Forecasting in venture capital

Good VCs need to be good forecasters. We need to predict which companies are going to succeed and second guess future trends so we can develop our firms to stay ahead of the curve. Deloitte have just published an interesting synopsis of Tetlock’s Superforecasting which gives us some insights into the types of people we need to recruit and the habits we need to cultivate to be good at the forecasting game.

The main and most surprising insight is that deep expertise doesn’t produce more accurate predictions. Tetlock ran prediction competitions over five years with multiple teams and found that the best predictions came from people with the following characteristics:

  • broad expertise
  • open minded
  • sceptical of deterministic theories
  • cautious in their forecasts
  • quick to adjust their ideas as events change
  • embrace complexity
  • comfortable with a sense of doubt
  • highly numerate (but don’t use sophisticated mathematical models)
  • reflective
  • learn from their mistakes

This is a good list for VCs too. One thing that stands out as a little different for me is that the best investors get behind big themes – e.g. the internet, open source software, SaaS, ecommerce, mobile, marketplaces – which feel a bit like the deterministic theories that super forecasters are sceptical of. However, even with these it’s important to keep an open mind and back off quickly if they aren’t playing out as planned. Mobile is a great example. As a category it’s yielded some amazing companies and investments, but many investors went too early and lost money – me included. I made my first mobile internet investment in 2000 in a business that was years ahead of it’s time helping banks to get their services on WAP phones. I’m not saying I’m a super forecaster, but I did learn from that and backed off from mobile until after the iPhone.

The other interesting point from Tetlock is that prediction skills can be improved by good sharing and debating within teams and by training focused on thinking in terms of probabilities and removing thinking biases.