The Venture Capital Math Problem

Yesterday Albert and I visited one of the investors in our fund. The good news is they are happy with the job we are doing. The bad news is they are frustrated with the venture capital asset class.

We got to talking about the venture capital asset class and it wasn't long before we got to the "math problem". The venture capital math problem is pretty basic, maybe something you'd do in high school calculus or even pre-calculus. Here's how it works.

The venture industry has been raising between $20bn and $30bn per year for the past few years. Here's recent data from the NVCA's web site.

Vc fundraising

Let's be generous and say that the average is $25bn per year (it's actually more). The math problem is to figure out how much in proceeds every year need to be generated to deliver a reasonable return to the investors.

Here's how I go about solving it. My math is not perfect and I'd like to hear from all of you how you'd solve it in the comments.

First, the money needs to generate 2.5x net of fees and carry to the investors to deliver a decent return. Fees and carry bump that number to 3x gross returns. So $25bn needs to turn into $75bn per year in proceeds to the venture funds.

Then you need to figure out how much of the companies the VCs normally own. The number bandied about by most VCs is 20%. That means that each VC investor owns, on average 20% of each portfolio company. We'll use that number but to be honest I think it's lower, like 15% which makes the math even tougher.

Using the 20% number, that $75bn per year must come from exits producing $375bn in total value.

But it is also true that many of these exits have multiple VC investors in them, sometimes three or four. So you really need to look at the percent ownership by VC funds in the average deal at the time of exit. That number is likely to be over 50% and maybe as high as 60%. If we use 50%, then to get $75bn per year in distributions, we need to get $150bn per year in exits.

Here's where my math starts to get a little fuzzy and where I'd love some other approaches. I assume that the distributions of exits each year is distributed on a power law curve like this one:
Power law curve

I've heard from investors in venture funds, including the one we visited with yesterday, that about 200 exits per year produce all the returns in the business. I think that number is too low because that is about the number of venture funds raised each year. That would suggest each fund has only one meaningful exit and that's just not right. I think each fund has at least five or six meaningful exits. So I would use a number like 1000 exits per year.

And I assume that the biggest exit each year is $5bn. Yes, it is true that some venture investments turn into businesses like Apple, Google, Microsoft that are worth $100bn and more. But it is also true that most VCs are long gone from those deals before the valuations get to that level. So for the sake of solving this problem, I'd assume the largest exit each year is $5bn and then you have a power law distribution of another 999 deals.

These assumptions are important to the results you get from this math problem and I'd love to hear everyone's thoughts on them in the comments.

I tweeted this power law math problem yesterday hoping to get some answers. Most of the replies said that I need a "scaling exponent" or a "power number" to get the answer I wanted. So my attempt to use twitter as a math crutch didn't work out very well.

So in order to finish this post and get to a discussion, I'll assume that the biggest deal, $5bn, represents 5% of the total value of all 1000 exits and that the total value of all exits is $100bn per year. If others come up with different numbers I will update this post with them here.

So here's the venture capital math problem. We need $150bn per year in exits and we are getting about $100bn. That $100bn produces roughly $50bn in proceeds for venture firms per year. After fees and carry, that $50bn is around $40bn. Which is only 1.6x on the investor's capital if $25bn per year is going into venture funds. If you assume the investors capital is tied up for an average of 5 years (venture funds call capital over a five year period and distribute it back over a five year period, on average), then the annual return is around 10%.

Here are the most recent NVCA numbers which I got from Anthony Ha's Venturebeat piece a few days ago:


I tend to look at 5 year and 10 year numbers since 20 year numbers include a period when there was a lot less money in the venture business. Those numbers suggest that the 10% per year returns are about right, or are at least in the ballpark.

So here's my conclusions from all of this math (some good, some not so good). The venture capital asset class does not scale. You cannot invest $25bn per year and generate the kinds of returns investors seek from the asset class. If $100bn per year in exits is a steady state number, then we need to work back from that and determine how much the asset class can manage.

If you use my 3x gross and on average 50% ownership by VCs, then the number that the asset class can take on each year is around $15bn to $17bn. It's interesting to note that the industry raised $4.3bn in the first quarter of 2009. That's a good thing. If we can keep it to that level, or less for a while, then we may be able to downsize and get returns back on track.

I'm optimistic that it will happen. In an open and free market, capital will flow to the places where it can earn an appropriate return. I suspect we'll see some of the large public pension funds who have been drawn to venture capital over the past decade decide to leave the asset class because it does not scale to the levels they need to efficiently invest capital. That will leave the asset class to family offices, endowments, and other smaller institutions who made up the largest part of the asset class in the 1980s and early 1990s.

I think "back to the future" is the answer to most of the venture capital asset class problems. Less capital in the asset class, smaller fund sizes, smaller partnerships, smaller deals, and smaller exits. The math works as long as you don't put too many zeros on the end of the numbers you are working with.

Reblog this post [with Zemanta]
#VC & Technology

Comments (Archived):

  1. Pascal-Emmanuel Gobry

    I think that’s roughly right (he said, with zero VC experience), but if you want to look at the future I think you also need to take into account the compounding effects of economic growth and technological innovation. Probably not this year but soon the world economy, and especially emerging economies, will keep growing. And not only will the overall pie get bigger, but I get the feeling that the market for technological innovation, i.e. venture-backed companies, will grow bigger faster than the overall economy. And so that means that in aggregate, the exits from VC-backed companies will, over the long run, keep growing at a fast clip. Which in turn means that if the money poured into VC only grows at a slightly slower rate, then the returns of the overall asset class will mathematically increase.Does that reasoning make sense?

    1. fredwilson

      i’m not sure about that. technology has a way of commoditizing things. it may be that technology enabled businesses make up a bigger part of the total economic pie, but i would not assume that in the aggregate they will be worth more.

      1. Pascal-Emmanuel Gobry

        I wasn’t making a point about technology in general as much as disruptive technological innovation, which mostly happens through venture-backed startups. I do think that the share of economic growth coming from these will keep growing, and at a faster clip than the broader economy. I’m not sure how that holds up under scrutiny but it feels roughly right.

  2. Deepak Das

    Clarification Q:# Where do the investors come up with the number that only 200 exits produce all the returns. You are right that an average of 200 funds cannot produce just 200 exits that hold up the industry % number. That is a key number here. You assume 1000 exits based on an average of 5 exits per fund. Now where do you get the approximation that 5 is a good number to assume for exits/fund? I think this # holds the key to achieving some accuracy in the effort to produce the return % for the asset class. The two are off by a lot to be accurate.

    1. fredwilson

      there may be real data out there on that number. i suspect that the services that track the venture business, like thomson ventureXpert and others would have it

      1. Deepak Das

        Found some data for liquidity events from 02-09. Wonder if this number represents all possible exits!?

        1. fredwilson

          This is great data. Thanks for sending me that link. I can’t make sense out of the table given that I’m using a blackberry browserBut I’ll update the post with that data later

        2. apmaran

          Looked for a couple of seconds at it and what strikes me is that your numbers and these numbers together tells me that I will expect IPO to come back strong very soon after a short period of focus on fast M&As to build larger new companies much faster, retain-regain. So, expectations on return will be low in 1-2 years? but will rise to higher levels than before after that which will put much higher pressure on finding IPO material now. The numbers curve that way.

          1. Deepak Das

            Wow. How did you extrapolate that by glancing at the numbers? I mean given the macro economic conditions that we are in right now, how do you predict that shorter term and longer term scenario?

          2. apmaran

            I think its because I like to bake bread…The concept of a good bread is to let the dough rise long enough but too long doesn’t bring anything good, so my totally wacky math analysis is that the total amount of M&A deals in 04-08 is level and deal size is level except -07. Ergo, not focus for VC’s, IPO is jumpy numbers up and down which indicates happy hunters, sometime you make it good sometime less, ergo VC is targeting.But in the year of the crisis -08, the IPO drops dead as a dodo. Still dead right now. And everyone see that their dough, sorry; portfolios; are not rising and costs and expectations are building up. So after a while, we must go somewhere, ergo M&A (alternative would be to do nothing). So to get any kind of raise in value, look for the best bride to marry.Now the bread is baked and of good value, but we need to get some real payback so we need to sell the bread as expensive as possible and then I think IPO will be back in shape to attract new buyers. As in game theory we move in flocks when someone spots something good and in 1-2 years everyone will be looking for some new good stuff to buy and I firmly believe that traditional industry, banks, oil, cars has not recovered in peoples mind then, maybe in 5 but not in 1-2.So I found it interesting to see the numbers in context to the discussion, the discussion in it self is a proof of real energetic thinking. What would have been problematic is if the numbers had been fast growing and average M&A deal in -08 would be 250 and average IPO 140 then that would mean that the VC market was not connected to the real world and wouldn’t be able to cope now at all. But all seems fine:- to me…

          3. fredwilson

            I agree about the IPO market coming back

  3. stone

    I think this means that the reward for entrepreneurship will be diminished because returns will be lower. I am planning for this, Fred. Sadly, I believe this is true. I hope this leads to fewer, more ethical VC’s. I also hope this leads to fewer stupid ideas getting funded.

  4. ceonyc

    Couple of stats to keep in mind:1) Most funds don’t get fully invested. 85% is probably a good number… so chop 15% off that raised number, because that’s not invested.2) The bottom quartile of funds is fundamentally a wipeout. If you look at distributions to paid in ratios via VentureXpert, the numbers are so horrid you’d think they were wrong. You have to go back to 1997 to find a vintage year where the lowest quartile mark even distributed half its capital back to investors. Since then, we’re basically talking single digit % distributions or nothing at all.3) I actually don’t think you’re that far off with the number of companies that produce most of the returns. That’s why most VC funds are one and done and two and done. They’re funds that never should have been raised in the first place.

    1. fredwilson

      do you still have a login to VentureXpert? we could use the average number of VC exits per year.

    2. ceonyc

      Check that… the 85% number is high… it’s probably closer to 75%

  5. drstarcat

    This definitely corroborates my experience at that the real cap on the size of the early stage market isn’t cash, it’s quality companies.

  6. jeremystein

    if the best deals chase the best funds then i would imagine that the lower 30-40% of the funds are dragging down the asset class as a whole. imho, past performance is most relevant in PE.

    1. fredwilson

      Yes and noMany funds have leveraged past successes into creating megafunds which means style driftThe only funds where past performance can predict future returns are those that keep the same people investing roughly the same amounts in roughly the same sectorsAnd even then, you have to factor in technology change and disruptionThe great VCs of the comm equipment era are largely gone from the asset class as that model doesn’t work as well these days

      1. jeremystein

        I should have been more clear– I meant relative to other asset classes. Past performance is not necessarily a good indicator for a money manager. There are plenty of them who merely found themselves in the right place at the right time. It is possible that they still have a long run negative statistical expectancy which will lead to a fund blowing up (I just covered this concept in a post).I think that is very different in VC. If you leverage a past success into creating a megafund, I think your statistical expectancy will increase regardless of past luck, even if that means style drift. In some ways, VCs are a bit different from public money managers (lets assume that we are excluding PIPEs). While the best VCs will still actively source deals, the extra visibility will help unsolicited deal flow. It may even lower the price of the deal which will increase returns. Of course, these funds may not produce the same types of returns– but I imagine that they will still outperform, or come close to the asset class average.Just like every industry VCs still need to adapt. Every model breaks at some point. A lot of the largest companies 50 years ago don’t exist today. If you are able to adapt, I believe that there will always be opportunities to invest in every sector.

        1. fredwilson

          I totally agree with this

        2. Sean Glass

          I think you’re saying that there is a “Brand” effect that increases returns for funds that have had success in the past. That is definitely the case. Entrepreneurs know that there is value to having Union Square or First Round or KPCB as a backer. If this brand lift gives the company a 1% greater chance of success, that can go directly to the IRR of the fund. I suppose it could lead to a reinforcing effect. Non-linear models may also be interesting to consider when modeling out a multi-found game (combined with a predator-prey or “genetic” model).

          1. fredwilson

            It is success leads to more success until you let success get to your head. I try so hard not to do that. Its hard

      2. jeremystein

        I think that statistical expectancy is the most important metric. Many returns aren’t sustainable because you can always run above or below expectation. But in the long run, your results always revert to the mean.

      3. jeremystein

        I should have been more clear– I meant relative to other asset classes.Past performance is not necessarily a good indicator for a money manager.There are plenty of them who merely found themselves in the right place atthe right time. It is possible that they still have a long run negativestatistical expectancy which will lead to a fund blowing up (I just coveredthis concept in a post).I think that is very different in VC. If you leverage a past success intocreating a megafund, I think your statistical expectancy will increaseregardless of past luck, even if that means style drift. In some ways, VCsare a bit different from public money managers (lets assume that we areexcluding PIPEs). While the best VCs will still actively source deals, theextra visibility will help unsolicited deal flow. It may even lower theprice of the deal which will increase returns. Of course, these funds maynot produce the same types of returns– but I imagine that they will stilloutperform, or come close to the asset class average.Just like every industry VCs still need to adapt. Every model breaks at somepoint. A lot of the largest companies 50 years ago don’t exist today. If youare able to adapt, I believe that there will always be opportunities toinvest in every sector.

      4. jeremystein

        I think that statistical expectancy is the most important metric. Manyreturns aren’t sustainable because you can always run above or belowexpectation. But in the long run, your results always revert to the mean.

  7. Simon Cast

    I wouldn’t assume a power of exits. From what I remember of the stats from StrategyEye, I think a Guassian (normal) distribution is more representative. The distribution would be skewed towards the lower end (probably in the are between about 30 to 100m exits). The number of low exits is small just as the number of high end exits is small as well.I can see that the bulk of the return is concentrated in the bell (albeit skewed) as that is the point at which 3+ returns are being made, whether or not the absolute amount of return is achieved at that point.If a skewed normal distribution is real then it means that there is a point of company valuation that should act as a break on VC investment. Anything above that point is likely to result in a very low return for the VC. Funds that investing in companies with valuations for less than $20m are likely to produce decent returns over time as the probability of exiting decently is higher than investing in companies with valuations above $20m.Having thought this through I do wonder whether late stage investing will work as expected. While the risk of failure is much lower, the probability of an exit that returns more than 1x is so much lower. I don’t think the lower risk is offset by the return.Fred, StrategyEye collects stats on exits and will have some representative data on the number of exits at different valuations. Might be worthwhile seeing if you can get your hands on them.

    1. azeemazhar

      We analysed several thousand VC exits from Venture Source and VentureXpert in 2007. We clearly saw a non-normal distribution. Exits can (and do happen at zero) but they also happen at $100bn. So it is a classic powerlaw. It may have the characteristics of a high kurtosis, highly skewed Gaussian at some scales but it isn’t. It’s a powerlaw or possibly a Poisson.

      1. Simon Cast

        That is odd as with the power-law distribution the number of exits for very low amounts is orders of magnitude larger than that in the middle. Normal distribution can still have exits at either end merely the number is really, really low.I would like to have a look at the data as it contradicts what I would expect. Not that I am saying that I am right merely that it contradicts. Bear in mind I am talking about the number of exits as various valuations.Admittedly, it could be that exits are random. The distribution of exits is neither normal nor a power-law. The current financial crisis is a good warning not to get carried away with models.

