A Moneyball-style revolution is taking place in venture capital.
[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']Just as the renegade general manager of the Oakland A’s flouted assumptions about baseball and replaced gut feelings and outdated statistics with more effective quantitative analysis, a new breed of venture capital firms are throwing out their Magic 8-balls and are using computer-based models to make smarter investments.
Venture capitalists keep making mistakes, but the algorithm is getting smarter. We want to establish ourselves in Silicon Valley with a different business model. Who else does this?
Matt Oguz, founding partner,
Palo Alto Venture Science
“The bottom line is that the game has changed … and there is a lot of digital exhaust out there,” said Chris Farmer, a partner at General Catalyst, which is considered one of the more progressive of the older VC firms.
AI Weekly
The must-read newsletter for AI and Big Data industry written by Khari Johnson, Kyle Wiggers, and Seth Colaner.
Included with VentureBeat Insider and VentureBeat VIP memberships.
Before it pours thousands of dollars into researching a potential investment (work typically performed by a well-paid associate), a small cadre of venture firms are using analytics tools to pull in megabytes of relevant data, whether it’s a game’s performance in the various mobile app stores or conversations about a new e-commerce site on Facebook and Twitter.
Relying on instinct simply isn’t good enough anymore.
If algorithms can predict the results of elections, why not the success or failure of a tech startup? Washington D.C. woke up to the power of data when numbers-cruncher Nate Silver proved critics wrong and delivered a gut punch to traditional punditry by accurately predicting the outcome of the presidential election.
In Silicon Valley, new firms are going a step further by creating an entire investment thesis around data. “Algorithms will be the heart and soul of due diligence — it’s not just a sanity-check mechanism,” said Matt Oguz, the managing partner of new investment firm Palo Alto Venture Science. “It’s the only way to cut through human bias.”
Oguz is taking a cue from Wall Street, which has been using algorithms for years to track the rise and fall of stock, and the macro-shifts in the financial markets. According to him, it’s only a matter of time before this algorithmic approach seeps into the private investment market like osmosis.
He is one of a growing number of investors developing algorithms. Some are vocal about their research, while others are keeping it quiet to prevent competing firms from following suit.
[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']
This next generation of venture capitalists are fixated on a billion-dollar question: Can an algorithm predict whether a tech startup will succeed or fail?
Venture capital is a numbers game
I recently received an invitation to meet a partner of a leading venture capital firms at a San Francisco coffee shop. The investor popped open his laptop to reveal a snippet of a “stealth” project he’s been working on for years.
At face value, it did not appear to me much more than a series of nondescript charts and graphs. However, he explained that this is the nascent research behind an algorithm that can take much of the guesswork out of venture capital.
In response to my befuddled gaze, he traced his finger over a graph that charts the rapid ascent of Facebook. At a certain point (marked in red on the graph), the algorithm triggers an alert: Facebook has become its own market. At that point, it is wise to invest in a company that would make money by piggy-backing off the social network, like Buddy Media. This is the most basic approximation of how the data might work to his advantage — but it illustrates the point.
[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']
This investor requested to remain anonymous. He is in no rush to pass on his research to competing firms with budget to throw at the problem.
Clearly, something is slowly shifting in the venture capital landscape. In the old days, investors would rely on the home run hits (Facebook, Apple, LinkedIn, and so on) to offset the failures in any portfolio. It took a combination of luck and intuition to pick out the hippie, greasy-haired Steve Jobs or college dropout Zuckerberg. With each fund, investors hope that at least one of their entrepreneurs will have the talent, guts, and persistence to compete in a crowded market.
August Capital investor David Hornik sees some profound problems with the algorithmic approach. “There are too many variables and too few data points to derive any meaningful signal from the venture noise,” he told me. “If you are a good venture investor, you are better off applying your own judgment than you are collecting data on how everyone else invests and performs.”
But firms may have no choice but to investigate these new quantitative methods. Recent studies reveal that over 75 percent of venture-backed companies fail. This number seems high — especially given the sheer amount of research that precedes the typical investment. In Silicon Valley, a firm will track between 6,000 and 9,000 new companies that form each year. For perspective, Trinity Ventures‘ general partner Ajay Chopra told me that his team will meet with 2,500 of these startups. With each fund cycle, Trinity typically selects about 20 of them.
[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']
Firms are facing pressure from all sides, particularly from their limited partners (“LPs”), who expect to see evidence of innovation (see recent criticism from the Kauffman Foundation about the “broken” venture capital industry).
Making money on a base-hit
The latest crop of firms to bloom in Silicon Valley are staking their success on the data.
This year, the firm raised $165 million, the largest first fund since Andreessen Horowitz got started with $300 million in 2009.
[aditude-amp id="medium4" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']
The new model stands out in a number of ways: It’s heavy on diversification, modest successes rather than home-runs, and rapid decision-making — startups can expect a “yes” or “no” in a few weeks. The fund is also extremely flexible: Its partners can invest up to $4 million and have no lower limit (they can go as low as $50,000). In addition, while most investors are taking an active role in their companies, the partners don’t take board seats.
The idea is that not every tech startup needs to have a billion-dollar IPO — a well-negotiated acquisition can yield a decent return.
The only hard and fast rule at Correlation is that it will not lead investments, only participate. It seems to be working. The current portfolio includes hot startups like Getaround, a peer-to-peer ride-sharing service; digital media company Say Media; and productivity tool PowerInbox.
“With our five person analytics team, we are constantly updating our dataset both for new exits and financings,” Kienzle told me. He would not disclose the factors it uses to determine whether startups will succeed, but he said that there are marked differences to Wall Street’s trading algorithms for the public markets.
