Sponsored by Oregon Business

Will this startup succeed?

| Print |  Email
Articles - May 2013
Monday, April 29, 2013


0513 FOB Profile Thurston
Thomas Thurston exploring laser mathematics at OMSI's Kendall Planetarium
// Photo by Adam Wickham

Thomas Thurston is in his office above Lela’s Bistro in Northwest Portland, describing an entrepreneur he talked to earlier that week. “It just bugged me. Something about this person — every time I engaged with him, it just bugged me.” The CEO of Growth Science, Thurston doesn’t usually traffic in gut feelings. On the contrary, the 35-year-old “data scientist” is strictly in the numbers game, creating computer algorithms to predict whether new companies and innovations will succeed or fail.

Until recently, Thurston worked primarily as a consultant to large Fortune 500 companies. But this past January, Growth Science shifted its focus to smaller companies and startups, deploying a model with a reported accuracy rate of 85%. “It’s not perfect, but we’re predicting way better than intuition,” says Thurston of his (top secret) formula, which combines elements of statistics, computer science and cutting-edge business strategy theory.

In February Thurston signed on as a partner at the San Francisco-based Ironstone Group venture fund, bringing his method for identifying winners and losers to an industry that often bases decisions on personal relationships and those ineffable gut feelings. “People who start companies are brave, smart, likable,” says Thurston. “But if we don’t get a green light with the model, we’re not going to invest.”

Mild-mannered and self-deprecating — “Gosh, I’m not an engineer, I’ve never taken calculus; my value is asking the stupid questions” — Thurston holds law and MBA degrees and worked as an attorney for several years before moving to Intel Capital, a venture capital group that invests in outside companies. While reviewing potential investment candidates, Thurston became interested in looking for patterns that might indicate which businesses would go boom or bust.

So he pored over Intel case studies, business school innovation theories and obscure German doctoral theses, eventually creating a model based on Harvard Business School guru Clayton Christensen’s theory of “disruptive innovation”: the process by which a product or service enters at the bottom of the market, then gathers momentum and eventually surpasses established competitors.

How does Thurston’s model work? It’s rooted in the mountains of data he has collected on market and corporate dynamics, including the anticipation of future changes in the marketplace. Patterns of success or failure then emerge depending on different industry and business behaviors. “The key is identifying variables that are predictive of success and failure,” says Thurston, who is very hush-hush about those variables. It’s a process that involves “lots of hard, hard work,” he says. “You go through a whole haystack to find one needle.”

Since founding Growth Science five years ago, Thurston, a University of Oregon graduate who lives in Lake Oswego, has worked on automating the model to make it affordable for entrepreneurs, as well as for the billion-dollar companies that have been using the company’s simulations to guide product launches and acquisitions. The son of U.S. government aid workers who grew up in Honduras, Bolivia and Nepal, Thurston describes his work as “a social mission.” About 80% of startups fail, he laments. “People are losing their jobs; it’s terrible.”

Aiming to make more businesses fail less, Growth Science not only runs the simulation, but if the model gives a thumbs-down, it also offers guidance as to how to yield a better success rate.

Extracting meaning and benefits from the reams of data unleashed by the information age has become all the rage in the past few years. Or, as Thurston puts it, data science is “taking off like a hockey puck.” Numerical models are being deployed to predict how the Supreme Court will rule on a given case, to find cures for disease, even to predict where crimes will happen. Data science entered the popular vernacular last year, when statistician Nate Silver plugged in information culled from polls, demographics, party registration and other sources — and correctly forecast the results of the 2012 presidential election in all 50 states.

Using computer simulations to predict the outcome of business decisions makes a lot of managers, executives and entrepreneurs uneasy, admits Thurston. “They’re afraid if the algorithm does the work, that brings their value and job into question.” Not to worry, he says, emphasizing that data science is simply about pattern recognition: “the tacit admission there are patterns in business — and that there are some patterns human brains are lousy at recognizing.”

But there will always be room for the human factor. An algorithm, observes Thurston, can’t tell if someone is annoying — like that hapless entrepreneur — or if a given product is interesting. “A computer,” he says, “can’t look at something and say it’s awesome.”