        1. PhilipJames

          I agree with Simoncast – what are the axis supposed to be on the sample power law graph above?I’m guessing its number of exits by value of exit or something similar. If so, it doesnt make sense. I understand the logic – there’s only 1 $10bn exit per year, a few $1bn exits, lots of $100m exits, and lots of others, and no one cares about the tiny ones.If you graph this, exit price by number of exits, you’d end up with a classic bell shape – most exits are in the middle. I believe that the average (mean, not mode) VC exit is around $80-120m. Call the distribution Poisson if you like, you can play with mu to get a variety of shapes.You could fit a power law curve to the exits if you graphed probability of achieving that valuation by valuation. ie, its highly probable that you can build a $1m company, but very improbable that you can build a $1bn company. Thats a rather artificial construct, and I don’t believe that the most common exit is also the smallest (how many $1m exits are there?). So the power law graph won’t hold.

      2. fredwilson

        Has there really been an exit at $100bn?That would be fantastic if its true

        1. azeemazhar

          Sorry. Fat finger error. I meant $10bn

      3. Michael F. Martin

        Yeah, I was just going to say. It can be hard to distinguish between a Poisson distribution and a power law. In most cases where people say “power law,” I think they mean “power law to a good approximation within this range of the domain.” True scalability is rare. If it’s a poisson distribution, then it could be approximated as gaussian pretty well in the limit of exits at very low frequency. So maybe exits this year might be well approximated by a gaussian. But averaging over long time periods, probably a poisson distribution is a better fit.Another thing to bear in mind is that the fit for a power law approximation is going to depend on whether you include the set of exits in the late 1990s early 2000s in your set.My guess is that VC investment scales better than you’d expect looking at the distribution of exits over the past few years, but worse than you’d expect from looking at the late 1990s and early 2000s. The margin for error in that guess is wide enough to easily capture the numbers that Fred projects above.

        1. fredwilson

          You guys are awesome. I asked for math whizes and I got them!!!

  8. azeemazhar

    In a free market, money will flow to asset classes with the right characteristics, as you saw.One thing that seems to be mis-priced around venture is the illiquidity premium (10 years is a long lock up).The second is the tail risk of venture investing. Since we know that venture returns are not normally distributed (not even with high kurtosis) but are powerlaw distributed, our typical models for asset allocation don’t work. These are based on an assumption of a Gaussian which is a bad approximation for liquid assets like equities (as recent months have shown) and a terrible one for venture.The allocation models of investors in venture funds assumed differently. Indeed, the tail risk of the asset class generated a notionally high option value making venture funds more interesting to allocation models of LPs.Frankly, if I was a pension fund and I couldn’t get into one of the fifty firms (in which I included both of yours), I wouldn’t come close to the venture class.There is another angle on the issue of the over-allocation of funding to venture capital firms. Very bright, persuasive people get into VC because it is a fantastic business. Even if you don’t enjoy the notion of working with tech firms or brilliant entrepreneurs, as an asset management business a two per cent management fee makes for a decent lifestyle even with zero return. Needless to say lots of smart bright people are drawn to it, creating lots of smart, persuasive PPMs floating around the pension fund world and lots of funds drawn in.One way of addressing this over-allocation would be for successful funds to reduce their management fees and perhaps increase the allocation of success fees. That way LPs would see the signal of low management fees as successful funds, thereby eschewing funds charging higher management fees. VC would become relatively less attractive to GPs, reducing the number of funds and hence the over investment in the class.Reverse logic. But I think it’ll work.

    1. fredwilson

      This is such a great comment and there are a half dozen truths in here that are not obvious to everyone

    2. Facebook User

      The 2% fee definitely misaligns incentives at the low end. But it’s legit to the extent that the fees are used for what they’re supposed to be used for, which is covering operating expenses of the fund. You’d get much the same effect by putting in clause that capped what salary the managers could take home from the fees. But most LPs aren’t going to have the stomach to enforce those clauses, so in the end you’re probably right.Actually there are a few private equity funds out there that don’t charge a fee but take more carry.

  9. disqus_2cZZY1Rtg0

    “I think “back to the future” is the answer to most of the venture capital asset class problems. Less capital in the asset class, smaller fund sizes, smaller partnerships, smaller deals, and smaller exits.”You can see evidence of this already happening as focus on smaller investments grows through things like TechStars, and larger funds moving into the small investments area (e.g. Sequoia with Y Combinator).Also, might be interesting to factor in how much it costs to start a company. The decline in the cost of technology and bandwidth removes some of the variables that would complicate a startup in the past, leaving the focus on the value of the idea.

    1. fredwilson

      This is certainly true in ITNot sure its happening in biotech and cleantech

    2. fredwilson

      This is certainly true in ITNot sure its happening in biotech and cleantech

      1. Sean Glass

        But doesn’t lower startup costs + more capital = creation of more competition = lower margins = lower returns unless you back the company that ultimately wins and absorbs the others?

        1. fredwilson

          Web services are largely a winner take all market if you define the market narrowly enoughIn ecommerce, we’ve got amazon, ebay, craigslist, gilt group, etcYou’d say ‘its not winner take all”But each of them owns a category

  10. Eric Olson

    @Proales: Interesting thought. However, one thing to remember is that Fred is referring to VC backable companies, which are a subset of all entrepreneurial activity. The answer to your question, in my mind anyway, is “it depends.” If all of the cities, towns, D-1 universities, etc. are telling people to start VC backable companies (as opposed to nice small businesses or businesses that don’t need capital) then perhaps we do need to rethink the long-term sustainability of that message. If these places are suggesting that more companies should be started (including companies that don’t need VC funding) that may not be a bad thing. A lot of great companies both big and small have been built without VC money and, therefore, don’t figure in to VC asset class problem Fred is discussing. (These “non-VC” companies have created a sizable amount of jobs in the aggregate as well so they do provide some solid economic benefits.)

  11. apmaran

    First, I’m not at all VC side, I’m the math side, ie entrepreneur.There’s a lot of holes in your math, first is that you assume a lot of relationships which is kind of a bad start for a math calc:-) and as it happens I’ve studied the exit market and it’s not linear at all because you need to look at each venture fund and it’s own flow chart as they are ruled by many known and unknown factors inside, then add them all up into a multiflow curve which is hard work and would be looking like +200 worms fighting hard to get first.When it comes to this type of empirical math you could do a comparison to ecological behavior in semi closed areas (as the VC market is?). There’s been thousands of studies on competing species and for me this is what this market behaves like.Play with the thought that the energy/food (investment objects) is in variation due to creation factors like mineral/rain/soil (knowledge/education/drive) and the different species (VC’s) living on the objects fight over it. Some loss some gain, the overall growth is determined on skills, institutional support and willingness to dare, something like that. Excuse my fluffy view on technological ecology versus real life investment hell; for me this is something Levitt & Dubner know much more about than me…

    1. fredwilson

      Is their a solution to the problem using your approach? It seems like an interesting way to think about it

    2. fredwilson

      Is there a solution to the problem using your approach?It seems like an interesting way to approach the problem

      1. apmaran

        If I could be blunt, you need to rephrase the question.This is some kind of game theory dilemma involved… and for me it means that the risk investment market isn’t a restricted area to work with, ie impossible to nail down in one line theories. Let’s set it into two perspectives (there’s probably many and variations on these) a. single fund and b. all funds.a is “simple”, fund managers know how much they raised and put in and know the expectations on them to deliver, some skills, luck and approach style later, voila enough return to be happy or in case of not, put band aid on and get back into the game.b is where we need to ask ourselves why we want “to figure out how much in proceeds every year need to be generated to deliver a reasonable return to the investors”. I can’t see the relevance of that, sorry, maybe you can be so kind to help me see why?For me the question would be more like -“what are the important drivers that needs to be maintained every year to make the VC market grow, create new funds and secure the return potential”, growth is in here because of two things, one is that I assume that everyone working in VC stuff wants the recognition and respect it deserves from other investment markets, entrepreneurs, etc second is that I see in your first table that in five years the cyclic movement is back to the same level as in -04 when it comes to number of funds but in money raised – it’s 50% up which indicates some growth idea.Proceeds are one driver to maintain growth, there must be several others that needs to be looked after, I would think the overall experience/knowledge capital is important maintained by looking at aggregated results and win/loose factors; trending, tech leaps, money and tax distribution systems/reforms and on. It comes down to how to grow the VC market with stability and sanity; as it’s easy to attract gamblers the opportunity to grow is very good but that will also disrupt the question you set. back to your question to me, if there’s a solution to the problem, I would put the question back; Is it a problem at all and is the question relevant to what you want to know?

        1. fredwilson

          This is great advice and insight but it makes my head hurt! 🙂

          1. apmaran

            Totally my intention 😉

  12. fnazeeri

    Couple of thoughts:(1) Instead of guessing the total value of exits, I think you could just run some reports in CapIQ or a similar database to get the total (global?) M&A deals between $20MM and $5BB over the past 10 years and add to that the total IPOs in the same range.(2) It’s hard to isolate the US data since I would guess that a significant portion of those dollars are invested in foreign companies. I’ve not seen a global VC number but you could probably find a close proxy by visiting the websites of the country venture capital associations like IVCA, BVCA, CVCA, IVA, et al.(3) If you have the two bits of historical data (i.e. money “in” and money “out” then the big guess is what percent of the “money out” do VCs own. My guess is that 50% is low since most deals at exit would have 3+ rounds of funding including late stage private equity that has bought out founders. If I had to guess at what percent VC/PE owns at exit, I’d guess it is more like 75%+.Ok, that was three thoughts.

    1. fredwilson

      Great thoughts. My hope is that this post will spur enough conversation and ideas that someone with acces to the data can do an accurate analysis of this math problem

    2. TU

      I think #1 would capture a lot of non-venture activity. You would need to qualify it further and do some more analysis. For IPOs, the typcial wall st. databases would allow you to identify anyone owning more than 5% of each IPO issuer, from there you could back out the funds.For M&A, I haven’t worked with that data in some time but my guess is that you would filter the list to where at least one of the players was private and then you either find some great database that ties back to their venture funding or you are stuck doing some serious legwork.But surely the some venture related association must track estimated venture backed company exit valuations in the aggregate?

  13. Peter Cranstone

    Fred,Let’s make the math simple. (and the see if it scales). You raise a $100m fund. The LP expects a 3x return back to him. So you have to turn that $100m into “at least” $400m (this covers your expenses and your carry). Now so as to not get eaten alive by inflation (none at the moment) and interest (low at the moment) you have about 4 years to return that asset to the LP. You have then a year to invest that $100m (get it working). Lets say that you invest $10m each time (huge number). You will have to do a deal a month (then take two months vacation for a job well done). You sit back and wait for 3 years for all your investments to exit (not happening now). You know a 1/3 will fail ($33 million down the drain) you know that a 1/3 will break even ($33m for the LP – not good enough you have $367 million to go). Your final 1/3 (actually 4 investments) all hit a five X. So that’s $200m returned to the fund. So lets check the math so far:1/3 – nothing1/3 – even – $33m1/3 – 5x – $200mTotal returned over 4 years $233m. Pretty good actually – I doubt the LP could double his money in 4 years.Now for reality. There’s no way you can do 10 ten million dollar deals in a year. It will take 8-10 years to spend that amount. There’s no way that 4 of your companies are going to exit with a 5x multiplier (not for the next 10 years anyway).So what are the chances that your class (VC) can return even a 2x on the $100m in a 7 year time frame. I bet it’s a lot less AND with a lot less risk than the LP can do himself by placing safer investment bets.The Black Swan for the VC’s is now clear – there are NO exits and I need to add a qualifier here before you jump on me – “In a time frame that makes it a viable asset class”. It’s the “AND” that is causing the problem. You will have to hold longer, your companies will HAVE to be profitable and they will have to be either acquired or IPO (huge barriers).You math above shows that the VC business has now changed forever. You got a reprieve in 2000 because Alan Greenspan made money cheap and Wall Street did some financial engineering – but it was all smoke and mirrors. The chances are that it won’t return for at least a decade which means that the Entrepreneurs really have to build “measurable, sustainable, profitable revenue from volume” for less than $15m invested and then drive up the value so that there’s a decent enough valuation that the price per share that you exit with is high enough. (Assumes that you got in early enough).And if your still reading (thank you)…so far I see you as one of the few VC’s who understands the business well enough to get in early and set the capital structure of the company… because what few realize is that capital structure of a company is going to play a HUGE role going forward IF you are lucky enough to get an exit.Cheers,Peter Cranstone

    1. fredwilson

      PeterI did read the whole comment (thank you)Most of what you say is right. But we have deployed about 100mm in our first fund over five years by starting with 250k to 750k seeds and then scaling our investments over timeWe now have 7mm to 10mm in most of our top third which is about 7 companies including one which we’ve exited from (tacoda)We called that 100mm over five years and will call the final 25mm over the next five years. Our capital calls over time looks like an s-curve with the steep part of it being in years three and fourThe model still works if you do it right

  14. Max Kennerly

    I do contingent fee lawsuits and so operate off a comparable business model. So when I write “venture capital” below, you can also read “contingent fee lawsuits.”Someone who is expecting a regular return in excess of other classes of investments from their venture capital is doing it wrong. Although investment into many companies at once may represent diversification from a venture capital standpoint, it certainly does not represent diversification from a global investment standpoint, not least because venture capital funds tend to focus on particular industries, and so can’t truly diversify even if they tried.If you want a regular return, you are more than entitled to invest across multiple well-established industries in a public companies that produce dividends or private companies with a long track record.Venture capital, however, is by definition used for industries and businesses that are not well-established and so cannot and will not produce regular returns to anyone, certainly not a purely capital investor who is not only removed from management of the business, but has another layer of management (the venture capital fund) between them and the money.Applying the 80/20 rule, 80% of even successful venture capital investments won’t produce a truly large return justifying the overall risk — you do it to get in on the 20% that do produce those large returns. Same goes for my business: 80% of the successful cases produce “okay” returns, not enough to justify doing contingent fee instead of hourly billing or the like. 20%, however, make it worth the risk.

    1. fredwilson

      I love the analogyWhat’s a big win in the litigation business?

      1. Max Kennerly

        Depends on the resources devoted to it, so let me give some examples based on actual costs and number of attorneys on the case.(Someone might ask, “why not use billable hours for resources?” Well, contingent fee attorneys almost never devote themselves entirely to one case, and each minute spent on the case instantly becomes a sunk cost, so we generally ignore time already spent on a case and focus on two things: actual costs and opportunity cost due to the lawyer(s) having to turn down other work. I refer to the latter as “bandwidth,” i.e. the availability of a lawyer to take on other work. Keep in mind also you’re paying these attorneys (including yourself) a salary, and thus have a significant carry cost, although the salary on a ‘per case’ basis is quite low given how most attorneys have over 10 cases, even those on substantial matters.)A large-damages personal injury / product liability / medical malpractice lawsuit can be done by one or two attorneys and costs below $250,000, with recovery of $5-$10m within 1.5-3 years. That’s a big win: you put in $250k out of pocket, likely didn’t impair bandwidth, and recovered $2-$4m in attorneys’ fees.The numbers aren’t too much different for most small business cases, with breach of contract, unfair competition, etc.A regional-market antitrust / mid-sized patent infringement case can be done with 3-6 attorneys, $1-$5m in costs, with a recovery of $15-$50m in 2-4 years. Another big win: you put in $1-$5m out of pocket, moderately impaired bandwidth, and recovered $7-$20m in attorneys’ fees.A massive shareholder class action / national antitrust / large patent infringement case can be done with 10-40 attorneys, $10-40m in costs, and a recovery of >$100m in 4-10 years. Think of the Blackberry patent infringement case, which ended with a $612m settlement and over $200m in fees (resulting in profits-per-partner than year over $4m).Point is, it’s quite hard to tell if you’re sitting on a Blackberry case, with big money in only a few years; an Exxon Valdez fisherman business damages case, with good money twenty years and millions out of pocket later; or a total loser.Can you make a regular return? Sure, and we do, but that’s not what justifies the risk. What justifies the risk is the potential for a Blackberry case, which allowed Wiley Rein’s profits to vault over every other (publicly disclosed) profit in the country, including perennial winner Wachtell Lipton, while still putting tons of cash in their bank to keep looking for the next one. Wiley Rein *could* have paid out double what they did to partners, but they looked ahead, much like your investor should have.

