Sign up FAST! Login

New York Times on Safe VC Betting using Big Data


Stashed in: Google Ventures!, Venture Capital!, Big Data!, Kleiner Perkins, Sequoia, Jerk Store, Twitter!, 106 Miles, Funding, #success, Big Data, Google Ventures, Startup, Awesome, Wafflebot!, Simpsons!, Big Data, Correlation is not causation.

To save this post, select a stash from drop-down menu or type in a new one:

Google Ventures and its take on investing represent a new formula for the venture capital business, and skeptics say it will never capture the chemistry — or, perhaps, the magic — of Silicon Valley. Would computer algorithms have bankrolled David Packard or Steve Jobs? Foreseen the folly of Pets.com?

The data provides one answer to those questions, at least for now: Since its founding in 2009, Google Ventures has stood out in an industry that, for all its star power, has been dealing its investors a bad hand. In recent years, an investor would have done better with a ho-hum mutual fund that tracks the stock market than with some splashy V.C. fund. Venture capital funds posted an annual average return of 6.9 percent from 2002 to 2012, trailing major stock indexes, according to Cambridge Associates.

Google Ventures, like all venture funds, does not publicly reveal returns. But its partners can count on one hand the number of its 170 investments that have failed, though it is too early to know how many will succeed, and it has missed investing in some superstar companies. Its successes include companies that have gone public, like HomeAway for vacation rentals and Silver Spring Networks for smart grid software, and start-ups sold to Google, Yahoo, Facebook and Twitter.

If fewer than 5 Google Ventures investments (out of 170) have gone out of business, then it is possible they are not taking enough risk.

Nor, with survivorship bias, does that failure-count alone indicate the return considering the sheer volume of bets GV is placing. I have every confidence they have results to be proud of, but then again, confidential returns are not a news hook solid enough to hang a hat on. To state the obvious, success for a VC is not as binary as it is for founders (sold/closed), the actual stake, price, and exit determine final cash-on-cash return.

It really feels like the writer went into this article with a thesis -- VCs are using data to decide! -- but then did not have the results to back up the thesis, and went with anecdotes instead.

Too bad.

Big Data tells Google Ventures which founders are worth investing in:

Here’s a riddle from Graham Spencer, a general partner at Google Ventures who oversees its data work: Is it better to invest in someone who started a company in a mediocre year for returns and did well, or started one in a good year with mediocre results?

Most people say the first case. But results from academic studies show it is the second, because that indicates the founders have a better sense of market timing, Mr. Spencer said.

Right. Wait, what? Really?

I need to find that study.

Yeah, I really wouldn't believe that study unless/until I knew more parameters. It's crazy to say that market timing is a determining factor for founders (much more so for VCs, but that's a different matter).

It's very unlike the New York Times to throw something like this out there without details to back it up.

Let me clarify my complaint: the data may say what the reporter summarized ('summer babies are richer') but that (alone) does not justify the claimed conclusion ('summer babies have the market timing genius to wait until summer to be born')

Ah, the old "correlation is not causation" logical fallacy that I've grown to know and love so well.

correlation causation lisa simpson rock tiger meme

Repeat entrepreneurial success is likely:

Some of the lessons seem obvious. An entrepreneur who has started a successful company is more likely to do it again, Google Ventures found, and start-ups based in tech hubs like the San Francisco Bay Area are more likely to succeed.

So they're looking for people in the Bay Area who previously have been successful?

Can they quantitatively measure how much of a jerk someone is?

“We would never make an investment in a founder we thought was a jerk, even if all the data said this is an investment you should make,” Mr. Maris said. “We would make an investment in a founder we really believed in, even if all the data said we’re making a mistake. But it would give us pause.”

Do they give every founder they meet a Jerk Score?

Bet on the jockey, not the horse.

In the world of Startups, what's the jockey? The founders?

Giddy-yep.