[aditude-amp id="medium5" targeting='{"env":"staging","page_type":"article","post_id":556711,"post_type":"story","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,enterprise,entrepreneur,","session":"C"}']
“You are trying to capture factors that are fleeting,” said Kienzle. “We have been careful to build a model that is robust over time to sector and stage [and] are going back through multiple cycles of venture.” Another challenge is that much of the data is not as publicly available — the legacy firms are notoriously bad at tracking their meetings and investments. A few firms have voluntarily handed over their data to further the research — but they will never sell it, he said.
Late-stage investment firms have been at it for years
For OpenView Venture Partners, a late-stage firm, algorithms are the future of venture capital.
“Where I think an algorithm will be most valuable is the timing on the market and the funding is the right amount,” said Adam Marcus, OpenView’s managing director.
In a phone interview with VentureBeat, Marcus said that his firm uses software to mine publicly available information like job postings and press releases and to capture social data around conversations.
Marcus said it’s ironic that venture capitalists laud the advent of “big data” analytics tools and technologies but are not applying them to their own corporate strategy. “When it comes to software as a service [Saas] technologies, we invest, drink the Kool-Aid, and use it to run our firm,” he said. “It’s a way for us to stay competitive.” However, he admits that it’s a tougher problem for early-stage firms, with fewer key performance indicators (KPIs) available for review.
Mapping the startup genome
Entrepreneurs have taken matters into their own hands. The launch of a research project known as “Startup Genome” stimulated debate in Silicon Valley. Still in beta, the project, has already received over 35,000 registrations.
Popular with angel investors and startups, the tool aggregates public information and data from the startups themselves (soon it will build APIs through Salesforce, Quickbooks, and other applications). It uses machine learning and natural language processing to benchmark a startups’ key performance indicators against similar companies. It also runs more than 50 red flag tests to calculate a company’s fundability and risk profile.
The 50-page report was authored by faculty-members at Stanford and Berkeley Universities. The key findings from an analysis of 16,000 startups include:
Factors that improve the odds
- “Pivots”: Tech entrepreneurs use this term to refer to the practice of trying out new ideas, shedding them quickly if they don’t catch on, and moving on to the next new thing. Founders have pivot at least once raise 2.5 times more money, have 3.6 times better user growth, and are 52 percent less likely to scale prematurely that did not pivoted at all or pivoted more than twice.
- Team size: Solo founders take 3.6 times longer to reach scale stage compared to a founding team of two people or more.
- Team dynamic: Balanced teams with one technical founder and one business founder raise 30 percent more money, have 2.9 times more user growth, and are 19 percent less likely to scale prematurely than technical or business-heavy founding teams.
Factors that decrease the odds or don’t make a difference
- Mentorship: Investors that provide hands-on help have little or no effect on the company’s operational performance
- Super fast growth: The most likely reason for startup failure is premature scaling. Don’t invest in entrepreneurs that get ahead of themselves!
- Experience of the founders: Successful founders are driven by impact, rather than experience or money.
Firms to watch
Two firms are strides ahead of the competition.
The statistics Ph.Ds focused on the question of early-stage investment. The partners wanted to know which factors made for a better chance of seeing a return on their investments. They found that the best way to make seed-stage investments was to make a larger number of deals — between 60 and 80 per year — that would generate a mean return of 25 percent.
“What really mattered was the diversity,” Google Ventures’s partner Joe Kraus told VentureBeat in a recent interview. “In the seed stage, you’re only gonna focus on the team; the product is too nascent.”
Related: Read our in-depth interview with Google Ventures partners
Likewise, General Catalyst has experimented with new approaches to investment. “Time will tell if it can work,” said Farmer, a General Catalyst general partner. Farmer is an expert on deal origination, arguing in a paper that an analysis of social media streams can be an effective way to evaluate and source deals.
Can algorithms replace venture capitalists?
These algorithms were not designed to do the work of venture capitalists — but they can supplement the traditional processes.
Let computers do the groundwork of monitoring everything through time. When humans dive into that information, they can pull out the pieces of the puzzle that look interesting
Sean Gourley, chief technology officer,
Quid
Sean Gourley, the cofounder and chief technical officer of “big data” startup Quid, said he was brought on as a consultant (this was how he met investor and adviser, PayPal cofounder Peter Thiel) to investigate this problem. “When a problem becomes more complex than a game of chess, a computer can’t solve it,” he explained. “Is monitoring global investment trends more complicated than chess?” Undeterred, Gourley hit on a similar finding. The computer was highly adept at picking out potential markets, and the right time to strike.
“Here’s how it would work: there are repetitive patterns that can be tells or signals that may show or warrant further interest,” Farmer said. “And then you do primary diligence … you do the traditional investment process.”
Farmer pointed to outspoken investors like Paul Graham, the founder of Y Combinator, who famously looks for entrepreneurial rebels that haven’t pursued a mainstream path. It’s an indication of the limitations of numbers-driven investing — an algorithm would favor highly experienced engineers or business leaders. Likewise, Ron Conway, one of the Valley’s most prolific angel investors, espouses investing in “people first.” According to Conway, markets are often in flux, but a great entrepreneur can change course to meet a pressing need.
Venture capital has exceptional investors and exceptional entrepreneurs. And yet, on the whole, the VC industry has not delivered returns that exceed the public market for more than a decade. With critics calling for more data-driven analysis to fix a “broken industry,” it’s about time VCs caught up.
At CloudBeat, VentureBeat’s upcoming conference, the Big Data theme will be explored by experts through real-world use cases.
