The odds of a tech startup surviving aren’t just bad. They’re terrible.
About 75 per cent of all startups backed by venture capitalists fail toeven return investors’ capital, according to research out of Harvardlast year. Even the most addicted casino gambler would say those oddsstink.
Yet Thomas Thurston says he can predict a startup’s chances of failure or survival with amazing accuracy.
“We have algorithms to predict if they’ll survive or fail. To date,we’re over 85 per cent accurate about whether (a startup) will be aliveor dead in seven to 10 years,” says Thurston, CEO of Growth ScienceInternational in Portland, Oregon.
It’s all in the algorithms. Big data is now being used to forecast thefuture of the tech world’s smallest but highest risk players -startups. Basically, researchers collect facts and figures aboutthousands of startups, gathering information on their technology,products, customers, founders, staff, revenue, clients, marketing,pricing, sales channels and more.
Then they use analytics to crunch the numbers. Certain patterns ofvariables and behaviour emerge: those associated with the failures and those associated with the survivors. Thurston says he canpredict a newer startup’s odds of surviving or going bust by comparingits data with benchmarks from the master database.
Although big data algorithms have been used to try to pick winners andlosers in the stock market, they’re a newer phenomenon in the startupworld. If, as Thurston says, his method is accurate 85 per cent of thetime, it handily trumps what he calls the traditional “gut-basedintuition” method, which he says is incorrect 70 to 80 per cent of thetime.
“People are using computers to predict (successful) startups, not theirintuition, and that’s the first time in human history. And it’sactually working about four times better than intuition. We have about41 different combinations (of variables) and we’ve got five or sixmodels. But that’s why computers are great,” says Thurston.
– ‘Moneyball’ for startups offered by Cdling
– Canadian cities nab three spots on startup cluster list
– Tracking the Startup Canada movement
The approach holds obvious appeal for investors. U.S. author andventure capitalist Josh Linkner estimates that only one in every 300startup pitches gets funded. Some VCs are looking at big datatechnology as a way to speed up that process and increase the odds ofbetting on a winner. According to a recent VentureBeat story, U.S. VCfirms developing or using some form of algorithm to make investmentdecisions include Palo Alto Venture Science, Correlation Ventures andOpenView Venture Partners. Surprisingly, the bulk of Thurston’s clientsaren’t typical investor types but Fortune 50 firms with $100 million ormore in annual revenue.
“Those companies are trying to grow and they’re interested in starting their own new businesses,” Thurston explains.
Can predictive analytics be used to help the startups make decisionstoo? Bjorn Lasse Herrmann thinks so. Not yet out of his 20s, theGerman-born wunderkind is now CEO of Startup Genome Compass. TheCalifornia firm has gathered stats on 50,000 startup players around theworld using more than 25 key performance benchmarks. Otherentrepreneurs can input their own data (for free) to see how they stackup against startups from Herrmann’s massive database. Herrmann’s goalis to help fellow entrepreneurs improve their chances at success. Buthe says the process can also provide a brutal reality check forambitious founders, as it did for the team at one startup.
“They said ‘We never realized how bad it actually was until after welooked at your findings.’ Within a few weeks they closed a few officesand restructured,” Herrmann recalls. “They realized, based on thecurrent metrics, they’re never going to get to success based on themoney they have in the bank. Now they have a much deeper runway and avery, very small team working on a more useful product.”
Toronto-based Cdling (pronounced “seedling”) Capital Services Inc. tries to predict a startup’s chances of success by asking the wider startup community to gauge those odds itself.
Cdling asks members of itsonline community to predict when the startup will do the following:raise $10 million, raise at least $150,000 in angel financing, reportmore than $100,000 in revenue, and when it will go out of business.(For an example, the crowdsourced Cdling forecast page for Montrealstartup Bunch can be found here.)
“It’s a form of expert opinion sourcing around a startup,” says Cdling founder Michael Cayley. “We’re part of a much broader trend, in a way a sort of more sophisticated approach that leading investors are trying to take, using networks to support the decision they’re making and developing data solutions.”
Predictions made by the most influential Cdling members are weighted the most heavily, giving predictions from experienced entrepreneurs and investors more weight than those made by neophytes.
Cdling also rates the influence (or social capital) of startup players such as founders and investors based on extensive analysis of their connections to successful financing deals. Due to this approach, combined with the crowdsourced milestone forecasts for startups as described above, Cdling considers itself to be a ratings agency for the startup ecosystem, much like Standard & Poors rates the ability of borrowers to repay debt and U.S.-based FICO assigns personal credit rating scores. Investors can use the survival prediction part of Cdling’s platform when making funding decisions. Startups can use Cdling’s social capital ratings of investors to strategically target their pitches to the most influential VCs andangels within their sector.
In the final equation, however, can crunching numbers alone reliablypredict startup winners and losers? Not in Traviss Corry’sestimation.
“My feeling is no matter how refined the (predictive) technology is,it’s still going to need some human judgment in the equation,” saysCorry, co-founder of Toronto’s Incubes incubator.
Corry pores through hundreds of applications every few months fromstartups applying to join Incubes. He says predictive analytics can bevaluable in determining, for example, if a startup’s product is theright fit for a certain market. But he points out that since big datais based on benchmarks from startups and products that already exist,it probably won’t spot the next Steve Jobs because his (or her)technology and approach have never been seen before.
“It’s difficult because the real out of the box thinkers are the onesyou’re looking for. It’s the Steve Jobs and the Bill Gates and theexact entrepreneurs who are always rejected by every system in place.How do you identify this kind of innovative thinking when what they’recreating is probably too innovative for you to recognize until you’vesat down with them?”
Here are some resources if you want to get a forecast of your startup’ssurvival chances and see how it measures up against other emergingcompanies at a similar stage in the same sector.
Cdling: Free for startups to join. Based in Toronto, so its Canadiancontent and membership run deep. There’s a social forum where memberscan provide a crowdsourced prediction of your startup’s chances ofsurviving, failing and hitting key funding and performance milestones. You can also see ratingsof who are the most influential and connected members of the startupecosystem, handy when targeting pitches, networking or trawling for talent.
Growth Science International: This Oregon firm’s predictive technologyis too expensive for startups to use at the moment, says CEO ThomasThurston, but his firm is working on a more affordable version aimed athelping startups improve their strategy, structure and survival odds.For now GSI aims its service at a few VC clients and Fortune 50 clientswho use it to assess promising startups or try to launch their own.
Startup Genome Compass: Located in the San Francisco Bay area, thisfirm works directly with startups by running their data through itssystem for free and giving them the good and bad news on how they stackup against other startups so they can shape up or call it a day.
YouNoodle: RIP YouNoodle. This San Fran firm got tons of buzz in 2008 and 2009by promising an alogorithm program that could predict the survival oddsand future valuations of startups. After key members of the foundingteam flew the coop, the firm restructured and the startup predictortool that got so much attention got axed.