Correlation Ventures uses all the big data they can get their hands on and bets on the horse, the track, the sun in their eyes, and every other piece of information they can get their hands on, not the jockey.  

http://correlationvc.com/about

Joyce affectionately refers to Correlation Ventures as "Wafflebot":

http://pandawhale.com/post/3560/an-algorithm-to-pick-startup-winners-technology-review

Twitter mentions correlate with success:

Other venture firms are at further ends of the spectrum. For some, like Kleiner and Sequoia, data-driven venture capital means simply tracking how many times a start-up is mentioned on Twitter, or where it ranks in the App Store.

Twitter mentions?

Markets are more correlated with success than teams or products:

For others, it means relying almost solely on algorithms to choose where to invest.

That is the strategy of Ironstone, a venture firm started by William R. Hambrecht, one of the pioneers of tech investing, who started a technology-focused investment bank in 1968 and helped take many of the biggest tech companies public.

Ironstone’s algorithms have produced conclusions that are heretical to traditional venture capitalists.

They say, for instance, that a start-up’s founding team has only 12 percent predictive value, even though most investors rank that as one of the most important factors. And just 20 percent of Ironstone’s analysis focuses on the start-up itself, and the rest is on the market it is entering, because they say start-ups are likely to change course and the market has more predictive power.

Market matters more than product or team or anything else.

I wish I could find an academic study at Harvard or Wharton that backed that up.

Food for thought: http://www.booz.com/media/file/sb61_10408-R.pdf

More food for thought: Team chooses market, so team matters most because it can assess if it's in a bad market and choose a new, better market.

Soon after I got to Wall Street as a grunt sales assistant to a VP I began interviewing the top performing professional traders and investors across our industry.  The random walk theory and controversy about "alpha" (the financial return an active professional can earn that is in excess to the return generated by investing in a passive market index) and whether or not any professional could outperform year over year was being blown away and tested by some very consistent high performers, like Paul Tudor Jones.  The question to answer was, what is it that makes a professional investor into a year over year success as a high performer?

It wasn't data, big or small.  How do I know?

We created a categorical sort of the entire universe of professionals, filtered by their use of data inputs and decision making framework.  We first determined if their data use was fundamental or technical and then identified if they were making discretionary or mechanical decisions.  I had many, many interviews with these top traders and they all clearly put themselves into one of these four quadrants as regarded their investing approach:

Fundamental data and discretionary decisions

Fundamental data and mechanical decisions

Technical data and discretionary decisions

Technical data and mechanical decisions

For the layperson, fundamental data are qualitative inputs, like rumor, hearsay, opinion and analysis of any market information from another human and conveyed by narrative or visual media--no matter how received and consumed.  Technical data are quantitative constants (most usually price, volume, time. etc.) that are received and consumed devoid of any wordy opinion or qualitative narrative context.

In general, discretionary decisions are malleable, whether based on hard data, guts or tea leaves while

mechanical decisions are resolute and automatic, typically based on numerical algorithms. Some professionals I interviewed even had legally binding contracts in place to prevent team members from overriding their own buy, sell and hold signals.

In relevance to the NYT article, it's disingenuous for them to say that Big Data is a new formula.  Investment professionals have been crunching big data for centuries.  From the comments of the Google Ventures team in the article it's probably safe to say that even though they use Big Data, they still make their investment decisions on a discretionary basis.  I would go so far as to say they are probably not even technical data investors, but simply use that Big Data as a contextual way to bolster their fundamental data approach... e.g. "is this entrepreneur a jerk or not?"

What is predictive and highly correlative to any professional investor's ongoing alpha success?  Well, that's my hard earned financial alpha and it's worth more than a pretty penny to get it.

Half of Playboy Playmates were able to beat the S&P 500 while only a third of professional money managers were able to do so:

http://pandawhale.com/post/21692/what-playboy-playmates-can-teach-you-about-making-accurate-predictions

There's more luck involved in investing than anyone wants there to be.

You May Also Like: