Using The Computer to Predict Tech Startup Success

Image: Timothy Archibald for Forbes

David Coats and Trevor Kienzle have built a $166 million fund to ply their Moneyball approach to venture

Ask Trevor Kienzle about his favourite movies and in an instant he’s gushing about Moneyball. “It inspired me,” says Kienzle, a bald, skinny fellow who hardly looks like the movie’s real-life hero, sports executive Billy Beane. But Kienzle has a bigger point to make.

The movie lionised Beane for building a winning baseball team on a puny payroll. The key: Powerful data analytics that pinpointed valuable players who were underpriced. Take a close look at venture capital—where Kienzle and partner David Coats work—and the same number-crunching magic can be put to work. At least, that’s what they have been contending for nearly a decade.

By 2007 Coats and Kienzle had quit their jobs at mainstream venture capital firms to see if they could set up Silicon Valley’s first quant venture firm. Their mission: To stockpile 25 years’ worth of data on every venture deal ever done, to comb through this data with proprietary algorithms and to pick investments via pattern-matching software. Naively, they thought they could be up and running by mid-2008. Instead, it took an extra three years to raise all the cash they needed.

Today Coats and Kienzle are having the last laugh. Their firm, Correlation Ventures, runs a $166 million fund that has made about 85 algorithm-driven investments in startups. While it’s too early to know for sure whether Correlation has the winning touch, its early showings are consistent with funds that eventually produce double-digit percentage gains.

Correlation’s most obvious winner is a $1 million mid-2011 investment in Virsto, a storage software firm acquired by VMware in February 2013 for about $200 million. That translates into a 5-to-1 payoff for Correlation, or about 185 percent a year before fees. Extra profits take shape as companies raise new money at higher valuations than Correlation paid for its shares.

 A look at Correlation’s portfolio shows that likely winners-in-the-making include Good Eggs, Aldea Pharmaceuticals and AirPR.

Many of Correlation’s picks involve young Silicon Valley firms that also catch the eye of notable venture firms such as Sequoia Capital, Accel Partners, Battery Ventures, InterWest and Canaan Partners. Those overlaps are deliberate. Correlation’s secret algorithms weigh heavily on the track records of the entrepreneurs, investors and other advisors. Correlation believes that reputations aren’t random. Also highly prized is a company’s ability to spend money efficiently. A prospect rolling ahead on a tight budget looks appealing. A spendthrift rival, not so much.

Beyond that, valuation matters. Correlation’s algorithms often turn up companies in out-of-the-way locales such as Oneonta, New York, where startups are rarer and bargains may be easier to find. Correlation’s selections include a weight-loss company in Chicago (Retrofit), a Nashville payments processor (edo Interactive) and a biotech company in Malvern, Pennsylvania (Galera Therapeutics).

Correlation typically invests $100,000 to $3 million at a time, making it a small to midsize participant in financing rounds led by a fresh venture capitalist. That means Correlation needn’t repeat the extensive due diligence of other investors that have already confirmed the startup’s health and probity.

As a result Correlation acts fast, typically injecting funds into companies just two weeks after initial contact. Traditional venture firms can take as long as six months.

“Correlation is fast, reliable and easy to do business with,” says Mark Davis, who co-founded Virsto, the storage software company that VMware ultimately bought. Correlation didn’t want a board seat—and didn’t offer much business advice—but that hands-of approach suited Davis. He was already graced with all the venture capitalists’ advice he needed.

In ego-rich Silicon Valley being bashful can be a competitive advantage. Coats’ ‘aha moment’ came in 2005, when he was a partner in a traditional venture capital firm and arranged a San Francisco dinner with five of his portfolio-company CEOs. “They all needed more funding soon,” Coats says. “I asked them what they wanted from their next venture capitalists. I expected them to rattle of areas of expertise. Instead they told me: ‘If you really want to know the truth, we just want someone who will give us cash and then leave us alone. We’re getting plenty of advice already.’ ”

The lesson sank in. When Coats and Kienzle set out in 2007 to raise money for Correlation, they billed themselves as the financial equivalent of an ideal, last-minute dinner guest—filling out companies’ fund-raising lineups without annoying anyone. The firm spent much of its first two years weathering the financial crisis and building its databases. (It now has detailed information on 60,000 venture capital financings since 1987—about 98 percent of all deals.) Eventually institutional investors such as the University of Texas’ endowment provided the funding that Correlation needed to roll forward. Star academics such as Harvard Business School associate professor Matthew Rhodes-Kropf and the University of Chicago’s Steven Neil Kaplan signed on as advisors.

Correlation’s critics, and even some of its boosters, fret that the firm’s real-life results might not live up to the dream portfolios created by its algorithms, especially if Correlation fails to win access to Silicon Valley’s hottest deals. So far, Correlation’s founders say, such adverse selection hasn’t been a problem.

That’s partly because Correlation has worked hard to build friendly ties with leading venture firms—going so far as to line up 30 venture capitalists as small-scale investors in Correlation’s own fund.

Coats says there’s a separate, happy surprise hidden in Correlation’s database. Over the years, he has found, some top-performing deals turned out to be ones that took annoyingly long to close. So if Correlation can’t join hotly oversubscribed deals, he says, that may not be much of a loss.

In baseball Moneyball’s Billy Beane liked players who were uniquely good at getting on base, even if they did so by unglamorous methods like drawing a lot of walks. For bargain hunters like Coats and Kienzle it’s the same story: Harness the data in a way that reveals efficient—and under-appreciated—winners.