Main image of article Can Big Data Pick a Winning Startup?
Can you predict whether a startup will collapse or succeed? Wall Street and Silicon Valley would certainly like to think so. According to Wired, a handful of independent entrepreneurs and big companies are using data analytics to better forecast whether that hot startup will become the next Uber, or merely the next Click here for analytics jobs. Thomas Thurston, who runs a research firm called Growth Science, told the magazine that his algorithmic simulations accurately predict 88 percent of the time whether a startup will implode within the next five years. He also claims a 66 percent success rate when it comes to guessing which startups will survive during that same time period. Despite Thurston’s reliance on algorithms, his model still requires a certain amount of old-fashioned intuition—a human being needs to decide whether a company qualifies as a fast follower or a first mover, for example—but his team is working on making everything as scientific as possible. He’s also not the only one working on a startup-prediction project: Google Ventures and other investment funds rely on their own formulas.

Upload Your ResumeEmployers want candidates like you. Upload your resume. Show them you're awesome.

Google Ventures’ track record is pretty impressive, with investments in Cloudera, Uber, Nest (eventually acquired by Google), and other name-brand companies; but given the firm’s secrecy, it’s difficult to tell how many of those investments hinged to whatever degree on a predictive algorithm. California-based Correlation Ventures, another fund, boasts of making “algorithm-driven investments” in startups, which often results in funding companies well outside of the traditional tech-power corridors of Silicon Valley and Silicon Alley. And no matter how advanced the math becomes, at least a subset of investors and advisors will always argue that their gut is superior to an algorithm when it comes to choosing startups in which to invest.

Related Articles

Image: solarseven/