        1. fredwilson

          Its too bad there’s no contigency model for defending suits from patent trolls

          1. Max Kennerly

            You may appreciate Jay P. Kesan, “Carrots and Sticks to Create a Better Patent System,” 17 Berkeley Tech L.J. 763, 770 (2002) (proposing “a one-way, pro-defendant fee shifting system for situations where patents are invalidated or revoked in a litigation or opposition proceeding based on certain categories of prior art that are reasonably likely to be discovered by a diligent patentee.”)There’s a short, biased report on introducing “loser pays” to patent litigation available at think full-fledged “loser pays” goes too far and would bias in favor of the biggest companies, but Congress could do a lot to create a market for defense of unmeritorious claims through fee shifting. Attorney’s fees create the market for most contingent fee pursuit of discrimination and constitutional violation claims, and support antitrust, copyright and patent claims.

          2. fredwilson

            Max – you are a smart guy. I’m gonna start reading your blog

          3. Max Kennerly

            Thanks, “I’ll read your blog” is about as high a compliment as can be given these days. I’ll try not to bore you too quickly.Since we’re going a little off-topic (and this looks like a post lots of people will come back to for the main topic), I won’t respond directly to Brad, but anyone who’s interested can read this post on my blog:

          4. bfeld

            If they did one-way loser pays (e.g. plaintiff has to cover defendants cost if the plaintiff loses) and they prohibited contingency fee relationships that would solve a lot of problems.

  15. Tim Cullen

    I’m not sure how it affects your calculations, but it might help to consider one additional input factor (touched on by PRoales and apmaran): entrepreneurs. I suspect that the irregularity of the returns and/or lack of scaling might be due to the variable supply of entrepreneurs. Macro-economic (e.g. unemployment) and public policy geared towards breeding entrepreneurs are just a few factors that can influence the supply of founders. In the model postited by apmaran, you can put lots of food in the closed ecosystem, but if the supply of animals is limited, the resulting vibrancy of the ecosystem will probably have a ceiling independent of the food supply.

    1. fredwilson

      I agree but that’s not an immediate effect, right?

      1. Tim Cullen

        I don’t think what I am describing is an effect, per se. It’s more like an inherent condition of the system.My basic point is this, building on the ecosystem metaphor: if the amount of food (money) is in excess to what can be effectively consumed by the animals (entrepreneurs), adding more food will not alter the outcome (beyond wasting food). Thus, if one thinks that VC is not scaling well as an asset class, then it might be that the proportions of the inputs are incorrect, and need to be readjusted–in this case more entrepreneurs might need to be added to the mix.Whether the factors affecting the supply of entrepreneurs (the economy, public policy, etc.) is fast-acting or slow, positive or negative, I leave to smarter people than me.

    2. qwang

      There is yet another input besides money and entrepreneurs. Isn’t there a structural limit to the supply of opportunities that capable of delivering out-sized returns? The kind of returns VCs expect can only come from new businesses that exploit deep structural and technological changes in industry, and that constraint does not scale with the amount of money or quality+quantity of entrepreneurs pumped into the system.They way I see it is, there are only so many problems at any given period that are ready to be solved, and prove lucrative while doing so. Some ideas are simply too early ( and others face overwhelmingly powerful gatekeepers (digital music), and other times, the technology simply isn’t there.

  16. terrycojones

    Hi FredYou say:> about 200 exits per year produce all the returns in the business. I think that number is> too low because that is about the number of venture funds raised each year.I don’t think your argument is right here. Arguing that there must be more than 200 exits because there are 200 new funds is a bit thick on the VC sense of entitlement 🙂 What if there were 1M new funds this year, would it follow that there therefore had to be 1M exits?Terry

    1. fredwilson

      You may be right terry but there have been 200 funds raised every year for a decade or moreIf the vast majority of them had no exits, don’t you think there would have been fewer funds backed?

      1. terrycojones

        Yes, your argument is right looking back and summing/averaging all funds and all exits over time. But that doesn’t make it right going forward, especially if we’re in a financial crisis, there’s no IPO market, etc. You’ll probably be right again if we look back in 15 years, but you may not be right if we look back in 5 years time at the years 2009-2014. Anyway, it may seem I’m nit-picking, and I agree that funding levels certainly will adjust (rapidly) to returns. The short term future numbers may not look like the long term past numbers. The number of funds and the number of exits are tied, but given that fund sizes and that exit sizes (as we’ve just been discussing) are non-linear, tying them together with a linear argument sets off alarm bells. It could still be right though 🙂

  17. Kenosha_Kid

    I wonder if another issue in the VC challenge is a valuation issue. In other words, what if 50% ownership is not the right allocation (i.e., too low)? In early stage investing, and especially in an industry as prone to constant change as digital media and IT, one could make a compelling case that value going into a deal is option value. When we’re talking about 50%, that’s more like full equity value, I think, especially when considering the absolute dollar amounts. Maybe, however, this also comes back to scalability, since too much capital chasing the same deals will lead to risk/return imbalances, much like the sub-prime credit bubble, etc.

  18. terrycojones

    On the math side, you can compute the sum you’re looking for using a simple guess at an appropriate power law. If we plug in a = 5B and k = -1 you’d have a classic Zipf’s law distribution of exit sizes (exit ranked 1 with value 5B * 1^-1 = 5B, exit ranked 2 with value 1/2 that, etc). But Zipf-like distributions are often more accurate (and more appropriate) with an exponent bigger than 1 (i.e. smaller than -1). Using f(x) = 5B * x^-1.1 and summing the top 1000 exits (via a 7-line Python program) gets you to $27B in total exits. So your guesstimate was pretty good 🙂 And due to the nature of the power law, summing the top 200 exits with that model you get $23.5B in total exits. So it doesn’t really matter whether you count the top 200 or top 1000.

    1. fredwilson

      First of all thanks for that math. I knew someone could deliver for me and I’m glad it was you terrySecond, are you saying that it doesn’t matter if its 1000 exits or 200, if you start with 5bn, you’ll end up with 25bn?If the number is indeed 25bn, then we have a very big problem

      1. terrycojones

        Yes: sum the first 200, or the first 1000 and you’re in the range of $25B. Sum the first 10,000 and you still only get to $33B. But that’s just under that model (with a -1.1 exponent), and the model may not be appropriate. If we take the higher $10B figure for the first exit and use an exponent of -1.05 and 1K exits, you get to $64B in total exits.It’s impossible to say which constants are better without more concrete data.If we had a list of the top 10 exits (say), then we could fit a power law: i.e., look at a range of exponents and pick the one that minimizes the error between reality and the model. That would be more useful. But even that is a guess, because we’re ignoring the lower order term in the power law equation, and because the existence of a power law is itself a hypothesis. But it would be a pretty good starting point.

        1. fredwilson

          Now we are getting somewhereWe need the data and I am getting it shortly

        2. lacker

          This fitting strategy reminded me of an old post -…says Fred:”The way I like to think about this is one deal returns the fund, another 3-4 deals returns it again, and the rest return it a third time to get to the 3x gross that a fund must hit to deliver good returns to the LPs.”If that really holds up then that’s enough info to fit a power law – if deals 2 through 4 are equal to deal #1 that’s enough information to get k. That’s about k = -1.08. If it’s deals 2 through 5 equalling deal #1, then it’s about k = -1.23.But I think the lesson here is that this modeling the area under a power law based on its largest element is very volatile. We are assuming that 5B top-exit is constant year to year, which is a very weird thing to assume. For example I would expect that the value of the 10th highest exit is more consistent because it has more fluke resistance. (Or replace 10 with any n > 1.) So I don’t think this sort of model is really a good one. If we want to make a good model we should start from the data we are most confident in, which is probably just a big list of exits with their values, and then we can fit to that.

          1. terrycojones

            Agreed. Note though that you’ve switched here to discussion of a single fund’s returns, which will show much more variation than the industry taken as a whole. In any case, you’re right about the models – they are just that. They’re bad at the tail too (especially if you just consider a single VC) since many models will produce numbers infinitely but a single VC has just a small number of portfolio companies, and some of them will produce zero return (and overall are a loss, and so should perhaps be counted as negative in a more accurate model).It will be interesting if Fred comes up with some numbers. I don’t have much time to be building models, but if they’re simple I’ll probably play with them a bit for fun.

          2. fredwilson

            I’m working on it. I do have a day job (like everyone else here) so it won’t happen overnight

          3. fredwilson

            I’m getting schooled and I love it

    2. terrycojones

      Ah, sorry, I misread your guess. You’ve estimated that the top exit of $5B would represent 5% of the total value of the top 1K exits. But if the distribution looks anything like Zipf’s law, that exit is actually 5B out of 27B, or about 20% of the total value. You would need a very different exponent in your power law equation to get the sum to be 100B if the first-ranked exit was 5B (using k=-0.75 gets you to $95B).You can see the difficulty: with just one data point (the size of the biggest exit) we don’t have anything much to help us pin down the constants in the equation. Is k=-0.75 out of the realm of the reasonable? I guess some people would argue that it is, given the huge number of natural phenomena that follow a power law something like Zipf’s law with an exponent less than -1.0 (i.e., in the direction *away* from -0.75).

      1. jrh

        Zipf’s law is really about ranks, not total amounts, but it was also my first thought on reading Fred’s post. (I work in computational linguistics.) If you plug in some numbers and look at alternatives, it doesn’t seem to matter much what the actual distribution is, and the values of the parameters don’t make a huge difference, as long as you’re in the reasonable zone. The lesser exits have very little influence on the total compared to the big exit, which is the assumption. Sounds like you got a similar result — the initial assumption matters more than the model you choose.Also, I think that the total exits is highly dependent on the total number of funds. So as you add more exits to your calculation, you are also adding more funds, so the average return is not going up (and probably down) even as the total return goes up. That’s not an explicit factor in Zipf’s law, or any other power law I’m aware of.If I were going to apply statistics to the problem, the interesting question would be how many large exits to expect over the time period of current venture funds. Seems like a Poisson would be the most useful model.

        1. terrycojones

          Hi jrh> Zipf’s law is really about ranks, not total amountsYes. That’s why in my original comment I put in parens “exit ranked 1”, “exit ranked 2” etc. But Zipf’s law relates the ranks to the amounts (be they word frequencies, company values, species counts, etc). It tells us that the frequency count of the 2nd ranked element will be about 1/2 that of the first count, that the 3rd ranked will be about 1/3rd of the first count, etc. So it’s connecting ranks with the total counts. Sorry if I was unclear in explaining what I had done.

        2. fredwilson

          I love reading these comments. Makes my day

      2. fredwilson

        Right. We are back to needing to know the exponent. Damn. I’m out of my league

  19. Michael F. Martin

    Since the IPO market is now closed, I believe the limits to the scale of VC investment will over the next few years at least be set by the money that flows to startups from public company buyers.My question about this is why do VCs emphasize their relationships to entrepreneurs to LPs? Are their VCs that emphasize their relationship to large companies?http://brokensymmetry.typep…In general, I think most VC funds could do more for the public company acquirers of their startups. For example, how many VCs have spent much time visiting with the engineers at Google, Cisco, and so on, figuring out exactly what they are most interested in seeing developed?I think part of the answer to this is that Google and Cisco don’t like people to know too much. But a VC with a good reputation should be able to negotiate confidentiality.Patent rights together with know-how from a small group of engineers with one or two business people is all that are necessary to develop a product or service idea enough to justify acquisition in many cases. The traditional model is on steroids.

    1. fredwilson

      That’s what many VCs did in the late 90s with cisco and the otherbig comm equipment companiesWe’ve sold companies to yahoo, google, and aol in the past few years and know those companies wellBut they are not going to tell us what their most strategic needs are and even if they did, it would take a few years to get a company funded and to the point where they’d buy itI believe you need to do a wayne gretzky and figure out where these companies will be in a few years and head there now

      1. Facebook User

        That answer makes all sorts of sense to me if the goal is to make a Schumpeterean transition from one technology to another.But what about for all the incremental steps in between? Startups and universities are way better places to spitball ideas.I’m pretty sure there’s a gap at the low end of the market in the verticals between consumer Internet and pharmaceuticals in terms of capital intensity. In other words, consumer Internet companies can almost bootstrap. Drug startups have patent rights that still work reasonably well. What about all the verticals in between these extremes? Think new materials, faster processes, &c.Anyway, I hope the companies open up more because our future as a country depends on them doing so.

        1. fredwilson


          1. Facebook User

            One possibility is for VCs to partner with specific large companies. It’s nice to make Yahoo bid against Google for one of your startups in the shortrun, but over the longrun you see more $$$ by aligning yourself with the buyers.This is not the equilibrium we see now, but perhaps that is because VC funds had more leverage when IPOs were a viable alternative exit.

          2. fredwilson

            I do not think partnering with large companies is the answerI think competing with them is a better way to make money

          3. Tereza

            Violent agreement here. Strategically, that alignment doesn’t work. It yields lower exit values and mediocre innovation. To maximize a (non-IPO) exit price you want potential that strategic buyers will compete and bid up your exit – a premium they pay because they did not have stake in early, and now need to make up for lost ground. That’s why it’s “venture capital”.Imagine a so-called venture fund aligns with a large co, and is taking on early-stage risk with an investment. If the venture still fails, despite the “strategic alignment”, they won’t buy it. Or – say you set up that relationship so they agree to cover your risk and compensate you for it. Then in effect they are your investor….or your insurance provider? That comes at the price of professional latitude, narrow and incremental innovation — lots of product extensions (yawn). You are effectively an outsourced R&D department, or at best, a company’s own venture fund on the lookout for “synergies”. It can be a legitimate function, but it’s not what people join the VC profession to do, nor what great entrepreneurs seek.Personally, I find VC and VC-backed companies more compelling than corporate-backed because after years on the corporate side, I’ve seen time and again it’s far too easy for corporations to choke off great breakthrough innovation.

          4. Michael F. Martin

            Well then we need an IPO market.I appreciate the perspective.

  20. Harry DeMott

    I’m not sure I can help with the math at all – but I can help with your investors issue.It is the same issue that every investor in every asset class faces – which is that no asset class outperforms every other forever. There are times when VC outperforms and times when distressed outperforms, but if you can figure out the timing on that you are smarter than everyone out there. (And given the inherent lack of liquidity in a number of asset classes you really better have your timing right) Right now liquidity is probably our performing illiquidity – but for many years private investments outperformed public ones (especially as private investments benefitted from better structures fro investors and a heck of a lot of leverage)You can aggregate and disaggregate all you want – but at the end of the day in every asset class there are winners and losers who combine to produce an average.Your LP’s job is to make sure he or she picks the winners in the asset class – and stop worrying about the asset class.

