Follow Datanami:
March 13, 2014

Will Your Startup Succeed? Ask an Algorithm

Tiffany Trader

As futurist Arthur C. Clark once posited: Any sufficiently advanced technology is indistinguishable from magic. The predictive patterns have always been there, but it’s only with the evolution of technology and the advent of big data tools in particular that data scientists are able to predict future circumstances with what can seem like uncanny success.

A recent article on icrunchdata demonstrates the disconnect that can occur as digital technology advances faster than public perception. Seemingly overnight, “what if” and “if only” scenarios are coming to fruition. In this vein, someone recently commented “if only big data could be used to predict startup success.” Apparently, one company is doing exactly that: Growth Science, founded by Thomas Thurston in Portland, Oregon, in 2008.

Figures on the company’s website relate that 70-80 percent of new businesses fail, and approximately 90 percent of the Fortune 500 companies from 1950 have folded. “It’s not just a business problem, it’s a social problem,” maintains Growth Science, which formed because “business needs better science.”

Growth Science specializes in Business Model Simulations that use algorithms to predict whether a business will survive or fail. The model minimizes the guess work that VCs have to use when evaluating startups by automating 80 percent of the decision-making process. In 2013, Thurston become the chief investment officer of Ironstone Group, a venture capital firm that relies on data science to grow disruptive businesses.

“Simulation lets you analyze new business models, products, services or acquisitions. You can pinpoint risks, course-correct and deliver better results,” notes the company. “To date, these simulations have proven lower cost, faster and more accurate than the leading processes of the world’s top innovators.”

Thurston further explains how the model works in an interview with icrunchdata, noting: “We start with predicting survival or failure – if the business will be alive or dead within its first 10 years. We think of it as ‘firm mortality.’ We can also get into predicting the magnitude of success (ex. base hit or home run) but it all starts with mortality.”

While about 80 percent of Growth Science’s Business Model Simulation is automated, Thurston hopes that it will one day be 100 percent hands-free, but he admits that the model falls short when it comes to the most human factors, such as the personality of the leadership, user experience and customer service.

These things don’t automate easily, according to Thurston, but at the same time, he thinks that most people overestimate the predictive value of these elements. “It’s not to say they aren’t important (they definitely are),” he states, “but we’re much more concerned with their predictive value relative to other variables.”

Some other interesting findings from Thurston’s models are that the leadership team does not have that much accountability for survival or failure rates. About 80 percent of the predictive value of the simulation is tied to externalities, while 20 percent looks at the startup itself – with the team only accounting for 12 percent of the equation.

Although the algorithms themselves are proprietary, Thurston has previously mentioned several companies with positive reviews, including Dropbox, Tango, Indow Windows, Practice Fusion, and CloudFlare.

Related Items:

Big Data Market Growth Exceeds Expectations

Big Data Getting Big Bucks from VC Firms

The Big Data Market By the Numbers