Failure Rates In Early Stage Venture Deals
My friend Jeff Jarvis invited me to talk to his class yesterday. Jeff is doing something really cool. He’s teaching graduate students in journalism school how to be entrepreneurs. So these students graduate with the understanding that they have another option, they don’t have to go work for a media company. If they are so inclined, they can try to start a media company instead.
We had a pretty far ranging conversation but the most interesting part of it was in response to the question “why do startups fail?”
I have seen my share of failure over the years, it’s part of the venture capital business. When we raised our first Union Square Ventures fund, I told prospective investors to expect 1/3 of our investments to fail. I always like the 1/3 rule, which is that 1/3 of the investments will fail, 1/3 will under-perform expectations, and 1/3 will meet expectations. Meeting expectations means 5x to 10x on our money. If you are into math, you can look at it this way:
1/3 average 7.5x – 2.5x
1/3 average 2x – 0.667x
1/3 average 0x – 0x
Total result – 3.2x
The important part of that equation is that you have to have high expectations going into an investment. If your expectations going into a deal are 3x on an average investment, you will fail in the venture business.
In reality, I’ve been able to do better than this over the years. But when talking to investors, it helps to set achievable expectations.
I went back over the past 17 years during which I have been doing deals by myself. During that time period, I have originated, led, and managed 32 investments, about 2 deals per year. I find that is pretty standard in the early stage venture capital business, 2 new deals per year per partner.
Here are the stats:
5x or greater – 11 deals – average 10.2x
1x to 5x – 7 deals – average 2.6x
Failures – 5 deals
Unrealized – 9 deals
Three of the unrealized deals are 1999 vintage Flatiron investments which will almost certainly end up in the 1x to 5x category so if I add them to that category and leave out the six unrealized Union Square investments that I manage, the distribution looks like this:
5x or greater – 11 deals
1x to 5x – 10 deals
Failures – 5 deals
That is decidedly not 1/3, 1/3, 1/3.
It is more like 40%, 40%, 20%.
I doubt that is sustainable going forward. If I take out the six investments I made in the “golden years” of 1996 to 1998, then the numbers look like this.
5x or better – 7 deals – average 7.1x
1x to 5x – 9 deals – average 2.5x (on the 6 realized investments)
Failures – 4 deals
That is 35%, 45%, 20%, which is more like what I’d expect to see from a good early stage investor’s track record.
In my next post, I am going to talk about why startups fail and how to reduce the failure rates.
November 29, 2007 10:09 AM, By Fred Wilson
Tags: failure performance returns success venture
Comments (19)
Impressive that you'd share this info. But if ya got it, flaunt it.
Posted by rick , November 29, 2007 12:57 PM
Hi fred, I'm the founder of a startup seeking 400k in seed funding I look forward to tomorow's post. I know having the big idea, proprietary software, solid contracts and just the beginning and the team is critical. Take a peek at www.personalpediatrics.com to learn about concierge pediatrics and Personal Pediatrics. Thanks, Dr. Hodge
Posted by Dr. Hodge , November 29, 2007 02:38 PM
Hey Fred. Bless your soul (as usual).
Curious: are there any common characteristics in the various deals/buckets?
For example... deals with >$10MM total invested (by all VCs) mostly had similar outcomes (and deals with <$10MM total invested mostly had similar outcomes)?
Or, deals that had strategic investors had similar outcomes?
Or, deals that had an advertising business model?
Or...?
Posted by Steve Kane , November 29, 2007 04:01 PM
Glad to see Jeff is teaching that class. I recently spoke at a b-school class going the other way: Teaching MBAs how to best deal with (albeit not become) the media. Something too few entrepreneurs, CEOs, VCs, etc. really understand...
Posted by Dan Primack , November 29, 2007 04:58 PM
Fred: great info, thanks! Hard data like this is very helpful for entrepreneurs.
Dan: sounds like a good subject for a post. (Or, include a link if you've covered it in the past.)
Posted by Scott Lawton (Blogcosm) , November 29, 2007 06:26 PM
Do VCs talk down their failure rates in order to justify higher equity positions? If the math really is 40% 40% 20%, then that might have implications for floor valuations VCs can accept...
Posted by Thomas , November 30, 2007 12:06 AM
i think the average for early stage startups is 1/3, 1/3, 1/3. You have to be able to get 5-10x on your money on the deals that meet expectations for the numbers to work.
That's what drives valuations.
Posted by fred wilson , November 30, 2007 10:03 AM
I just explained on Tech Crunch about the business plan situation, I mean you got to ask your self every single question before you even start business, people think that if somebody else did it why can't I ? Well that somebody
A. Got lucky ( cool domain name, had supporters etc.)
B. Wrote good business plan
C. Had backup plans
D. Layout the operational management structure.
~ Live Crunch
Posted by Live Crunch , November 30, 2007 12:00 PM
jeez Fred - you only made 6 investments from 96-98? You blew the chance of a lifetime! ;-)
Posted by mike , November 30, 2007 12:59 PM
This kind of transarency is so welcome in the VC world. I assume investors are comfortable with it because the numbers are good. For the entrepreneur it makes a huge difference to building trust and a productive relationship. One point of friction between entrepreneur and VC that I have seen is based on the fact that a VC is hedged by having a portfolio whereas as the entrepreneur's financial health is totally tied up in that one venture. So when a VC wants to "shoot for the moon" to get that 10x return, the risk to the entrepreneur needs to be considered. These issues are all toally resolvable within a relationship based on transparency and trust. Kudos to you for showing real leadership in making that happen.
