Money Ball was a good movie (I have to admit to not having read the book) with a crucial theme: statistics beats human intuition in complex situations. With the predictions for the 2012 elections we saw the same theme play itself out as Nate Silver crushed the pundits by using data. There are is also a fantastic example in Daniel Kahneman’s Thinking, Fast and Slow (which I did read) of a relatively simple data based model outperforming all “experts” when it comes to predicting the future prices of French wines. As Kahneman points out a lot of this has to do with the brain’s preference for a coherent and entertaining story (system one) over spending the time and effort to crunch numbers (system two). Pundits make for better television than computers but it doesn’t make them right – instead they tend to attribute too much weight to their unsubstantiated hunches.

That of course brings us to another situation of complexity in which telling a good story may be the difference between a successful and failed pitch (get it?): venture capital. Is it possible to bring Billy Beane’s approach to investing in startups? Someone wrote a tongue-and-cheek post about this on GigaOm late last year titled “The secret algorithm one VC firm uses to pick entrepreneurs.” It was somewhat funny but missed the point – what if it’s not about the team? What if the statistics you need to pay attention to are the ones that show traction of the actual product? Obviously this assumes there is a first version of the product so this does not really apply to the earliest stages of investing.

Once a product has launched though, there is a ton that can be measured about user engagement, network growth and density (if there is a network effect), social media mentions, Google queries (via Google Trends) and so on. With the trend towards self service on-ramps this is even true for enterprise or B2B products. My current working hypothesis is that this type of data is far more predictive of the success of a startup than a highly qualitative assessment of the founding team. Having said that there appears to be some statistical evidence that repeat entrepreneurs are more successful *if* they stay in the same field (I unfortunately can’t find the link to the study right now).

We have been using some of these quantitative techniques at USV for a while but informally. For instance, we often ask startups for their analytics logins during due diligence and compare what we find there to patterns from our successful investments. This morning we will be meeting to discuss how to systematize this effort and support it with some actual technology so that we can more easily discover and track new companies.

Money Ball was a good movie (I have to admit to not having read the book) with a crucial theme: statistics beats human intuition in complex situations. With the predictions for the 2012 elections we saw the same theme play itself out as Nate Silver crushed the pundits by using data. There are is also a fantastic example in Daniel Kahneman’s Thinking, Fast and Slow (which I did read) of a relatively simple data based model outperforming all “experts” when it comes to predicting the future prices of French wines. As Kahneman points out a lot of this has to do with the brain’s preference for a coherent and entertaining story (system one) over spending the time and effort to crunch numbers (system two). Pundits make for better television than computers but it doesn’t make them right – instead they tend to attribute too much weight to their unsubstantiated hunches.

That of course brings us to another situation of complexity in which telling a good story may be the difference between a successful and failed pitch (get it?): venture capital. Is it possible to bring Billy Beane’s approach to investing in startups? Someone wrote a tongue-and-cheek post about this on GigaOm late last year titled “The secret algorithm one VC firm uses to pick entrepreneurs.” It was somewhat funny but missed the point – what if it’s not about the team? What if the statistics you need to pay attention to are the ones that show traction of the actual product? Obviously this assumes there is a first version of the product so this does not really apply to the earliest stages of investing.

Once a product has launched though, there is a ton that can be measured about user engagement, network growth and density (if there is a network effect), social media mentions, Google queries (via Google Trends) and so on. With the trend towards self service on-ramps this is even true for enterprise or B2B products. My current working hypothesis is that this type of data is far more predictive of the success of a startup than a highly qualitative assessment of the founding team. Having said that there appears to be some statistical evidence that repeat entrepreneurs are more successful *if* they stay in the same field (I unfortunately can’t find the link to the study right now).

We have been using some of these quantitative techniques at USV for a while but informally. For instance, we often ask startups for their analytics logins during due diligence and compare what we find there to patterns from our successful investments. This morning we will be meeting to discuss how to systematize this effort and support it with some actual technology so that we can more easily discover and track new companies.