official filings and live feeds from the exchanges. Competition in the quant hedge fund business is therefore based on which funds can better interpret the data to achieve desired outcomes. In venture capital, the situation is very different. Firms will have ad-hoc access to information based on their relationship with founders, and some firms have better access than others.

Machine learning is past-driven, venture capital is future-driven

Once you have the data, understanding which parts are informative is very difficult. Mike Maples at Floodgate has talked extensively about how their team looks for technology waves that can help companies grow to very large scale. The returns generated today are the results of technology trends started five to ten years ago, and in some cases even twenty years ago.

Lack of liquidity in the market doesn’t let you course correct

If you deal with publicly traded securities, you can quickly get in and out of a position. As your model develops and gets more accurate, you’re able to rebalance your portfolio to increase returns. In venture capital you don’t have that luxury, which drastically reduces your margin of error. This applies to both errors of omission and errors of commission:

  • If you’re an early stage investor and your model doesn’t surface the seed/Series A round of Google/Facebook/Uber, you’ve missed the boat. That particular round isn’t going to happen again, and your portfolio construction strategy may not allow you to invest in subsequent rounds. With a stock, you can miss out on some of the appreciation, but you’d still be able to buy in the public market for many years
  • Similarly, if your model had surfaced companies that aren’t performing (bad business models that were predicted to be good, markets that took too long to develop, etc) you are not able to decide to get that money back and invest it in something else. Even if your model has now improved, you have less swings left at bat because a part of your fund has already been deployed.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store