Posted by bernard lunn , November 30, 2007 01:01 PM
Interesting read.. /ac.
Posted by aaron , November 30, 2007 01:26 PM
I think in a world of increasingly smaller investments to get a net company started and increasing number of net startups, the measure of a vc's career success should be net dollars returned overall vs failure od success in terms of number of deals.
For example, you could put $200k in 5 deals and make 5x on each and still lose $10m on one deal and have a negative return on an overall dollar basis but a great record on a deal times return basis. Similarly, you could make 5 small bets that fail but one that performs well to get an overall 3-5x return. At the end of the day lp's want to know if you made more money overall than their investment and if it beat their alternatives to deploy capital.
Posted by rk , November 30, 2007 03:29 PM
Mind posting this on Only Human? Its a site that allows people to share their mistakes in hopes of helping others. You don't even have to sign up to post your story.
I think our users would love to hear more about your experiences.
http://www.weareonlyhuman.com
Posted by Dennis Eusebio , November 30, 2007 04:34 PM
Hi Fred
I'll keep this brief, to let someone with better knowledge of stats improve it, and provide less opportunity for them to jump all over me...
One question is to ask whether you're really doing well or if your performance could be due to chance
For example, and this is just an example analysis, suppose we model the return of an early stage investment with an exponential distribution. Your initial assumption is that the average return will be around 3.2x, so we can pretend that each investment is an exponential random variable with mean 3.2 (and hence lambda parameter 1/3.2 = 0.3125).
You're counting 26 deals. If we take the sum of exponentially distributed random variables, the result follows the Erlang distribution. The Erlang dist has a mean of k/lambda (where k is the number of exponential random variables being summed). For k=26 lambda=0.3125, that's a mean of 83.2. Plugging in approximations for your returns, you have 11x7.5 + 10x2 = 102.5 in total return. So you're well above average, given all assumptions. But how far above average is that really?
The variance of the Erlang dist in your case is k/lambda^2 or 26/0.3125^2 = about 266. So the standard deviation is sqrt(266) or about 16.3
Without going into p-values and Null Hypothesis (I'm about to get on a 6-hour late plane, I hope), that places you just over one standard deviation away from the mean of that distribution. You can use the probability distribution function of the Erlang dist to see just what proportion of samples would be further from the mean, etc. But given the number of assumptions already, it may not be worth pressing on. One standard deviation away from the mean is nice, but it's also fairly inconclusive. We'll just have to give you the benefit of the doubt, I guess :-)
Did I say already that there are a lot of assumptions in the above?
BTW, I've been planning a blog entry on figuring out how good you are at something, and comparing that to efforts in totally different domains (was Babe Ruth a better hitter than Tiger Woods is a golfer?) and it's based on number of standard deviations from the mean.
Terry
Posted by Terry Jones , December 1, 2007 02:43 AM
Fred
Thanks for opening up an interesting discussion, which recalled a study I read earlier this year by Juan Villalobos of Angel Venture Partners of success rates in early-stage investments: http://tinyurl.com/yunza3.
Michael
Posted by Michael Rivers , December 3, 2007 01:52 PM
Hi Fred,
Not sure if I understand this clearly, but when you say 1x or 5x times return on capital invested, is that within a certain time period associated for the return to get counted? For example when you say "Three of the unrealized deals are 1999 vintage Flatiron investments which will almost certainly end up in the 1x to 5x category" what timeframe are you looking at?
Thanks
HimS
Posted by Himanshu Sahani , February 25, 2008 11:16 AM
Hi there,
My name is Ethan Suplee , I just signed up. I hope, I am posting in the right section, and I am sorry, if I posted in the wrong place,
moderators, please move this post accordingly.
Early last year, I stumbled upon information about
investment private clubs, that are not readily available to the public. The yields, which people could
receive by participating in such funds, appeared
surprisingly high at
first, but as I studied more about them, I began to make a lot of money. These days, I consider
myself being the successful
investor. My next step, is marketing field, so, I
made available the blog, which
shows detailed information
about real investment
funds. If you would like to see my
blog, [url=HTTP://WWW.FINANCESECURE.NET]click here[/url].
I would appreciate
user's feedbacks.
Posted by Investblogger , February 29, 2008 02:40 PM
Your initial assumption is that the average return will be around 3.2x, so we can pretend that each investment is an exponential random variable with mean 3.2 (and hence lambda parameter 1/3.2 = 0.3125). Thanks for opening up an interesting discussion, which recalled a study I read earlier this year by Juan Villalobos of Angel Venture Partners of success rates in early-stage investments
Posted by ersin , May 7, 2008 04:00 PM
Fred, these are impressive results and to my understanding much better than average VC returns, which are negative, right? Don Dodge posted min-analysis some time ago where he wound up concluding there was a lot more VC failure than is normally thought.
There are elite guys like you and Jeff Clavier who "beat the averages", but isn't "making money" with startups an unrealistic expectation, since those VCs and companies that succeed are still around to talk, but those who fail are not blogging about the burgers they now flip to pay the bills?
I'm also noting that without "time" as a factor the return is not meaningful. 3x is easy....if you use a 15 year horizon!
Posted by Joe Hunkins , November 29, 2007 11:15 AM