Part 1
Part 2
Part 3

Part 4

After college instead of working in politics as he had always planned, he had joined a few friends in a startup. One of them, Zach, a stats major, thought that he had cracked a hidden pattern of startup founders and aimed to make that his business. By running data analysis on as many successful founders as he could identify, and 1.5 million professionals in New York and Silicon Valley, Zach had pinpointed a few dozen data points he called predictors. These included university program, graduate program, childhood neighborhood, current neighborhood, age, race, gender, employment history, and attendance at certain events.

With funding and a handful of tech industry leaders signed on as mentors, an email went out to an algorithm-picked list of what Zach called high potential individuals, inviting them to a networking event with hints of possible future funding and mentorship. Most of them didn’t even have a startup idea when they received the email.

Seventeen of the 350 were selected after further screening as having the best chance for success, and they went on to start companies. After five years, one company got a lot of media attention and many users, giving Zach and Stephan a 20x return on investment. The experiment had been proved a success.

But something still bothered Stephan. He wasn’t convinced that the algorithm-generated list of people did any better than a random sample would. They had limited the pool to professionals in New York and Silicon Valley, and Stephan thought that a good idea could come from anywhere. He was convinced that the success they’d had was an artifact of the mentorship and support, nothing more.

After years of pestering Zach, they would test it with two new classes of founders. One would be filled with algorithm-selected individuals; the other, random people. Each would receive the same amount of attention, support, and funding, starting with identical emails inviting them to a networking event:

You’ve been chosen as one of the most likely people to create a successful startup. We realize this note comes out of the >blue. We lead The Future Founder’s collective, the venture fund backed by Bloomberg LP. We don’t like waiting for the >future to come to us, and we began to search for a way to use data to find the future ourselves. Over the past few months >we’ve been conducting an extensive study of everyone in the Bay Area and New York, using publicly-accessible data to >>analyze future potential.
Our model predicted that, right now, you are one of the top 350 people most likely to start a company and succeed.
We would be delighted if you joined us...

Twenty of the algorithm-selected individuals started companies that the collective funded, whereas only thirteen of the random population did. After five years, however, both groups produced exactly one successful company. In fact, the random sample’s company netted Zach and Stephan a 25x return on investment, and the algorithm-picked company netted only a 17x ROI.

Stephan wanted to go public with the results and change the company’s philosophy and operating procedure, to choose people at random to participate. Zach wanted to keep the results hidden and continue using the algorithm because, he said, “It’s a better news story if people know we’re using an algorithm and not picking names out of a hat. And it’d be an easier sell to keep the mentors if we’ve got a better method to it.”

Zach, being the founder, had naturally kept a majority stake, so his word was law. Stephan quit because he “couldn’t live in good conscience defrauding the investors.” But his years in tech had made him fantastically wealthy.

----------------

Want to keep reading? Here's Part 5!