Venture Capital

Thoughts on quant approaches to early stage investing

By August 15, 2013 8 Comments

‘Quant VC’ is a hot topic at the moment. Techcrunch and VCs have been writing about it, VC funds have been hiring data scientists, and Kleiner Perkins has taken the step of branding its proprietary data mining tool as Dragnet. There is even a company called Mattermark with the tagline ‘Big data meets venture capital’.

All of this activity is happening for the same reason as in many other industries: there is now a lot of data that can be used to support decisions. Data sources to be mined include AngelList, Crunchbase, VentureSource, the App Store, the Google Play Store, the Facebook Platform,  Klout scores, Twitter mentions, LinkedIn.

There’s a lot of good insight to be drawn from this data, but nobody (with the possible exception of Mattermark) is saying it’s possible to build a predictive model that estimates a startup’s chances of success. That’s mostly because the world is changing too fast. Startup success generally takes five to seven years but technology and markets have moved on too far for repeating a formula that worked five years ago to make sense. 2006-2008 was a mini-bubble period where gen 1 social networks Bebo, Myspace and Facebook were the big story and investor interest in business models was at an all time low.

So rather than predicting success the data is used to support decision making by surfacing trends and prospective investments and for analysing competition.