AI Case Study
GV evaluates new and follow-on investments using machine learning
GV's staffers use "The Machine" to get approval to make an investment. By plugging a deal's details into the system, such as round size, syndicate partners, past investors, industry sector and the delta between prior valuation and current valuation, the algorithm ranks deals on a 10-point scale. Green, which represents 8 and above means go, red means no and yellow means proceed with caution.
Investment Banking And Investment Services
"Axios has learned that the firm, formerly known as Google Ventures, for years has used an algorithm that effectively permits or prohibits both new and follow-on investments.
Staffers plug in all sorts of deal details into "The Machine" — which is programmed with all sorts of market data, and returns traffic signal-like outputs. Green means go. Red means stop. Yellow means proceed with caution, but sources say it's usually the practical equivalent of red.
It was initially designed and used as a due diligence assistant that could be overruled but, according to three sources, it has evolved into a de facto investment committee.
Inputs into "The Machine" include round size, syndicate partners, past investors, industry sector and the delta between prior valuation and current valuation. The algorithm then ranks deals on a 10-point scale, with green said to represent 8 or above."
"GV was formed in 2009 as one of the first venture firms to employ engineers whose primary job was to work with portfolio companies on technical challenges. But, in the early days, there weren't too many portfolio companies yet, so the engineers were tasked first with building a dealflow management tool dubbed "Vortex," and then with what would become "The Machine."
Another impetus was that few of the early GV investors had much, if any, investing experience. So "The Machine" would leverage the firm's strengths (engineering) as a bulwark against its weakness (proven VC chops).
The engineers also were asked to have "The Machine" help source deal opportunities, but that wasn't viewed as a terribly successful effort."
Details unspecified but it appears to be predicting a numeric output based on several inputs
"round size, syndicate partners, past investors, industry sector and the delta between prior valuation and current valuation"