  21. Rob K

    Fred- Great post. I pointed out this math problem in a prior response to a post about the VC “industry.” I think your math is directionally correct and the industry is over-invested.I few observations:I think your 50% ownership number is way low. You need to include biotech/devices, which is much more capital intensive but still has very high “miss” rate. A Phase 3 FDA trial has a 50% success (failure) rate and takes $50MM plus to get there. Even in tech it seems low to me. Series A by itself will take 30-50% of the company.I don’t think there is any way to get to $100B of exits per year. You can count on 1 hand the number of $1+ billion dollar exits and probably fewer than 20 exits from $300MM to $1B.I think it’s pretty obvious that funds themselves don’t scale. So many funds have moved to “growth equity” models because they can deploy more capital (and reap more fees), NEA, Oak, Austin, Polaris, Highland, General Catalyst, the list goes on and on.Very few funds like USV are raised each year- new funds with unique strategies, proven partners and the right amount of capital.

    1. fredwilson

      Hi RobIts nice to hear from youAny idea what the right number is for total exits per year (dollar value)?

  22. lacker

    I responded earlier on twitter but I think the math is tough to get across in 140 characters 😉 I think the “scaling exponent” earlier is actually important here not just in a mathy theoretical way.Basically there are two sorts of power law curves. When the scaling exponent is greater than 1, most of the area comes from events near the median. When the scaling exponent is less than 1, the area is unbounded because of the possibility of a single huge exit that skews all the averages.So when you assume every year has the same maximum exit, you are implicitly assuming the distribution is the first type of power law. But I’m not sure that’s the case. You’ve mentioned before that just being in on the Google investment has skewed the stats for the top venture funds. I’m not sure about stats for the industry as a whole, but I wouldn’t be surprised if total returns in some years were several times the total returns in other years.In general, if returns are really volatile (math: E[X^4] unbounded), then you can’t rely on averages to make good predictions, because it will be common for a single year to be greatly divergent from the previous years. And then there is no “steady state”.

    1. fredwilson

      Its clearly an unbounded problem but I am trying to think of it the other way to simplify things. Probably a bad idea

      1. Mark Essel

        Your gut instinct serves you well. There’s no simple model for what you want because if there was, what would be the implications for VC funded growth, your field, and our economy. Would it be predictable?Far from it, if anything history has shown us that the future is unpredictable! You’re changing the odds right now as you ask this question. Investors are reading your post and the comments and shaking their heads in confusion.Maybe some superstar VC funds and incredible startups will outperform the sum of the rest of the industry, maybe not (trying to predict tails is just bad mojo in math, and tails drive your returns). There’s probably an upper limit on how much the some of our industry can shift (extremely successful VC juiced accelerated startups are an industry shift in my mind), and some of that shift will be due to non VC startups.

  23. Stewart Alsop

    Here’s how we express it to limited partners (the institutional investors who provide capital to venture capitalists): You need to believe that any investment in a particular fund can pay off without any miracles.If you invest in a $100M fund that owns 15% on average at exit, you can get a “reasonable return” (close to 2x net) if the fund produces 4-5 outcomes of 5x or better. Total market cap needed to get that return: about $1.3B — not a miracle even in today’s economy. If your GPs are good enough to create $2B of market cap, then you can get to 3x net and a very positive return for the alternative asset class, what’s usually called “venture-style returns”. If you invest in a $500M fund that owns 15% on average at exit, you need to create $13B of market cap to get a positive return and closer to $20B to get venture-style returns.The simple question is: Are you investing in a group of investors who can produce those kind of returns with that much money? Are you investing in miracle workers?

    1. fredwilson

      Stewart – exactly right.I believe 100mm is the perfect size for early stage VCAnd you may wonder why we are managing 125mm and 155mm funds?And you may say ‘my god how did I get there’ (to quote the talking heads)The answer is our LPs really wanted us to do that (individually, not in the aggregate)Its hard to say no to money. We have to learn how to do that

      1. Facebook User

        You know the only caveat to that is that you have to consider where the money would go as an alternative. I have to say that I’m happier seeing money thrown at startups that will eventually fail than I am at buyout funds that just reconfigure a bunch of assets in a way that analysts on the big board will like better.

        1. fredwilson

          Yes as a society but not if you are a lifetime public employee watching your pension get badly invested

  24. Navanit

    You correctly assume that the distribution follows a power-law, but you apply an arbitrary constraint on the maximum ($5bn). You constrained the maximum and then concluded that the distribution is not scalable!Even if historically VCs haven’t had a larger than $5bn exit, there is no reason it won’t happen in the future. Unless VCs have some sort of self-imposed upper-bound on exits, this asset class is perfectly scalable as long as VCs have the intestinal fortitude to hang on to some hits for longer than they have in the past. (Also they need to invest in earlier stage startups and diversify more than they think they need to, to benefit even more from scalability).

    1. fredwilson

      The LPs don’t want VCs to hold onto public stock. They want it sold and/or distributed. So holding on for the big payday is not easy to do

      1. Navanit

        Why is this? Is this because they want to cash-in on the first sign of reasonable profits, or are there other considerations (structural ones, such as capital gains tax, etc.).Thanks.The Hollywood ecosystem and the VC model have similar properties. If you want to dig deeper into the math, I would highly recommend “Hollywood Economics” by Arthur De Vany. Of course, Nassim Taleb’s mathematical pieces also shed light on the subject.

        1. fredwilson

          A bird in the hand is worth two in the bush. When you have no birds, one looks as good as two

          1. Navanit

            So it’s just lack of intestinal fortitude. The people meant to benefit from one dimension of scaling don’t realize that they need to hang on to hits longer than their intuition is telling them to. Lack of discipline in other words.This is one interesting side-effect of high liquidity – this problem doesn’t seem to exist in the publishing and movie industries. The publisher and author don’t generally sell off their stakes early once the book/movie starts to become a hit. They stick it out to the end.

        2. fredwilson

          A bird in the hand is worth two in the bush. When you have no birds, one looks as good as two

    2. fredwilson

      But you are right that I should not cap the max at 5bn. Its not the right mathematical model

  25. thinkCHUA @ LivingIF

    At one time I was an analyst with a very large LP. We the history of roughly 400 venture funds over the years and roughly 100 active funds. We had data on essentially everyone, and because of it we scaled back our investment significantly. Venture Capital returns are not very good, when you see venture expert, the PWC surveys, etc, they are self-reported, none of the bad funds self report. I would ballpark that about 2/3 of our funds lost money and these included the very big names that everyone thinks makes money and have easy fundraising because of their brand. We started to realize the best VCs were small funds that did a lot of work (think 2-4 partners, $50M per). It was always the people with the angel investor, roll-up their sleeves and work with/for the companies they invested in that had the returns, not the money managers in plush Sand Hill offices.That said, the returns are not evenly distributed, they are very, very skewed to a few funds (some say the top 10), that is why people focus on top quartile, you can get lucky and have a fund in there, but if you can’t do 2-3 funds there, you are probably not someone to invest in. So as an asset class your LP is correct, VC stinks if you don’t have access to the top funds. So if your math is correct and as a class you are generating $50B in annual returns, and 2/3 of the money invested is breakeven (I am being VERY generous), it means that $8.3B of the invested capital is getting $50B in returns (sorry for the simple math which could be off), these funds report their results, a bunch of break even funds do and you get to the numbers you have. In summation, the great VCs get outsized returns and the others are lucky to return capital, the great returns are so great that the overall average is pulled up.(here is one view of it, though I am not going to invest completely on this analysis, see charts: http://www.alignmentcapital

    1. Scott

      I went to lunch with a seed-stage vc a couple weeks back. He stated that 11 funds have generated 90% of the returns.. Said VC is not a numbers game, not a connections game, not an academic game, but an action game. It’s all about action.

    2. fredwilson

      This is exactly right. I have not seen the data but my experience over almost 25 yrs in the business says you’ve nailed it

  26. Scott

    Very tricky problem indeed.Of course almost all venture investments have some sort of debt-like preferred return, so the exits can be a lot less and funds can still get their capital returned. I think this is why there are a lot of funds that just do “okay” and keep the average returns clustered around a few % IRR. To get capital returned and participate in the upside with common is of course much less frequent but is what generates the top quartile’s big returns.

  27. Jimmy

    “That would suggest each fund has only one meaningful exit and that’s just not right. I think each fund has at least five or six meaningful exits.”No, I think one meaningful exit per year on average is about right.

    1. fredwilson

      But its not one per year since 200 funds are raised each year. 200 exits per year and 200 new funds per year means one exit per fundAnd that cannot be right

  28. Mark Zohar

    What does all this say about the investment acumen and due diligence of pension funds and institutions that have poured billions of dollars into hundreds of VC firms over the past 10-15 years? What data are they looking at to assess implied risk and expected returns? More significantly, is there empirical data to support asset allocation towards venture capital as a whole going forward? And, if so, towards which sectors or funds of the VC industry?The PE and VC industries have always claimed that as an asset class they outperform the S&P over the long term. That may be true on an average basis (though I’m not sure it reflects returns net of fees and carry). But the reality as noted in the comments is that the distribution of returns is highly skewed to the top quartile (as expected).So here’s my question: who the heck is funding the bottom three quartiles of VCs? I received an email yesterday from for their upcoming New York Venture Summit. Take a look at how many VCs are going to be there:…. For the most part, these are mid-tier VCs, probably falling in the second of third quartile. Who is funding these guys? And more importantly, what actual, historical or comparative data are the VC investors using to support their investments?Clearly, a major culling and consolidation of the VC industry is pending as investment dollars and asset allocations decrease in absolute terms and investments are targeted at the top quartile; ie., the big get bigger (NEA, Sequoia, etc) and the mid and small VCs wind down, get niche or fade away.

    1. fredwilson

      Small is not badBad is badI agree with you on everything else

  29. jrh

    I’m not sure the total distribution of exits matters that much. It’s a few big exits that make the venture game worth playing. Everything else is just stuff you do so you have a shot at the big exits.There are more returns down the curve, but there are more players, too, so it’s a wash. The question then becomes — will there be really big exits? How many and how big?For the first couple decades of the venture business, the big exits were based on economics. The companies generated real revenue and often astounding profits.For the last fifteen years (starting with Netscape, imho) we’ve had some astounding exits that never had the underlying economic returns to back them up. We still had “traditional” exits based on economics, but the mix was leavened with some fluffy and flaky valuations. (E.g., Geocities was a cool site, but it’s impossible to justify the price Yahoo paid for it. )So I would boil it down to two questions:1) Are there legitimate opportunities for huge exits based on technological advances and sound economics? My answer is yes, but probably fewer than before, because many of the most obvious opportunities have already been taken. But the rapid drop in technology costs, and the rapid movement of people online, combine to make a lot of very cool things feasible and potentially very profitable.2)Is the wave of economically unsound exits (both ipo and acquisition) permanently over, or just in remission? I would argue that it’s gone for now and probably a long time.My WAG is a >50% drop in big exits, and thus a >50% drop in the asset class.

    1. fredwilson

      Good bet

  30. Brian

    What kind of indicator do you have on number of comments and general health of the VC business?I mean if this post reaches 500 comments, does that mean that the business is weak as few VC’s and fans of VC’s have little to do?Think # of comments can be an economic indicator?

    1. fredwilson

      I’m not that cynical though maybe I should beIf it gets to 500 (that would almost 2x my biggest comment thread) it tells me this is a question on everyone’s mind and people have opinions they want to shareAnd that is a very good thingVCs telling LPs that everything is OK is a bad approachVCs and LPs discussing what’s wrong and how to fix it is a good thing

  31. Scott

    Fred — As you probably know, anything in this realm really interests me, and it’s good to see that you recognize there’s an issue with too much money in unproven funds. I believe the model should be democratized and spread out, which obviously can be done through a large number of mini-funds; rather than a small number of massive funds.I think we’ll not only see new models in the near future, but also an emergence of mini-funds ($5-$30 million in funds, managed by 1-2 people). Of course, with a fund size of only $5m, investment nature would be much like Ycombinator, and the attorney fees would have to be figured out. It gives the investor more focus, and resourcefulness–almost like twitter, where you don’t think communicating something in 140 characters is possible, but then you cut down words, and you realize how much you could do with so little.- Scott from

    1. fredwilson

      I like the idea of smaller funds with one exception. If you can’t continue to participate in the deals for as long as you need to, you can get wiped out and that’s a huge problem with small funds

      1. Mark Essel

        Maybe the rules of the game need legal revisiting? Does wiping out earlier rounds make sense for successfully growing startups?

        1. fredwilson

          Yup. Because its how you get more gas for a car that is running on fumes but has the potential to get somewhereAnd that’s what’s wrong with the Obama administration’s approach to the bailout (until recenty)Not enough willingness to cram down bad money

  32. Manish Vachharajani

    Perhaps the “math problem” is larger than that of investment style, size of individual funds or anything other problem specific to the VC industry. Perhaps it is simply the number of quality companies, but it may be somewhat more subtle than that.In general, the more capital one has to invest, the harder it is to generate market beating returns. The question is, at $25B invested annually, has the VC industry reached the point where it becomes impossible for the industry to collectively generate returns that justify the illiquidity and high risk of the asset class?Certainly, there will be VCs who win big because they are managing a fairly small amount of money. But as an asset class, it may be a different story. Warren Buffet is a big advocate of the idea that the larger the invested dollars, the harder it is to generate outsized returns. Buffet has been warning shareholders to expect Berkshire Hathaway to only beat the market by a very modest amount because of this problem. According to its 2009 annual report, Berkshire invested $58B of float last year – only twice the size of the VC industry. Granted Buffet doesn’t do tech, and granted VCs have some advantages, namely more partners looking at deals allowing them to target smaller investments that have a better chance of generating outsized returns. But VCs also look at much narrower segments of the economy.So, the question remains, how big does the industry have to get before it becomes impossible for it to succeed as an asset class, and are we already there?

    1. Kendal

      Fred, this post has a bigger meaning to me. What you are requesting here, is to allow social networking to work together and provide you with some great financial data and information. Answers, lets call them. Back in the day, one would have to hire bankers, consultants etc.. for these answers. Obviously this is a small scale example, but you get the idea.

      1. fredwilson

        You are on to me. Shhh

    2. fredwilson

      I think we’ve been there since the mid/late 90s

  33. peterhoskins

    The limiter in this discussion is the number of exits and the area under the curve equaling $100 mil. If cash is not a limiter and entrepreneurship is not artificially constrained then the area under the curve is increased by quality entrepreneurship (plus a bunch if things like good exit economy). If the product if venture investment is good companies then the supply side economist would suggest all the VC industry is limited by is the creativity and ability of entrepreneurs. Which clearly has practical limitations of a different type.

  34. Nick

    This is the same ‘crowding’ problem as various hedge fund strategies (especially arbitrage) and micro-cap mutual funds etc. Have faced for years.I’ve never considered it in the VC context though, thanks for raising this.

    1. Michael Chen

      I like to see this issue as analogous to Moneyball. Since LP’s are comparing relative efficiencies of monies allocated to different asset classes. (rather than just an absolute singular black swan event ,like winning the championship, or in VC parliance getitng a single fund returner a la Google or Ebay)The smarter money managers (across any asset class, in particular VC now) should be ascribing to the Oakland A’s/Billy Beane/moneyball school of thought of wringing maximum relative efficiency out of their funds, in particular because of size, it’s the only way to stay competitive. The analog of the Venture MegaFunds (KP, Seqouia, etc) is like the Yankees, they have the biggest payroll, every player (generalizing now) wants to play there (biggest media market, biggest salary, biggest whatever).Yet we see teams doing well (Brewers , occasionally, Oakland, Tampa) and often outperforming the Yankees despite having payrolls less than what just Arod makes. Like in baseball, it’s a case of more is not always better. To wrap the analogy, paying a lot for many of the hottest free agents generally will get you wins, but they are not efficient wins. Developing young, unorthodox talent gets you both wins and efficiencies. And I think in the battle of asset allocation, VC only needs to be a more efficient place to park your money in order to remain relevant, which can be done with if you are willing to take a different structural approach to the business.

