New Partner at Quiet Hambrecht Fund Takes ‘Moneyball’ Approach to VC

Thomas Thurston says we’ve entered “the era of the Moneyball VC,” and no less a name than investment banking pioneer Bill Hambrecht is placing a bet on this Northwest-based data scientist’s formula for picking winners.

Over the last seven years, Thurston (pictured) has been developing algorithms—first at Intel Capital, and later in collaboration with Harvard professor Clayton Christensen, and under his own shingle at Growth Science in Portland, OR—to model markets and business behavior, and thereby predict the success or failure of a given innovation.

Now he is joining Hambrecht-led venture fund Ironstone Group as a partner to try this method in a venture capital industry increasingly open to data-driven approaches.

“It’s sort of this data scientist’s dream to find some amazing patterns and then get to put them out into the real world,” says Thurston, 35.

Since launching Growth Science five years ago, Thurston has used algorithms as the basis of a corporate consulting practice, helping large companies evaluate acquisitions, investments, and strategy shifts. He has also collaborated with Hambrecht—founder of investment bank WR Hambrecht + Co, and leading proponent of the OpenIPO method used most prominently by Google in 2004—who, Thurston says, has used the algorithms to inform his investment decisions. One of these includes a 2010 investment in Palo Alto, CA-based mobile video communications company Tango.

Hambrecht

A Hambrecht representative confirmed that Thurston has become a partner in Ironstone, but offered no additional comments. “In a couple of weeks, we’ll have more information about the fund,” the representative tells Xconomy.

Ironstone Group, currently structured as a public Hambrecht holding company, will be focused on seed and early-stage investments up to a “few million dollars,” Thurston says. He would not disclose the amount of capital at the San Francisco-based fund’s disposal.

“We’re not raising $100 million from a pension fund that we have to deploy in three years,” Thurston says. “There’s no unhealthy pressure to do deals. Instead, if we see awesome, disruptive companies, we’ll do a deal.”

The fund has no specific geographic focus—though Thurston says he would love to do more deals in the Northwest, having lived in Oregon on and off since 1994—and will invest in sectors including technology and healthcare.

Thurston joins the chorus of those finding fault with what he describes as the traditional intuition- and personal-relationship-driven approach to VC investing. “That’s wrong 70 to 80 percent of the time,” he says.

And he is not the first to apply the Moneyball moniker in venture capital. In fact, the term is coming up more frequently. (Michael Lewis’ 2003 book Moneyball, and the 2011 Brad Pitt movie by the same name, chronicles Oakland Athletics’ GM Billy Beane’s use of obscure statistics and data analysis to upend the old way of scouting baseball talent.)

In June, Xconomy profiled Correlation Ventures, a $165 million fund based in San Diego using analytics software to decide more quickly whether to co-invest in potential deals. Others pursuing a data-centric approach to investing include Matt Oguz, who last summer announced Palo Alto Venture Science, with a mission to chase the “cognitive biases and hype investing” out of VC with “data-driven analytical models.”

Since Moneyball, data-driven decision making has gained enormous currency, helped by the “big data” transformation of nearly every field of technological and scientific endeavor, and high-profile data-driven successes in other fields, most prominently Nate Silver’s nearly perfect prediction of the outcome of the 2012 national election.

It should come as no surprise to see these experiments in the venture industry, though many investors are skeptical, particularly around whether the data available to grind through the algorithms is meaningful.

“I’m a huge fan of data-gathering efforts on patterns of startup success and failure over meaningful time series and sample sizes… but since most seed-stage investing is done before you have much data to work with, I continue to believe that it’s a high-touch process that lends itself to human intelligence (especially emotional intelligence/’people sense’) more than algorithms,” says Chris DeVore, general partner at Founders Co-op in Seattle, in an email. “As the age-old analytics adage goes, ‘garbage in, garbage out’ and in a data-poor environment like seed/early-stage investing it’s easy to convince yourself you see ‘patterns’ that don’t really exist.”

On a recent visit to Canada, Tim Porter, managing director at Madrona Venture Group, heard another pitch about the Moneyball approach to VC. Madrona tries to be “as data-driven as possible in our investment evaluation,” he says in an email. “I think the main difference here is what happens after the investment—at Madrona we feel it’s important to invest more than money, and really help to be company builders.”

Porter also echoes DeVore’s concern about data availability particularly for early-stage companies. “This approach to investing likely works better for (a) later stage investing or (b) follow-on investing in an existing portfolio where you do have more performance data,” he says.

Efforts such as the Startup Genome project are helping make more data available. Thurston has his own approach to solving the problem, detailed below.

He does not pitch his data-driven approach as a wholesale replacement for the VC investment process, but rather as a tool to be used in concert with human skills.

“Sometimes computers can be really good at finding precisely the types of patterns our brains are ill-suited for. Therefore the right pairing between our brains (for some patterns) and advanced technologies (for other patterns) holds tremendous potential for the future of business, strategy and innovation,” Thurston wrote recently on his Growth Science blog.

“Technology doesn’t go to the birthday party of your top customer’s daughter or share ideas over a beer with your team. By the same token, your brain doesn’t accurately detect subtle probabilistic interdependencies while crunching a terabyte of data through complex algorithms with perfect fidelity.”

Thurston has encountered fear from entrepreneurs who don’t believe “any algorithm could catch that spark of genius that makes their business special,” he says, and from corporate M&A types who worry, “Do they have their place at the table, or does it undermine their value-add?”

He acknowledges that “good old VC” work is still required to identify promising companies on which to run the algorithms. After hearing a company’s pitch, Thurston would plug scores of data points into a few different algorithms “that triangulate onto the answer,” a process that takes a couple of days. He compares it to “how engineers would use a computer to simulate an airplane under different conditions before building a real one.”

Thurston guards the precise details of the algorithms as trade secrets, so much so that he has not pursued patents. “We expect the algorithms to be an advantage a lot longer than a patent lifetime so it didn’t make sense to disclose, even for a filing,” he says.

He shares these general details with Xconomy:

“We feed in a lot of data about a market, competitors, larger social-behavioral-economic trends, the relative advantages and disadvantages of the startup relative to these factors, which value drivers a startup has chosen to focus on. While this is a rough estimation, around 30 percent of the analysis is on the company itself, while around 70 percent is focused on modeling the market(s) the startup is going into.”

The algorithms—which were honed on a hand-built database of information on thousands of businesses, mostly at the time they were pitching investors for funding—tend to smile on business models that are “cheaper and worse” than market leaders, but still good enough. These are more likely to be disruptive—and also more likely to fly under competitors’ radars—than the “better mousetrap” plays, which, according to his statistics, fail more often, Thurston says.

“All we can do is hope to manage probabilities much better than other people have before,” he says.

One of the biggest differences between Thurston’s approach and the established practice of many VCs is emphasis on betting on the right people and expecting them to adapt to fast-changing markets.

Porter argues that the team is “the single most important factor in a startup’s ultimate success.”

Thurston says his algorithms value the team “as much as the empirical research says we should… but we don’t care nearly as much as most VCs do.”

In the end, the algorithm will be the final arbiter. “If the algorithm doesn’t like a business,” Thurston says, “we’re not going to do a deal no matter what.”

Trending on Xconomy