Venture capitalists often talk about using pattern recognition, a fancy way of saying that they look for similarities between startups and other companies that have previously seen wild success. 

In a new paper, Massachusetts Institute of Technology economists took a more scientific approach to predicting the kinds of startups that might one day become the next Googles and Facebooks of the world.

"The big, big, big question that we had was, how can you measure the quality of startups before they achieve an outcome?" says Jorge Guzman, a doctoral student at the MIT Sloan School of Management and lead author of the paper.

Guzman applied algorithmic pattern recognition to for-profit business registrants in California from 2001 to 2006. From 70 percent of those companies, he extracted the most common traits that young startups exhibited before they were acquired or went public. 

Using the standard procedure for verifying the quality of a predictive algorithm, he then checked these traits against the other 30 percent of the sample to see if the algorithm accurately predicted startup quality. It did with a 75 percent success rate.

So what did Guzman find? 

  • Companies that have names that are short and/or contain words associated with technology are more successful than those that don't.
  • Young corporations are more than six times more likely to grow than companies that aren't incorporated.
  • Startups that obtain a patent are more than 25 percent more likely to experience growth. 

Guzman admits that the results are somewhat superficial. The algorithm is limited by the fact that there's only so much public data to work with when studying private companies. But the demonstration could have important implications for the future. 

Believe it or not, this doesn't involve helping VCs with their jobs, or eliminating them, for that matter. Guzman sees this method eventually helping communities, as opposed to individual startups or venture capital firms. 

"A very direct application of this that we hope to kickstart within the next month or two is that we're going to use business logic to start creating an ability for cities and regions around the world to start tracking and understanding their innovation-driven entrepreneurship initiatives," he says.

Guzman found that startups located in Silicon Valley are 60 percent more likely to experience growth compared with companies based in other areas of California. Going forward, he hopes to be able to identify why that is so that researchers can bottle some of the Silicon Valley science that goes into building a thriving startup community.

Published on: Feb 9, 2015