      1. hungryscientist

        Michael,I like your baseball analogy. Looking at the VC asset class in the context of gaming theory is helpful. I think the point Fred makes is the VC is a classic case of the Commons Dilemma…. To many fisherman fishing the same waters.As VC asset allocation declines due to the shrinking economy the question is how long does it take for the fishing grounds to recover? Is the recovery sustainable? Or will the first signs of recovery trigger more fisherman and put us all back in the same situation? I don’t necessarily agree with Fred that in this case the “… capital will flow to the places where it can earn an appropriate return.”Thanks Fred for the thought provoking analysis.Patrick

        1. JLM

          A lot of this discussion gets down to the quality of the deal flow. To use your fishing analogy —The worst fisherman in the world baits his hook and casts his bobber into a pond which is filled to overflowing with hungry fish and he catches 10 and goes home and fries his fish.The best fisherman in the world baits his hook and casts his bobber into a pond in which there are NO fish and goes home empty handed.Who is the better fisherman?Timing and opportunity and deal flow are EVERYTHING!

      2. fredwilson

        Good analogy

        1. deancollins

          So the question is “What do company growth builders” want from VC’s that isn’t directly tied to a cheque.Not everyone wants to play for the Yankees, and some are more than happy to support their ‘home team’ on a reasonable salary.What can Union Square do for entrepreneurs that it doesn’t already?

          1. fredwilson

            we try to do whatever we can to help the entrepreneurs we invest inI am sure there’s more we can do and we are adding new tricks all the time

  35. Manu Kumar

    @fredwilson: You are absolutely right. In fact, we had a similar discussion on one of your previous posts as well. Here is the link to my comment on that post:…The problem with VC is partly a result of LPs desire to scale the returns they saw in the venture asset class when it was small. But, as I say in my post on the K9 Ventures blog (The Venture Boutique post at… ):”The other form of scaling in venture capital is that of the size of the venture funds. Limited partners invest in venture since it is supposed to be a high-return investment. Lets assume that Shylock Ventures is capable of producing a 20% IRR on a $100M fund. Those returns do not (and cannot) scale with the size of the fund! If Shylock Ventures were now to raise a $500M fund or a $1B fund, it probably will not be able to maintain its IRR.”Your post does a great analysis of the fundamental problem with the asset class. The amount of $$s invested in venture capital need to decrease, or, the number of high value exits need to increase dramatically — or better yet, some combination of both. What the media and the pundits describe as a crisis in venture capital fundraising right now, in my view, is exactly the correction that is needed. Venture needs to return to its roots.Thanks for bringing this up to the forefront of the discussion.

    1. JLM

      The challenge may be that the quality of the deals simply does not correllate with the magnitude of money available for deals. The supply of money can increased fairly easily while the supply of deals cannot be so easily increased.If every possible deal in a given calendar year were evaluated and ranked in descending order of attractiveness and the amount of money available were doubled, it would simply result in more LOWER quality deals getting done in that year thereby reducing the average rate of return on the entire portfolio of deals.

  36. David Cottingham

    Hi Fred,To follow on from @terrycojones’ answer regarding your original question concerning the area under the curve, here’s my take on the maths. If you assume a simple power law equation of y=a(x^-k), where y is the value of the exit, and x is the number of such exits, then you can just integrate the function between x=1 and x=44 (which gives you a total number of exits of 990, if you’re adding up all the numbers between 1 and 44). The parameters a and k are ones that you set depending on what you want the curve to look like.Given that you want one exit at a value of $5bn, you need to set a=$5bn, as when x=1, a(x^-k) = a. The value of k you pick varies how fast the curve falls away: higher values cause steeper fall offs. What the value of k is set to is crucial to the outcome.The integral of the power law equation above (provided k is not equal to 1) is (a/[1-k])x^(1-k), which we’ll call v (total value of all exits). For k=2, v=a (roughly). For k=3, v=0.5a (roughly). For k=0.5, v=11a (roughly).Looking at it another way, to obtain v=$150bn, we’d need v=30a (as we set a=$5bn). This (roughly) implies that k=1/8. That would mean around 40 exits at the $3.1bn value (it’s a very slowly decaying curve!), which sounds very optimistic. Of course, that depends on the whole thing actually following a power law…Hopefully the maths is correct here — someone correct me! ;-).Cheers,David.

    1. fredwilson

      This is awesome as we know the number is not 150bn because the curve doesn’t slope that wayI think its not even 100bnProbably between 50bn and 100bn and more date will help us determine the exponentI’m learning about power laws!!!!!

  37. VLong

    Fred,I’m wondering how many companies in VC portfolios are actually currently IPOable, i.e. with at least 4 quarters of consecutive profitability, with a solid, stand-alone business model, not just a product or service, 30%-50% top and bottom line growth, and an addressable market of at least $ 5 billion. Now assuming an IPO market does come back, apply what you think is a reasonable P/E and you’ll have an idea of what the industry return could be. I bet it’s a very small number of companies because a lot of startups have been consciously or unconsciously built as products or services, not stand-alone companies or they’re just bad ideas.

    1. fredwilson

      I think a fund of 20-25 companies can produce between two and five IPOable companies if we are patient and let them stay independent and grow. We means the entrepreneurs and VCs.

  38. Nikko Strom

    At a macro level, it is trivial that the VC business cannot scale unbounded. By definition startups can only be a fraction of all business activity in an economy. Assuming markets are rational (any believers out there still?), the VC business will grow to a scale where the return is no longer favorable compared to other asset classes. So the question is where is that equilibrium? Is it at 10% returns? It should depend on what returns are in other asset classes.Markets are not perfect of course and it is plausible that investors will always demand higher returns on a VC fund than on other alternative investments, but it would be interesting to understand how much more? Then, if you perfected your model so that you can predict returns based on total VC investments, you could plot that equilibrium!

  39. Jason Fritz

    Isn’t this just market efficiency at work? No asset class should be able to generate outsized returns adjusted for the level of risk. Where high returns are available, money will come until the value of those assets are pushed up high enough to lower the future return.

  40. Ravi Mishra

    This sort of makes sense at a very basic, logical level – there are only so many good ideas out there. Pumping more money into the system just means crappier ideas get funded and (ultimately) fail, depressing the return of venture capital as an asset class.

    1. David Semeria

      I’ve flicked through the comments, but haven’t had time to think carefully about them all.My first intuition is that you have opened a can of worms here Fred. Some topics are just too complex to tackle in one blog post. For example (and excluding the fact that many VC funds are only ever fully paid up until the end) you’re tracking exits based on investments over multiple time periods. Even taking into account recent low inflation, compounding over 7 years will still affect the numbers.That, however, is just a quibble. The main problem lies in the currently fashionable issue of mark-to-market. Just because an investment hasn’t exited doesn’t mean it’s not performing for the fund. The real return metric should be the excess value generated by the entire portfolio regardless of exits. Yes, I know, you wanted to keep the numbers real – but all the same…Unsurprisingly, many of your posts are related to each other. For example, you have recently posted about the problems of valuing non-listed holdings and the fact that many investments can live longer than the fund. These aspects are all relevant to this discussion.Zooming out for a moment, I thought that all the posts asking whether capping the amount of ‘useful’ VC money implied there was a physical limit to the number of startups a given country could support were the most interesting.

      1. fredwilson

        I like opening cans of worms. I find it very enlightening and this post will lead to many more

  41. JLM

    It might be useful to look at the problem from a different perspective starting with the average % owned by founders when the pay window creaks open. This number — the amount owned by founders after multiple rounds of financing — is a fairly well agreed number. Do you agree?If one assumed that the average equity owned by founders in a liquidating event was approximately 20% and that LPs/VCs owned the balance with LPs owning approximately 20% of the remaining 80% then the ownership point of departure would be more like:Founders — 20%LPs/VCs — 80% collectively (100% – 20%)VCs @ 20% of 80% = 16% netLPs @ 80% of 80% = 64% netThe total funds at work are then diluted by the fees paid during the life of the fund at the rate of 2% per year. This impacts the total funds available for investment and the basis for calculating the LPs ROI but not the actual funds invested in a particular deal.There is a big difference between looking at the rate of return of a single deal and the rate of return of the entire fund because the entire fund is unlikely to get invested in toto and the drag of the 2% management fees on the total funds actually invested.Assuming a $100MM fund invested at 75% for five years —Total fund = $100MMAnnual management fee = $2MMAnnual mgt fees over 5 years = $10MMTotal funds invested in real deals = $65MM <<<transactional investment=”” basis=”” total=”” funds=”” invested=”” in=”” fund=”$75MM” ($10mm=”” +=”” $65mm)=”” <<<lp’s=”” real=”” fund=”” investment=”” basis=”” required=”” fv=”” of=”” 20%=”” irr=”” in=”” real=”” deals=”$162MM” required=”” fv=”” of=”” 20%=”” irr=”” in=”” fund=”$187MM” grossed=”” up=”” req’d=”” fv=”” of=”” 20%=”” irr=”” in=”” real=”” deals=”$253MM” ($162mm=”” 64%)=”” grossed=”” up=”” req’d=”” fv=”” of=”” 20%=”” irr=”” in=”” fund=”$292MM” ($187mm=”” 64%)=”” average=”” liquidation=”” multiple=”” in=”” real=”” deals=”3.95X” ($253mm=”” $65mm)=”” average=”” liquidation=”” multiple=”” in=”” fund=”3.89X” ($292mm=”” $75mm)=”” there=”” are=”” few=”” simplifications=”” which=”” must=”” be=”” indulged=”” but=”” this=”” seems=”” to=”” demonstrate=”” the=”” difficulty=”” of=”” the=”” return=”” threshold=”” problem.=”” you=”” can=”” quibble=”” about=”” the=”” rate=”” of=”” return,=”” the=”” percentage=”” of=”” the=”” fund=”” actually=”” invested=”” but=”” all=”” of=”” that=”” seems=”” to=”” be=”” within=”” the=”” sensitivity=”” analysis=”” of=”” the=”” industry.=”” i=”” suspect=”” that=”” 15%=”” is=”” going=”” to=”” look=”” like=”” a=”” great=”” return=”” in=”” a=”” few=”” years.=””>

    1. David Semeria

      By the time exit day comes, the employees usually own a decent chunk as well.

      1. JLM

        Absolutely, agree completely. That’s one of the simplifications just changes the math a bit and makes the heading a bit different — “founders & employees” own X%. Also discount positive cash flow during the investment term.

    2. fredwilson

      I think the 80pcnt is too high because in addition the the 20pcnt for founders, you need 20pcnt for managementMaybe 60pcnt, but not 80pcnt

  42. Ray Acayan

    With all the math presented here, it might be possible for quants to design a synthetic derivative that mimics the investment returns of a venture capital fund. VC funds could then offload some of their risk to other investors by establishing a short hedge on that derivative.

    1. fnazeeri

      You mean like a credit default swap? 😉

      1. Ray Acayan

        More like hedge fund replication applied to VC and private equity funds (using Poisson / Power Law distribution).

    2. fredwilson

      ‘Beware quants bearing models’. Warren Buffett

  43. jdrive

    Worth using the most recent data if we are going to do this sort of analysis…. All data are from Cambridge Associates, Thomson & NVCA – as of 12/31/08First, Long-term VC returns:10 years40.2% gross (but the dot-com runup effect still big; will dissolve in 2010/2011)20 years22.2% gross (lrunup effect not a major factor)For 20-year, by stage:Early/seed 21.6%Balanced 14.7%Later Stage14.7%All VC17.1%T-Bills, Bonds, S&P, NASDAQ, PE, LBO -– over the same time period: none close (half or less in all cases). Obviously, 1 & 3 year numbers paint a different story, but not much sense measuring ten-year + funds over 1-3 years.Upper quartile years that truly mattered: 1994-1997. Both earlier and later standard deviations much closer. Said another way, during those years specific funds mattered a great deal; in other years differences persisted, but were much less statistically meaningful.Liquidity — VC-backed IPO & M&A — number of deals per annum (avg hold time around 6-7 years now, blended)1999 5532000 6632001 4322002 4112003 3862004 5472005 5112006 5282007 5292008 332Now, dollars raised (in billions):1999 612000 972001 362002 42003 112004 192005 282006 312007 362008 28(Incidentally, in 2008, just two PE/LBO firms, TPG & Warburg, raised more money than the entire VC asset class, a combined $33.8 bil). Total PE/LBO raised in 2008: $178+ billion.Geography matters. SV was 39% of all VC in 2008; Combined with SoCal, Seattle et al, West Coast totaled 55%. New England was 12%; NY 7%. Category also matters. In IT, from 2001-2008, between $12-15 bil per year was invested. In Healthcare, between $6-10 billion, per year. (Energy/Cleantech only 6% of 2007 dollars, 12% in 2008).As one would expect, markets tend to self-correct, albeit with various amounts of lag-time. Forecasting out $25 bil/year raises ad infinitum makes no sense, and is not supported by history (granted, data above only go back to 1999-a crazy year. If you added in another decade the averages would drop significantly. Here, too, is a runup effect). Same story with returns, exits, etc. VC asset class has high variability and is not just hits-driven, but also vintage-driven.Finally, VC firms are almost all twelve-year creatures (10 in the books, plus 2 [or more] 1-year extensions), with five-year drawdowns. In the first five years, funds typically deploy 65-70% of capital, with the rest held in reserve and doled out over the remaining life of the fund.

    1. Facebook User

      Although the dot com years are obviously less of a factor over 20 years than they are over 10 years, the drop from 40% to 20% suggests that if those points were left out of the 20 year analysis, the returns might be on par with publicly traded stocks, no?

      1. jdrive

        20-years:DJIA 5.6%IXIC 8,2%S&P 4.2%Russell 2k 6.0%So, no.

        1. ventureblogalist

          This is a great post.Good data Jim, thanks for providing. Joshua Lerner analyzed returns, net of fees, for 1,252 U.S. venture funds going back to 1976. The median return for all venture funds was just under 5%. I didn’t dig here, so this could be department of awful statistics speak.Rob

          1. jdrive

            Yep. Know the study, have met Joshua. 5% was the all-venture number, 28% for top-quartile. That ’94-97 major outlier period between quartiles makes a huge difference (and it was a huge spread!), as does the less significant but still quite measurable performance variance over the rest of the years. Dot-com era also has some impact, though not as much (will less so over the next year or two).The basic assumption is that running with the stars pays. True in most everything, from bank stocks to basketball. Problem is, past performance is not necessarily (blah blah blah)… Hard to know what you’ve got. Jordan? Sprewell? Marbury? A player you can’t remember? The peanut guy? Maybe a scalper?In addition to my day job, I am also an investor in multiple asset classes, and vc is one of them. Know what I look for? The people. Know what I hope for? A rising tide. That’s the other thing Joshua et al figured out; timing matters. A lot.