+2 #1 DavidGuest 2013-04-29 23:28:44
Very interesting work. And really good to see this coming from the Portland business scene. If this type of work can do 10% for business what Silver's models have done for election predictions, it would turn the current business investment/fund ing model on its head. Hopefully for the better.
Quote | Report to administrator
0 #2 Color Me SkepticalGuest 2013-05-01 19:14:10
I haven't met Mr. Thurston, but I'm skeptical of any "closed source" model when there are thousands of papers published on this very subject every year in B-schools all across the world. You can go all the way back to 'Altman's Z Score' http://en.wikipedia.org/wiki/Altman_Z-score in 1968 and build something that will get you started in a spreadsheet
Quote | Report to administrator
+1 #3 Valley GuyGuest 2013-05-02 04:00:57
The fact of the matter is, most business decisions have for hundreds (thousands?) of years been based mostly on intuition and gut instincts. If we can add to these human talents the power of data science to more accurately predict business success, it will be good for everyone-more successful businesses means more jobs and more income for families. I'm all for it.
Quote | Report to administrator
+1 #4 Proof in the Pudding - JBGuest 2013-05-03 16:36:03
I know Thomas and have worked with him closely. The "closed source" aspect simply means that his secret sauce is proprietary; hard to blame someone who has personally spent the time to compile the vast data sets he has. I doubt anyone else would freely disseminate the trade secrets that they make a living off of. You don't need to reverse engineer his models to determine if their accuracy.

And let's not confuse the latest b-school buzzwords of "innovation" (and their related-though- often-misused cousins "disruptive" and "agile") with real world practicum. While those professors and students spend their time debating the definition of "innovation" and what it smells like, Thomas can actually tell you. Contact him and ask him how his Rottura Capital portfolio (the other Portland area based public equities investment manager he is a partner of) has performed since its inception- the proof is in the pudding.
Quote | Report to administrator
+4 #5 2 kinds of skepticismGuest 2013-05-03 21:08:40
This is Thomas (from the article), great discussion. I tend to be a very literal person, to a fault. So if it's okay I'd love to share something it took me years to learn when most people would have seen it right away.

When this research had its first "eureka" moment there was a lot of peer review, additional vetting and sanity checking that followed. As in any science, this process consumed years. Daily life was all about things like sample size, robustness, replicability, parsimony, falsifiability, correlation coefficients, out-of-sample validation, etc. You get the picture. This turned out to be the first kind of skepticism. It asks "do the models really work?" This kind of skepticism is healthy and should be insisted on. It's good.

The second kind of skepticism looks the same on the surface, but is quite different underneath. This second skepticism isn't worried about any risk that the models don't work. Rather, it worries that the models actually do work. It's fear about the social and moral implications of a world where business can be accurately predicted using algorithms. For example, entrepreneurs with this fear may be worried their flashes of genius won't be grasped by "Hal." Managers and investors can have a different fear. They may feel threatened, much as the old-time baseball scouts hated algorithms in Moneyball. These fears are reasonable and we owe it to ourselves to not dismiss them. I sincerely hope and believe our algorithms are good for the world, but it's important to have a dialogue about the social and moral implications of any new technology that could change things in a big way. In the case of Moneyball, the old scouts didn't lose their jobs. In fact, today there are more scouts than ever and they've learned to combine algorithms with their own knowledge to become even better at their jobs.

It really helped me to realize there are two main kinds of skepticism when people first hear about our work. Both kinds are fair and bring their own value. Of course, there are some people who can't be convinced of anything, no matter how much proof you offer. There will always be unreasonable people. Yet most of us are reasonable, even if we're a bit skeptical, and that’s a great way for a healthy conversation to begin.
Quote | Report to administrator

More Articles

5 questions for Ruby Jewel creator Lisa Herlinger

The Latest
Saturday, October 24, 2015

What's it like working with your sister and how do you compete in Portland's crowded artisan ice cream space?


Back in Black

Guest Blog
Friday, November 20, 2015

It’s been a volatile year in equities and heading into the holiday season, it doesn’t look like these market extremes will dissipate.


The Shift to Community Health Care

November/December 2015
Wednesday, October 28, 2015

A conversation with Patrick Curran, CEO of CareOregon.


Not Your Father's Cafeteria

November/December 2015
Wednesday, October 28, 2015

Corporate food service reaches out to foodies.


The Harder They Fall

November/December 2015
Wednesday, October 28, 2015

Storyteller-in-Chief by the managing partner of Schwabe, Williamson & Wyatt.


100 Best Nonprofits: Working for equality inside and out

October 2015
Wednesday, September 30, 2015

Striving for social equity is the mission of many nonprofits, and this year’s 100 Best Nonprofits to Work For in Oregon survey shows employees are most satisfied with their organizations’ fair treatment of differing racial, gender, disability, age and economic groups. But as a national discourse about racial discrimination and equity for low-income groups takes center stage, data show Oregon’s 100 Best Nonprofits to Work For still need to make progress on addressing these issues within their own organizations.


Where Do We Go from Here?

Guest Blog
Wednesday, October 21, 2015
102115-thumbBY JASON NORRIS | CFA

Volatility reigned supreme over the summer. The old Wall Street adage of, “Sell in May and go away,” was prophetic in 2015.

Oregon Business magazinetitle-sponsored-links-02