Friday, September 23, 2011 at 5:28 pm.
Moneyball for tech startups
If you haven’t read the book or seen the trailer, the basic idea is: Oakland Athletics general manager Billy Beane — played in the film by Brad Pitt — turned the baseball industry on its head by using objectivity and data to help pick a baseball team, instead of subjectivity and gut. It was controversial and not perfect, but he did a pretty good job at it.
As I was re-watching the Moneyball trailer this week, I wondered: Could this same principle be applied to investing in tech startups? Tech is, after all, an inherently data-driven industry, but much of the talk around investment seems to be about relationships and personality. With the recent boom in startups and smaller investment vehicles, you’d think that data- and value-driven startup investing — looking for specific indicators and trying to exploit them — could be an opportunity.
So I asked four of the top investors in technology what they thought about this: Fred Wilson from Union Square Ventures, Chris Dixon from Founder Collective, Paul Graham from Y Combinator, and Ben Horowitz from Andreessen Horowitz. I figured that if something new was working, these four — a mix of East and West Coast, early and later stage investors — might have heard about it.
Fred Wilson said:
“We have not been able to quantify it. We haven’t even tried. Although I am sure someone could do it and they might be very successful with it.
To us, the ideal founding team is one supremely talented product oriented founder and one, two, or three strong developers, and nothing else. The supremely talented product oriented founder should have been obsessed about a product area/idea for a long period of time and just has to build something to satisfy their passion/curiosity. That’s about it. Joshua Schachter/Delicious, Jack Dorsey/Twitter, Dennis Crowley/Foursquare are the iconic examples of this kind of person in our portfolio.”
Chris Dixon said:
“One of the main activities of good investors is trying to find ‘accurate contrarian theses’ about what make good startups, markets, founders etc. So there is a lot of Moneyball-esque activity. I’ve seen a few attempts to do it quantitatively (I recall an academic paper on it and also some studies done internally at VCs) but I think those are often flawed because the quantitatively measurable things are either obvious (e.g. founders who sold their last company for a boatload of money are more likely to be successful than founders who failed), irrelevant, or suffer from ‘overfitting’ (finding patterns in the past that don’t carry forward in the future).
Personally, I think the biggest ‘Moneyball’ opportunities in seed investing are around the processes used. For example, I think the format of spending a few hours getting ‘pitched’ is a deeply flawed process for getting to know whether a first time founder will be successful. You can think of [Y Combinator] as an example of trying a new process. I’m personally constantly experimenting with different ‘getting to know founders’ processes.”
Paul Graham said:
“I know of no reputable investor who invests based on data. I once heard of someone who planned to, but I forget who it was; probably nothing came of it. [...]
We are the far opposite end of the spectrum from an analytical approach. We decide based on gut feel after a 10 minute convo. It may seem ironic that we who have the most data make the least use of data. But perhaps not: perhaps it’s because we have so much data that we know it all comes down to the personalities of the founders. Or maybe we’re just lazy.”
Ben Horowitz pointed me to a company called YouNoodle, which, according to a 2008 TechCrunch article, “aimed to tell investors how likely they would be to see a return on their investment, and would even go so far as to estimate the exact valuation of a startup a few years into the future.” (It obviously didn’t work very well, predicting an $87 million valuation for TechCrunch; AOL bought it for far less.) YouNoodle has since changed its focus, and is now a “global entrepreneurship network.”
So, what’s the consensus? It seems that this sort of technique is possible. And given that four of the most important tech investors in the world seem skeptical of it, if someone can figure out a good formula that works, they may be able to exploit it.
The major challenges seem to include:
- A lack of good data (unlike baseball, where stats for some players go back to Little League)
- A more complex and different “game” than baseball, which is played for 162 games per year and has clear, objective winners and losers, unlike tech startups which can build for years before they win or lose
- A “draft” in baseball that doesn’t exist in tech, unless programs like Y Combinator and TechStars effectively turn into that sort of thing
- Different incentives to work together than in baseball
Still, it’s a fascinating idea. I’d love to hear more about people’s experiences with the topic, and what the future may bring. But first, I’m off to see Moneyball.
Follow-up: More Moneyball for tech startups