Wednesday, July 20, 2011
Google hopes to apply algorithms to the problem of identifying start-ups worthy of venture capital investment.
To some, it is a telltale sign of an overheated industry, symptomatic of a late and ill-advised rush to invest during good times. But Google says it has a weapon to guide it in picking investments -- a Google-y secret sauce, which means using data-driven algorithms to analyze the would-be next big thing.The point about investment being an art is well-taken. First, the fact that human beings have free will would, alone, make it impossible to reduce the whole process to an algorithm. Second, and setting that aside, any attempt to model the decision process would require the accounting-for of a mind-boggling number of variables. But that doesn't mean there is no way to take advantage of the brute strength of computation. On that score, it is interesting to see how Google has separated the wheat from the chaff by evaluating what their algorithms have unearthed.
Never mind that there often is very little data because the companies are so young, and that most venture capitalists say investing is more of an art than a science. At Google, even art is quantifiable.
"Investing is being in a dark room and trying to find the way out," said Bill Maris, the managing partner of Google Ventures, the corporate investment arm. "If you have a match, you should light it."
Google says the algorithms have taught it valuable lessons, from obvious ones (entrepreneurs who have started successful companies are more likely to do it again) to less obvious ones (start-ups located far from the venture capitalist's office are more likely to be successful, probably because the firm has to go out of its way to finance the start-up.)The second finding exemplifies the saying, "correlation isn't causation;" and if anyone thinks Google is going to start investing on the basis of "distance from Mountain View," I have bridge to sell. Distance has actually been eliminated as a relevant variable (if it was really being considered as one), but the finding with respect to distance may lead to other relevant data: Perhaps the "longer-distance ideas" were better (in and of themselves, or better-promoted). If there were some way to compare the overall merit of the successful long-distance proposals to successful ones from nearby, perhaps Google may be able to estimate how many successful ideas have been missed nationwide by looking at a ratio of successes to failures in its own back yard. Perhaps other findings will tell them how to hunt for talent nationwide.
Is investing an art, in the sense of requiring good intuition about the character of the innovators and market conditions? Or is investing a science that can benefit from algorithms and computational techniques?