          2. fredwilson

            Reporting biasI’m an investor in a vc fund raised in 2000 that has returned between 5pcnt and 10pcnt of my investment and I don’t expect much moreIt would be interesting to see if they are in the thomson/nvca dataA bad apples like that can spoil the whole bunch to quote michael jackson before he got weird

          3. jdrive

            There’s an excellent chance that they are in the numbers. All funds raised in the US from acrrediteds/qualifieds (or from anyone, frankly) do have some federal reporting requirements (Reg D exempt – a designation you can only get by actually filing) which fund lawyers do in the normal course; and the data is publicly available, if hard to find. It’s a physical record process; you would have to truck down to the SEC in VA, and once there, make sure you actually add up all the various filings related to a given fund. This is where you often see inaccurate media reporting on fund sizes; they don’t find all the filings. The NVCA is hip to all this, but also has many other sources, including US, Euro and Asian LP reporting inputs, etc., etc.The ‘reporting bias’ issue tends to only be a factor in very small funds (south of $10-20 million that ‘forget’ or do not know they have to report). There aren’t many of those. What you often do not see is ongoing reporting of mark-to-market activity. And that is a problem, but also gets into the hornets-nest of FAS 157.

        2. Facebook User

          Over the same period of time? Those look low unless you’re looking at 1990-now

    2. fredwilson

      Great data jim. Thanks for sharing it. I tend to throw out 99/00 and did exactly that in my follow up post this morningI am going to take all the data I got in the past 24 hours (don’t need logins, just readers who share), and do some deep analysisMore to come

  44. Guest

    This distribution is correct for the industry as a whole. However, I have also seen data for the fund distribution by IRR, and it has the exact same shape. What is striking is that this is by far the only class for which this can be said, all others are almost perfect Gaussians. I have a slide with the data plotted, the visual is striking, but I believe most of the readers here are aware of that. So if we superimpose the funds on your gross industry distribution, you’ll see that the “winner” funds are the ones that give you the big hits on the left, and the “loser” funds give you the tail…Why is that? Let’s conceptualize: in the public equity market you have a preselected universe of issues where you can bet, and it is rather stable. Wilshire 5000 gives you 99% of the total market cap and there is limited turnover, mostly on the well-defined bottom. So, what you have mathematically, is the conditions where you if use random strategy (throw darts) you will get a normal Gaussian at the end.With venture, you have unlimited reservoir at the bottom. Everyone with access to Powerpoint and Word can create a pitch. Moreover, the bottom is not predefined. Both Google and a hypothetical blow-up “Goofle” looked similar to the untrained eye, when they first pitched. Contrast this with public equity, where you already know the pecking order by the market cap. Even if you like a company that trades at $0.20, you are not going to pay $20.00 for their shares…Put all of this together and the conclusion is crystal clear. The VC class does not scale for one simple reason: dearth of good VCs. To be successful in this class you need to have capabilities that far exceed the “random selection” approach, and very few have them. The source of pitches is practically unlimited, even if you are good at sieving out 90%, you are still drowned out in low-quality pitches. You need to be able to dismiss 99.9% or better of the low-quality pitches. Only few can do that, and they have limited bandwidth, that’s why the class does not scale.And that is why your investors are happy with you and are not happy with the class. There are just too many morons at the bottom: MBAs with less than high-school physics investing in cleantech, scientific illiterates investing in biotech, etc. As an amplifying factor, you have this negative selection via uniform valuation and terms. The industry thinks that all early stage deals have to be valued at $x M; the quality ideas and inventions are unhappy to be bundled with all the junk and look for boot-straping, and angels, thus increasing the number of bad deals that get funded…As I see how many cool and useful discoveries are being made in Universities and Institutes, not to mention the IQ power in garages and basements, I am absolutely positive that $150B in economic value is a low hurdle for the innovation sector. That’s just 1% of GDP, as a measure. So the limit is not there; we just need better VCs controlling the faucet of where the money should go and how to manage them, post-funding…

    1. Adam Lindemann

      Totally agree with what Krassen wrote.

    2. Mark Essel

      Fantastic way of looking at things, thanks Krassen. I like the way you look at the potential for VC high payoff startups as a fraction of the GDP, my instincts base the fraction size as the total amount of change our economy can allow due to momentum from previous sub-optimal industries, and our economy’s inherent agility.

      1. fredwilson

        This is how my partner brad would look at it too. But he’s not a math geek so he doesn’t have the tools to build the model

    3. fredwilson

      Take that analysis up to boston and give it to the nvca at their annual meeting today!

    4. Stephen Fleming

      From an entrepreneur’s perspective, this is even more reason to focus on bootstrapping and angels… since, if you get caught up in this math, you get stretched on Procrustes’ bed whether you fit or not.More on this in a blog post triggered by the Angel Capital Association meeting a couple of weeks ago:

    5. sling

      Yes,the recent model that has govern VC investment decisons is wrong. It goes something like this: look at the team first- is it an A Team, B Team, etc. Then look at the idea, is it an A Idea, B Idea; then look at ROI, is it an A ROI, B Roi, and so on.Put them itno a grid and choose the oppertunity that has as many As as possible. There are further rules such as : if one potential ivestment is an A Team/B idea, and the other potential investment is a B Team/A Idea, always go with the A Team/ B Idea.If anyone wants to pay me!!!, I will take the time to use game theory and a bit of logic to show that this model is an irrefutable loosing hand over time.

  45. Adam Lindemann

    Fred, a most enjoyable post. It would seem to me based on your analysis, that venture capital companies and the venture capital industry needs to be put through the same rigorous due diligence process as it puts entrepreneurs and their companies through! If we assume that GPs truly are as smart as they say they are to their LPs then they must already know that the business model of their own industry does not work. This leads me to an extremely depressing conclusion, which is that GPs know it does not work and yet have been perpetuating the business because it has been working for them. I think your back to the future scenario should take us back to the time when great entrepreneurs and great venture capitalists had a respect for wealth that had been generated in the past and a sense of responsibility for creating real wealth in the future.

    1. fredwilson

      It took brad and me a whole year to raise usv 2004. Its not the the LPs don’t do the work. Its that its damn hard to figure out who is good and who is not in the venture business

  46. Mark Essel

    Loving the math, very shaky on the business side of the equation.1) is there enough information for good statistics2) does a fixed model make the most sense with the way the industry works?More on 2The way the industry works: I’m “being creative” here so bear with me. Successful or big payoff VC funded startups usually represent large shifts in technology or in industry. Current industry has a certain amount of chronological momentum, and can therefore sustain or allow for fixed gradient magnitude before it collapses. So unless the way our economy works drastically changes to allow for more agility in regards to rising and falling of incredible valuable companies, big payoff VC startups will be capped at a fraction of our total economy.The Model: (did I tell you I love simulations and modeling, especially the unknown ones)I like apmaran’s concept best (consider it pseudo digged), and my preferred model would be chaos theory with local regions of stability where your power law model would fit just fine Fred.Why chaos theory? It fits the limited food supply, and competition models. And there’s even extinction potential.Pick your poison gents, there’s no easy predictors to be had.

    1. Mark Essel

      I suppose entirely new business sectors unrelated to industry momentum could provide a ripe area for VC backed startups, like adding an orthogonal growth direction to the VC fraction of the economy.New sectors: open spectrum for new wifi like companies would be nice, add more based on the imagination of the commenter.

      1. fredwilson

        I love open spectrum, if only we could get some

    2. apmaran

      thx Mark, agree on modeling it regionally as the institutional impact is different and creates minor but not unimportant differences; but I wonder what was the butterfly in this 🙂 In the case of trad IT VC’s where not as important to create the big ones but when Internet hit the masses, then it made a real impact, so “something” is begun outside the sphere and when the super need for something is on, then the boost comes. Strange strings of influence mathematically.What if a new something outside VC focus is on the verge of breakthrough, I suppose that would add 20-25% instant growth to the VC sector if it caught it. Changes the game plan, expectations, results and power, again.(low density fiber plastics are a very underdeveloped something that will change energy and transportation)

    3. fredwilson

      I knew you were going to give me chaos theory when I started reading this comment. That’s way beyond my math skills but I’d love to see that model get built

      1. Mark Essel

        It’s not as far as out as I first thought. Took a great class on it at Stony Brook in ’95. In my foggy memory there were attractors that caused regions of stability, multiple attractors made for some interesting simulations.from (very good intro read or refresher in my case)http://www.abarim-publicati…”Chaos Theory however taught us that nature most often works in patterns, which are caused by the sum of many tiny pulses. “

        1. fredwilson

          Thanks. I’m going to study it since it nicely maps to my work life these days! 🙂

  47. Steven Kane

    as always, thanks for your brave and thoughtful posti will leave the granular math detail to those more qualifiedbut i will humbly submit that using NVCA data for an eercise like this is akin to asking the tobacco lobby for data on the safety of smoking

    1. fredwilson

      Well its pretty damning steve. I don’t think their data helps their cause.They are convening in boston right now. Why don’t you go picket the event?

      1. Steven Kane

        Agreed. So imagine what less biased data would do to the calculation.At the least, as one of the other commenters points out, few if any failedfunds report data to the NVCA, so there is “survivor bias.”I’m not meaning to disrespect the NVCA. Its a trade group, no more and noless. So lobbying/PR is a core mission, which inevitably means happy spin.

  48. bussgang

    Fred – this is a great post and the overwhelming number of comments suggest you have indeed struck a chord. Ironically, I had this exact conversation with a CEO in Silicon Valley today – although, I took the mico perspective and walked through the simple math of the individual $500m fund at 20% average ownership requiring 3x fund performance and thus $7.5 billion of exit value (most likely across 30 portfolio companies, of which only 10-15 will generate the $7.5 billion in returns at an average exit value of $500-750m). Pretty hard if the IPO market doesn’t come back (and I’m not optimistic).Thus, I agree with you that the industry will need to shrink to $15 billion per year (see my post, “VC Rightsizing”, The fund by fund evidence suggests the industry is getting there and that LPs “get it” more than some of us may give them credit for. For example, Atlas went out for $400-500m and closed at $280m. Bain Capital Ventures was rumored to be considering an $800m fund and closed at $400-500m. And Highland is going out with a $400m fund as compared to their last fund of $800m. And these are the strong brands and teams with good history! Many others simply won’t succeed in raising their funds at all (see Dan Primack’s growing “VC zombie” list).At Flybridge, we’ve taken the $50m per general partner approach and concluded that we should stay focused on a single stage (early) and office (Boston) – not allowing ourselves to get bigger than 5 general partners all sitting in one office. We’ll see if other firms that have taken a more aggressive growth approach with staff, capital and offices continue to thrive. We VCs are competitive, type “A”, success-oriented types and it takes tremendous discipline to stay small and focused when you see everyone around you getting big and making money launching new “product lines” (i.e., additional funds)!

    1. fredwilson

      I love the one office, one strategy, less than five partners, less than fifty million per partner modelPut that playbook in the hands of experienced and strong VCs and you’ll make good money over the long run

    2. Mark Essel

      Sounds like Flybridge is “full of win”. Looking forward to hearing about the long term growth of your fund, and best of luck getting first dibs on Boston’s finest startups. Just curious how do you folks manage a presence/connection so that the locals know your group is the best to work with (and don’t go running off with a SV offer)?

      1. bussgang

        Mark – very few early-stage Boston-based start-ups opt for SiliconValley VCs (the bloom is off the “drive by boardmembers” rose) andhonestly, very few Silicon Valley VCs parachute into Boston to scoopearly-stage deals. They have their hands full at home! Like politics,the early-stage business tends to be all local (+/-).——————————Jeffrey J. BussgangFlybridge Capital Partners500 Boylston StreetBoston, MA 02116E: [email protected]: 617 307-9295F: 617 307-9293Blog: http://www.seeingbothsides.comTwitter: http://www.flybridge.com________________________________

        1. Mark Essel

          Hey thanks for the generous education, the more I learn about the business the more fascinating it becomes. I read somewhere (not sure of blog source) that a number of companies move to SV (maybe those are later rounds of funding?)I hope I’m right, but it looks like medical IT is going to get a strong advocate in Aneesh Chopra soon. Not sure how that will shake out with political support to businesses growing in that field, but it could be good news.

  49. Ortal Alpert

    Thanks for a very nice way of looking at things,First, the math, I will skip the integrals part (boring), but if it is exactly a power curve (e.g. A*exp(-(1/b)X) the area under it would be A*bA would be the value of an exit that happens once, and just once per year (arround 5bn I guess)b is the number of exits with roughly 1/3 of that value (how many exits with ~1.5bn value are expected in 2015?, I would bet on 10-30, but I am an optimist).In such a case, the area under the graph (total exits) would be 50-150bn and the VCs would get about 50-60% of that. I am just guessing these numbers here, so just multiply the biggest exit you think will happen in 2015 by the number of exits with 1/3 that value you expect in 2015 (we should be talking 2015, if we consider investments today).A few more pointsThis total value under this graph, grows, if you look back a decade or two, the values we are looking at today for a good startup (not google), would have been unprecedented. This is a result of the last decade investments in startups. Companies like microsoft and google (used to be a very nice exit) are now purchasing other companies and products from other startups.In most big technological companies, between 1-3% are spent on R&D, I think our business is similar, to sustain growth, you have to invest about 1-3% of the GDP on entrepreneurship (this is not all VC money, it’s the internal R&D in companies, it’s the money spent by private people on their neighbors great idea and so on). But it does come up to about 140-500bn in the US alone.So your 15-25bn spent through VCs is a few percent of that “world R&D budget”.We have decided to increase the R&D budget, in order to boost next decade’s economy, which I think is a good move. but perhaps 3X is not reasonable anymore, perhaps we should be cutting costs and look at a model where 2X is enough to make everybody happy.

    1. fredwilson

      Thanks for the math shortcut. I love that. Makes my life a lot easierI think what we need is tiers of capital that add up to that 500bn. Venture capital should be limited to high risk commercial ventures with high potential

  50. Justin Dangel

    Here is a simple partial answer to the problem: cut fees. 1 and 10 instead of 2 and 20 would shave the gross return requirement to 2.75 instead of 3, a meaningful difference. 2 and 20 emerged when funds were very small. A more mature industry with larger funds should accommodate this. A primary reason that the institutional early-stage market is not generating returns is that the intermediation costs are too high. The reality is that VC’s don’t add the value that justifies their fees and should cut. I suspect they’d rather blow up the market then let that happen.

    1. fredwilson

      I’m with you on mgmt fees but not on carryI want founders and management to own way more than 20pcnt of our companies and LPs should want GPs to get more than 10pcnt of the upsideI see your point on larger funds but I don’t think VC funds should be bigger than 250mm ever

    2. fredwilson

      I’m with you on mgmt fees but not on carryI want founders and management to own way more than 20pcnt of our companies and LPs should want GPs to get more than 10pcnt of the upsideI see your point on larger funds but I don’t think VC funds should be bigger than 250mm ever

  51. Fun Dump

    Before you look at the math, look at the definitions. An asset class is a “group of securities that exhibit similar characteristics, behave similarly in the marketplace, and are subject to the same laws and regulations.” Classifying “securities” according to Duration and Payoff we get the following:Name,Duration,Payoff——————–stocks,variable,variablebonds,fixed,fixedlottery,fixed,variableannuity,variable,fixedThus, venture capital is very similar to a lottery or, perhaps, to an activity like prospecting (i.e. fixed duration, variable payoff). If this is true, then is venture capital an appropriate asset class for pensions? Furthermore, shouldn’t one be analyzing the size and value of the opportunity space (e.g. oil reserves, drug targets, computer-mediated social interactions) that one is prospecting in rather than a lump of heterogeneous venture capital funds? Investing more cash than the opportunity space will yield is foolish.

    1. fredwilson

      Excellent point

    2. fredwilson

      Excellent point

    3. Sean Glass

      Great point re: asset class… your comment makes me think of some of the arguments Ben Graham and David Dodd make in Security Analysis.Graham would say that there is no more an apt example of speculation than venture capital. He would term it “Venture Capital Speculation” and would probably quiver at the idea of it being called Venture capital “investment” as there is no safety of principal and no margin of safety :-)…He does say speculation done right (intelligent speculation) can be very profitable, just that it’s much harder than through intelligent investment

      1. fredwilson

        Its only speculation if you have no vision

  52. mike

    Great analysis and quite informative to an individual not involved in the VC industry. However, the comment at the end that you need to keep investment at 15bn a year to keep things in balance is a bit naive.The reason I say this is because if VCs are charging 1-2% carrying costs and fees (or whatever it may be) regardless if you make money or not, what is stopping anyone from raising money? This is why the VC business has grown so large is because everyone, as someone stated below, can make a pitch with Word and PPT. What I feel will happen is that the industry will choose to self regulate itself to protect the existing players, and make it harder for new entrants.I would use the Physical Therapy (PT) industry as an example. A lot of people are drawn to the PT industry because it is quite a high paying job in the medical industry, without the rigours or debt associated with going to medical school. Due to this, there was a huge rush for medical student wannabes into the field in the early to mid 90s. PTs throughout America realized that having more PTs would increase competition, thus lowering their overall salaries, so they forced Universities to enact quotas on entrance in PT school (for example Ohio State only allows 50 or applicants a year) to limit the supply.Thus if the existing players in the VC world would like to keep their fees steady and prevent this “math problem” from seeing the light of day, I would tend to think it is in the VC communities best intierest to regulate how much they wish to raise and how many people can be involved. Now is this thought naive? Probably so, but to protect the community in the long term, it is a logical step.

  53. enoss

    doesn’t this REALLY mess it up? $100b only looks like $18b then what happens? if 2010 goes from $18b to $50b that would be a HUGE increase and still leave things way bad.

  54. SGLee

    You’re right about any given model not scaling – the markets are finite in size introducing limits. There may be an analogy to say hedge funds trying to beat an industrial index. As the size of the fund approaches total market capitalization, it would approach average returns, less fees. Re: Models incorporating power law components – to be sure finance is an interesting area of study, and potential application for this kind of model –An examination of the literature suggests that in an overlapping model class, that associated with the work of Barabasi, Albert, many others, there is a fairly small set of canonical models, plus a number of more or less standard variations. However, the implicit premise may not be sound, namely that a model might exist which fits/approximates and is appropriate for the data. You mentioned considering data which goes back 5, 10, …, years. I wonder whether there may not be degrees of freedom in the problem which would undermine the expected predictive power of any models. For instance, valuations, arguably, depend to a non-negligible extent on the given setting – here the kind of prevailing economy. It may not be entirely certain that we do not find ourselves in an economic setting which differs significantly from the ones of recent years. If so, even given quality data, seemingly apt model, and so on, the model, given a different set of fundamentals, might not possess the inductive power associated with it.

  55. Barbara Duvoisin

    Just a quick note to congratulate you on bringing such an important discussion into the public arena. The idea that VC fund growth has a natural constraint is a gorilla in the corner that most VC’s wouldn’t have the guts to discuss. Although a math geek, I have nothing quantitative to contribute in the face of such greater intelligence (reading the comments is like being back at b-school in the midst of a great brownie-fuelled (or Red Bull-fuelled) tutorial session without the one-upmanship!), but qualitatively must agree with the comments that while VC funds have a natural constraint due to the returns required and the time limitations, entrepreneurship will never have any constraints and should be encouraged at all levels. I’ll try to bone up before the next geek blog entry so I can contribute more quantitatively.

  56. Chris Yeh

    As an asset class, VC has become terribly overinvested. While the dot com bubble was partially to blame, the hidden cause was the massive rush into VC by endowments and pension funds.All the major institutional investors were jealous of David Swensen’s returns at Yale, and concluded that the reason Swensen outperformed was that he invested in illiquid asset classes like VC.As a result, the VC industry ballooned to its $25 billion/year size. It was simply a matter of supply and demand.And at $25 billion/year, VC is simply too big to deliver good returns. True, there are only so many exits, but that too is a symptom and not a cause. VC companies tend to play in the specific fields of technology and biotech. Both those markets are a certain size; simply pumping more money into company formation doesn’t change the fact that the major technology buyers such as Wall Street aren’t going to increase their spending more than 5-10% per year.In the end, all company value comes from the net present value of future cash flows. And cash flows come from customers. And customer spending fundamentally scales with global GDP.When VC was a small asset class, it had plenty of headroom to grow. But the larger it gets, to closer its returns will get to the overall equity return, with the added negative of illiquidity and volatility.

    1. ventureblogalist

      As ballpark references, total money currently managed in VC asset class ~ $250B. Total global software market ~$250B growing at 10% (I think maintenance is 45%). Neither of these numbers are even close to GDP which masks new business dynamics. A decrease in customer spend does not mean there is not significant growth.

  57. Chris Douvos

    Fred:Good stuff and good comments. I wrote last year about the problem of generating “venture returns” at the large fund level ( and you’ve aggregated the problem up nicely. It’s a big fallacy of composition (combined with a Lake Wobegon problem): the LPs all complain that there are too many people in the asset class, but no one wants to get out because everyone thinks their managers are “better than average.”One thing I worry about (and this is a longer discussion) is the role of luck and timing in results. Said another way, when you run the regression of what makes a good VC, for most, the error term/unexplained variance is really large. We could stipulate that USV is da bomb, but every now and again someone else will get lucky or you’ll get unlucky and that will muss up results. The law of small numbers drives some funky outcomes. Because the evaluation horizon exceeds people’s attention spans and things are so opaque, you get can be rewarded for being lucky, but not right or punished for being right but not lucky.That’s what keeps people coming back; they’re just playing a Dollar and a Dream ( good,CD

  58. Geoff Morrissett

    Using round numbers from “normal” market conditions in the historical data that I have, the gross annualized return of venture capital (VC) is 17% – 22% (agreeing well with NVCA’s 10 year and 20 year data in the above article’s table) while the net (of 2.0% – 2.5% in management fees and 20% in carried interest) annualized return of VC is 14% – 19%, which tries to cover the limited partners’ (LPs’) expected annualized return (or weighted average cost of capital (WACC) to VCs) of 15% – 16%. I assume that VCs own 70% – 75% of their portfolio companies at the time of exit and that the average investment holding period is 5 – 7 years. A $25 billion investment today should yield a gross value of $73 billion – $144 billion (gross value multiple of 2.9 – 5.8) or a net value of $64 billion – $121 billion (net value multiple of 2.6 – 4.8) in 5 – 7 years.Historically, 9% – 15% of VC investment targets yield those returns while 23% – 28% of VC investment targets yield those returns or better. I believe that NVCA data suggests there are 700 VCs with an average fund size of $150 million and 3 general partners (GPs), which make 1 – 2 investments (think targets) per year during the first 5 years of a 10 year fund. Assuming each VC firm concurrently manages 2 funds and 3 – 5 VCs participate in any given investment target, then there should be 840 – 2,800 unique VC investment targets per year. Therefore, 193 – 784 unique investment targets should yield the above returns per year upon exit.Sensitivity analysis can be performed to adjust the above assumptions in order to understand changes in market conditions (i.e., lower annualized returns, longer investment holding periods, etc. in today’s market that might better align with NVCA’s 1 year, 3 year, and 5 year data in the above article’s table) and/or reflect one’s experience, knowledge, performance objectives, etc.

    1. fredwilson

      This is very specific dataWhere do you get this from?

      1. Geoff Morrissett

        The data was gathered and analyzed by The Wharton School, University of Pennsylvania. Granted, the mix between IPO (minimal) and acquisition (available) exit types, exit multiple values (lower), and investment holding periods (longer) have all changed significantly today from the respective “long-term” averages. Is this a business cycle or permanent change? No one knows with certainty. Hopefully, all of the above provides context (at a minimum) for the VC math problem.

        1. fredwilson

          Is that work available for review?

  59. jire

    Fred, you’ve actually proven a different point… that is that a 2.5x net is an unrealistic number for a large asset class… the net IRR on that would be something like 18-24% depending upon the holding period… why does the bar have to be set that high? public investments have yielded essentially 0% for the last 10 years… a 10% compound annual rate of growth with the ability to defer capital gains for many years is not so bad, is it?… the asset class works just fine, and would be perceived as fine if venture capitalists stopped complaining that they suck

    1. Marc Step

      Larry Kabul of Labrador Ventures has given a couple of presentations on the VC model in the last few years:… (2007)… (2008)Both speak to the issues of VC with similar conclusionsAlso, note that returns for seed VC at ten years is still pretty good (your table). Goes to the point that more smaller deals, at the right valuation, helps returns. (Wonder how Angels are doing —…But regardless of math (by the time anyone figures it out, all will have changed), prudence just dictates that there is too much money chasing poor deals. And many of those deals are probably “run fast” type deals. Be interesting to see some comparison of funds that invest in technology with real IP (e.g., medical devices) as compared to those that are only software based where competition is easily inspired.

    2. fredwilson

      The return should be that high or more because you are asking someone to lock up their money for 10 years and in all likelihood more like 12-15 years. I know they won’t be out all that money for the entire time because of the way capital is called and distributed.But this is a highly illiquid asset class and the returns should justify that illiquidity

  60. elliottdahan

    I agree that there are too many traditional VC firms chasing too few exits. I believe Limited Partners and market forces will whittle down the number of VC firms.But – I think that the Risk Investing industry needs to be viewed in a different way.The Risk Investing industry is divided into 3 systemically, operationally and attitudinally different components: (1) Seed Investing; (2) Traditional VC Investing and (3) Exit Risk.The criteria and expectations for Seed Investing are different than for VC Investing.The Funding Criteria for Seed investing is “Belief” – in the guys in front of you, can they really build this great new thing – in the marketplace, will it really buy as much and pay as much as you projected . . . . .The Funding Criteria for VC Investing is “Traction” – validation from potential end users of the technology / service / product – how many ? how quickly ? competitive reaction ? . . . . .Follow on investments for the Seed Stage come from VCs and corporate investors. Follow on investments for VCs come from other VCs, corporate investors and M&A.The Geographic reach for Seed is usually local/1 day trip and the geographic reach for VCs is regional / national and even international.The most cost and labor intensive elements of Seed investment are the sourcing, screening and post investment oversight of the Seed company.A traditional VC firm cannot source, screen and provide post investment oversight enough Seed investments to justify the number of Seed investments needed for a profitable fund based on the mortality rate of Seed companies.What is needed is a Public-Private For Profit dedicated effort to work with, support and compensate the Seed Infrastructure (Incubators, Economic Development Agencies, Tech Transfers). This infrastructure, the public component, already exists and provides the efficient sourcing, screening and post-investment oversight needed to develop Series A worthy companies. What is needed is a dedicated effort that is not geographically constrained. What is needed is a thorough Virtual Incubation system that brings both Community and Collaboration to all elements of the total Investing community.The Venture Capital stage of the Venture Risk Investing industry has a valuable place – to expand Seed/Startup companies with money, targeted managerial talent and business development/partnership assistance.But, simply shrinking the size of a VC fund will not work for all of the reason presented above.By dedicating a private/public collaboration to increasing the value and viability of early stage companies you are also increasing their valuation for their Series A round; thereby leveling the playing field with what will be a smaller group of Traditional VC funds.This Seed dedicated effort can take two forms:(1)Standalone Fund(2)Operating Division of a Traditional VC FirmPlease review the powerpoint – The START Fund –…I look forward to all comments.Thank you,Elliott DahanManaging PartnerThe Growth Group652 Cuesta DriveMountain View, CA 94040Email elliott(a)thegrowthgroup.comPhone 650 903 9990URL

  61. valto

    I think this just shows that money does not solve problems. The problem here is not too much money, but how it’s now distributed.VC industry is starting to look like the newspaper industry 😉 – better to wake up, the blogs are coming…More at:

    1. fredwilson

      What are the blogs of the VC business?Things like Y Combinator are great but they are feeding us even more opportunities so I see them as additiveAlthough I also see blogs as additive for the newspaper business if they’d just see themselves as curators and aggregators instead of content creators

      1. valto

        I guess the closest thing to “blogs of VC industry” today are angel investors. But that’s for today. Also Y-combinator and the likes are great too and could be considered as “blogs of VC industry”. However their “next step” need to change away from just VC’s.The fundamental change will become, when there are “platforms” for anyone to start a “blog for VC industry” and that’s what we are doing in, we feel that in long term the money will be spread to more potential startups and more of them will not go via IPO but just buy back of shares, mergers etc. with lower ROI. But that’s OK if the time for ROI is shorter and cost of management is lower.So – be more direct, spread wider, lower the management cost, speed up the ROI cycle and you can accept lower ROI.If you think about the structure of today, from where the VC money really comes from, you start to see the “big picture”. – basically it means that individuals like you and me pay for pensions funds etc. and these funds then invest to VC funds. VC’s then make investment decisions and “manage” the investments, all the way to take it public (hopefully). Basically just to sell it back to us…When more people will start to understand this cycle because of more info and transparency online (if they are interested), people will not accept this structure. Because in the long run what matter is, if the companies in question sell what matters. And that is not a question of size.

        1. fredwilson

          Got itI hope this worksIt would scale much better

  62. mtalaba

    This point is related to performance, and is probably a whole other subject for discussion, but the deeper question about improving VC performance and return have to do with the how and why of selecting the right entrepreneurial team. We all know that the startup world is littered with the bones of great ideas and products that fizzled. Some of these failures were nobody’s fault, but most of the disappointments have to do with the team–not just the individuals on the team mind you, but rather, the chemistry and ‘teaming characteristics’ of the team. Certainly there is value in aggregating all this activity mathematically to make sense of what has happened in the past, but I suggest that the real ‘treasure’ and game-changing factor would come from a better way to reliably determine whether or not the ‘custodians’ of the VC funding are really a good bet. If this is of interest to anyone, check out Role-Based Assessment, which was created by The Gabriel Institute in Philadelphia.

  63. Jeremy Horn

    Awesome analysis and comments. An interesting extension of this analysis would be to see how other methods of fundraising are taking up the slack resulting from the mathematical constraints being experienced within the VC community. Are methods like Angels, F&F, etc., able to provide the scalability to the investor-entrepreneur ecosystem as a whole, are they already providing the flexibility (how much flexibility are the adding, what sort of ceiling can we project on those components, how much flexibility and growth potential do these non-VC sources enable — marginal? orders of magnitude?).Fred, great conversation you got rolling here. Recommending this article to my readers for their Weekend Reading……Jeremy HornThe Product Guy

  64. Andrew Sispoidis

    I’m not approaching this from a mathematical perspective but rather from a qualitative one. The class as a whole might be better served to base it’s performance on volume of transactions rather than exit size. I firmly believe that smaller deals with smaller, quicker exits are a more sustainable mechanism. I also think that alternative methods for realizing returns from portfolio companies is an idea worth investigating, some form of either convertible debt or dividend payouts might be a good start. Additionally, I think that a more tight-knit collaborative relationship between venture funds and portfolio companies where some operational support could be facilitated across multiple ventures would be a great way to reduce the risk and the size the investments. Finally I think that structuring deals where the capital flows to portfolio companies on a annual, biannual, or quarterly basis rather than in large truanches might be a way to ameliorate the IRR issue.

  65. sigmaalgebra

    First, I would advise you and your limited partner (LP) that looking at the ‘asset class’ is likely drawing the wrong lesson from the H. Markowitz and W. Sharpe work on portfolio management where a central feature of the work was the assumption of very limited input data maybe roughly appropriate for investing in stocks in the NYSE.Or, considering the ‘asset class’ could be important if your LP were picking venture firms just by throwing a dart at a full list of venture firms. That you met with your LP, explained what you are doing, and heard that he is “happy with the job” you are doing is a huge split from the Markowitz-Sharpe assumptions. That is, your LP has, is considering, and is “happy with” much more data then just looking at the ‘asset class’. Since he is one of your LPs, it is important for him to care about Fred Wilson and Union Square, and, appropriately, he did meet with you, etc. Once he is “happy with” Union Square, really he shouldn’t much care about what happens to the venture capital ‘asset class’ and instead can be happy all the way to the bank, whatever the asset class is doing. Or, if your LP were just throwing darts, then why did his boss have him spending time meeting with you? Indeed, why is his boss paying him at all?!For some math, your LP is estimating the ‘conditional expectation’ of returns given that the venture firm is Union Square. Or, if real valued random variable Y is the return and X is the venture firm, then your LP is estimating E[Y|X = “Union Square”]. Or, to do this, take the joint distribution of Y and X, at the point on the X axis where X = “Union Square”, extract the data from the joint distribution there (SQL query: SELECT Y FROM Joint_Distribution WHERE X = “Union Square”), scale that data to a distribution, and take the expectation. Yes, it is true that E[Y] = E[E[Y|X]], but, still, E[Y|X = “Union Square”] is free to be very different from E[Y].Yes, that E[Y] = E[E[Y|X]] follows from the Radon-Nikodym theorem, with a famous proof by J. von Neumann, long about 30 miles south-west of Union Square!Here is a little on how your LP might apply the Markowitz thinking: Get the Markowitz expectation and covariance data just within Union Square and solve the corresponding quadratic optimization problem to maximize the return. For how to apply the Sharpe work, that would be more complicated and using assumptions out in the ozone. Again, though, you would be ignoring crucial information beyond what Markowitz was assuming.Or, before we go across the street, do we look at the odds from all pedestrians or do we just look right in front of us at the traffic, the width of the street, the weather, how well we can see, etc.? Or, if we can see our way clear to cross the street, and get something nice on the other side, quickly, safely, then we go for it, and averages from much different circumstances do not concern us. This analogy should be close for all of entrepreneurs, venture partners, limited partners, and exit buyers.So, the key for one venture partner to doing better than Markowitz and Sharpe is additional information. I would also recommend that additional information is also the key to improving the ‘asset class’.In particular, here is a neglected source of such information: If we are to use computers to convert data into valuable information that we deliver over the Internet, and broadly this is what we want to do, then one question often relevant is how to do the processing to get especially valuable information? Whatever processing we do will necessarily be mathematically something, understood or not, powerful or not. For more powerful processing (so that we can want to cross the street) well understood (so that we can see our way clear to cross the street) we should proceed how?Hmm …. We are not the first to ask this question. Physics asked it and concluded “Mathematics is the part of physics where experiments are cheap.” by V. I. Arnold (as in the Kolmogorov–Arnold–Moser theorem, right, J. Moser long at Courant not far from Union Square) as inhttp://pauli.uni-muenster.d…The US DoD drew similar conclusions, and US national security has been greatly helped. E.g., what role is there for the fast Fourier transform (Cooley, about 40 miles north of Union Square and Tukey, again, maybe 30 miles south-west) in passive sonar? Given a dense metal we want to compress a lot, quickly, with a shaped charge, what shapes do we want for the metal and the charge, i.e., more efficient than spherical? Credit has gone to J. von Neumann, again, about 30 miles south-west. There was much more in spread spectrum radar from algebraic coding theory, shapes for stealth, Wiener filtering, target detection with the Neyman-Pearson result, design of turbojet engines, and on and on.As DoD problem sponsors and proposal reviewers know, mathematical solutions commonly use a unique technique to see the way clear across at least part of the street — that crown jewel of civilization, the highest quality knowledge we have, mathematical proof.So, right: For more powerful processing to get more valuable information, we should proceed mathematically. Mathematical proof helps us see our way clear “where experiments are cheap”.Track record? On average? Hmm …. Or, did Bill Gates do this? Warren Buffett? Steve Jobs? Nope! But, if see way clear across the street to something nice on the other side, then go for it anyway! Or, after all guys, we’re supposed to be looking for things that are new yet we want to know that they will work, right?For the math in the blog entry, one issue is ‘stages’: Of course, we are now necessarily close, in both wording and meaning, with R. Bellman’s work as in, say,Stuart E. Dreyfus and Averill M. Law, ‘The Art and Theory of Dynamic Programming’.So, first-cut, we should consider the usual stages, seed, A, B, C, and exit. To continue we should have some data on investments and ‘transitions’ at the various stages. One side result might be an adjustment in the investment amounts at the various stages. Bellman’s work provides a well organized way to do this. Or that work is the center of best decision making over time under uncertainty and, yes, is close to Black-Scholes, exploiting the term-structure of interest rates, etc.From ‘power law’ considerations you are looking for the distribution of returns and, it appears, then using that distribution to estimate the expectation of returns. But (1) the ‘power law’ assumption is shaky; (2) even if a power law is appropriate, the right power law exponent is not known; and (3) you do have data on about 1000 exits a year to work from. So, to estimate the expectation of return, exploit the ‘law of large numbers’ and just take the ordinary average of the data you do have on the 1000 or so exits.Does a distribution exist? In a well defined problem, of course it does. Can you find the distribution in fine detail? In practical problems, sometimes, but often, no. Does this mean we can’t work with probability? Usually, no. That a distribution exists does not mean that always we should find it as our next step in practice! Yes, rushing to find the distribution is sometimes implicit in some teaching; not mine!Here is a case where just qualitative assumptions (“Look, Ma; no data!”) let us know nearly everything about a distribution: Suppose things are arriving, one at a time, and we count them. Suppose the number of arrivals in an interval is independent of the arrivals before the interval (‘independent increments’). Suppose the distribution of the number of arrivals in an interval does not depend on when the interval starts (‘stationary increments’). Then the arrivals form a ‘Poisson process’, and the time between arrivals has the exponential distribution with one free parameter, the arrival ‘rate’. A simple ratio of the number of arrivals in an interval of time over the length of the interval is an unbiased (that is, a good) estimator of this arrival rate. Details are at the beginning of the chapter on Poisson processes inE. Çinlar, ‘Introduction to Stochastic Processes’,again, 30 miles south west!Moreover, if arrivals are coming from many sources, the sources are acting independently, and at each of the sources there is a mild assumption about the distribution of time between arrivals, then as the number of sources increases the arrival process converges to a Poisson process. This is the ‘renewal’ theorem with a proof in W. Feller’s second volume, again, from about 30 miles …! So, first cut, the sources might be entrepreneurs and the arrivals, business plans as good as Google!Why the ordinary average? Because it has terrific ‘statistical’ properties. The law of large numbers says that (under meager assumptions that likely hold in this case, power law or not) as the number of data points increases, the average converges to the expectation with probability 1 (that’s the gold standard on convergence!). Next, the estimates are always ‘unbiased’ which means that if you take, say, 500 data points, get an estimate, do this often, and get the expectation of the estimates, you again get the expectation you want. Next, there is a famousPaul R. Halmos, “The Theory of Unbiased Estimation”, ‘The Annals of Mathematical Statistics’, Volume 17, Number 1, March, 1946, pages 34-43.(wonders of the Internet, start at…) that argues, under reasonable assumptions, the ordinary average has the least variance (is the most accurate in the most accepted sense) among all unbiased estimators. In particular, then, Halmos showed that using the data to estimate a distribution and then using the estimated distribution to estimate the expectation will not be a better estimator than just averaging the data! All without any data at all! Ah, the power of mathematical proof — should be a way to make money with such a powerful approach to information!Note: Halmos is a good candidate for the best writer of mathematics in the 20th century. He was a student of J. Doob (as in martingale theory) at University of Illinois and was an assistant to von Neumann about 30 miles …!Martingale theory: Every stochastic process is the sum of a martingale (essentially a fair gamble) and a predictable process. Hmm …! Every martingale either runs off to infinity or converges to a point — it can’t just keep wandering around near home.If Union Square and its LP want higher returns, for Union Square and the asset class, then I suggest exploiting mathematics to find how to get more valuable information from available data and suspecting that for the rest of this century such mathematics will be more powerful than Moore’s law will be and has been. Yes, knowing how to program is important but usually routine. Knowing what to program is crucial but not a part of programming! The mathematics has been neglected. Sooooo, there’s an opportunity here, including for the whole ‘asset class’. That’s some of what your LP wanted to hear, right?

    1. Longtime Entrepreneur

      You folks demonstrate why true entrepreneurs disdain you. Where is the discussion about building a great company, products that change things?

      1. sigmaalgebra

        For your:”You folks demonstrate why true entrepreneurs disdain you.”I AM an entrepreneur.I don’t “disdain” Fred’s discussion: He is responding to a concern of his limited partner (LP) that the venture capital ‘asset class’ is not doing well. For the LP, and, then, for Fred, that can be a serious concern. In particular, the LP and his organization likely are pushed hard to follow the thinking of Markowitz and Sharpe on ‘asset allocation’. Markowitz got a Nobel prize in economics for his work, and later so did Sharpe. This material is now rock solid in ‘finance’ for ‘money managers’, especially ones close to large state pension funds and university endowments who have a tough time going against two Nobel prize winners.I outlined for Fred how he can tell his LP why it is good that he and Union Square are doing well for the LP even if the asset class is not.My post was long, but with more length I could outline some more that Fred might tell his LP or what the LP might tell his organization about investing with Fred; that would be more length.For your:”Where is the discussion about building a great company, products that change things?”I DID emphasize that for Fred to look at additional information, for just this issue, is what is crucial and, especially, what separates Fred’s work from the relatively simplistic considerations of Markowitz and Sharpe. Much of the additional information is what is available to Fred from his entrepreneurs about “product” and “building a great company”.And I DID discuss one key to those things, “a neglected source” of information to “change things”, a source more powerful than Moore’s law, that can be the key to higher returns “for all of entrepreneurs, venture partners, limited partners, and exit buyers” and for the asset class. Don’t have to believe me; just look at what the US DoD has done with applied mathematics.Just read, slowly, carefully, word by word, and look up on the Internet words that need more explanation such as the work of Markowitz and Sharpe, the Radon-Nikodym theorem, the Halmos paper, and the list of examples I gave of how mathematics has helped US national security.For very specific examples of how mathematics can be the key to the powerful ‘secret sauce’ for more valuable information for a great company, I should not explain those here, but there are venture partners enough in the US who have my most recent thinking.

      2. fredwilson

        we have those discussions too herei’d suggest you go look at the ten most recent posts and i think you’ll find plenty of that

    2. fredwilson

      lots of math in here that i can only try to comprehend, but i do appreciate the commenti like your assertion that math is the new moore’s lawwe’ve been hiring a lot of mathematicians into our companies recently and this gives me the confidence to push for even more of thati also think semantics is a big part of the future of the web

      1. Sean Glass

        Great comment. I think that in addition to better information, the way that top tier VC’s separate themselves from 2-4th quartile VC’s is through better relationships. If you consider the idea that the value of a network grows at an exponential rate (the size squared), then the VC’s that build a large network of close relationships will build a proportionately large differentiator versus other funds. This is why great VC’s keep up relationships with so many different people (entrepreneurs who have succeeded, failed, are too early, that they missed out on investing in, etc…)

        1. fredwilson

          Yup. Think of this blog in that context

  66. Robert Levin

    I just finished reading an interesting paper on the benefit of Private Equity to the value of the exit.The abstract, “The findings indicate that PE-backed firms generally have higher earnings quality than those that do not have PE sponsorship, engage less in earnings management, and report more conservatively both before and after the IPO. Further, PE-backed firms that are majority-owned by PE sponsors exhibit superior long-term stock price performance after they go public.”Not sure how this will be directly useful in your analysis of returns, but it bolsters a hypothesis I’ve operated under for years. Companies that get the benefit of professional, experienced investors gain far more than the ones who are just owner managed or self financed. While I admire anyone who can build and grow a company without outside capital, when it comes time to realize an exit, they tend to play catch up with those who had outside investors earlier.You can get the whole paper here….

  67. GarySilverWDX

    There are still unfunded quality deals that VC’s are not sophisticated enough in those relevant technologies or that vertical marketplace to understand, then too much money chases only the more obvious and trendy opportunities. This is NOT a slam against VC’s as a group, but against the limitations of any outsider (relative to the entrepreneur) of knowledge, understanding, and crystal ball vision of the outcome in taking all sorts of risks (on technology, market acceptance, team, etc.)The question is whether this means the VC model is broken? Perhaps it is semantics to say it is or isn’t. Can you say it is the model that is broken when the limitations are a function of the fundamental limitations and knowledge of those who own and run the process, and perhaps not a fundamental flaw of the basic structure of the process itself? That is not to say that the process does not have room for improvement, however the VC model may just unfortunately be (in its basic form) the best humans can do. The VC model does not scale beyond those opportunities that are easily understood by industry/technology outsiders (investors), or opportunities caught in a popular trend.So the model may not be broken, but just inherently limited by its human inputs, and “broken” is just an argument of semantics.On a separate front, if the math is right, then the model is out of equilibrium and investment will shrink or IPO/M&A prospects will need to get a lot brighter in order to re-balance. And while as an entrepreneur it pains me to suggest that, my hope is for the pickup in M&A. But out of balance is not the same as broken.

  68. Sven

    It’s all about valuation of the last money in. In Q4-2007 the 12 months average post money valuation for a start up was $115m but the average Exit valuation was $75-80m the math is not working (except in Baseball and Las Vegas).The VCs have to learn to set valuation right and to manage the portfolio towards average Exit valuations and not towards home runs.

    1. fredwilson

      The average valuation was 115mm?That cannot beWe do about 20 deals a year between initial and follow onsAnd maybe one of them, a late stage round, will be at that valuationThere’s something wrong with that number

  69. jeremylichtman

    No industry can scale indefinitely.That’s what Warren Buffett has been saying for a number of years in his annual reports – that Berkshire is too big to get the returns that they used to get, since there just aren’t that many large deals he can invest in (and the smaller ones, while profitable, aren’t big enough to give a high percentage rate return overall).Now that said, there’s still a lot of room for small VCs to make very good money investing in startups.

    1. fredwilson

      I agree

  70. David Mas

    You’re right, the VC class is not scaling. I ‘m not sure you need to resort to power laws to make your points, aggregate figures are already very convincing. Whether the 100$bn are coming from a single exit or from ten thousand, it doesn’t matter to much for the investors, as long as they get their returns :)This “math problem” has been investigated by Gompers & Lerner (1998): Money Chasing Deals?: The Impact of Fund Inflows on Private Equity Valuations. (…They show with a simple market model that the good deals are a scarce ressource and that too much investment in the asset class leads to les interesting deals being financed and thus to lesser returns.That’s a big point about VC : you don’t only need money, you also need entrepreneurs and innovations. As all market, the VC financing market is there to adjust offer and demand. Diminishing returns means too much money invested and should lead to less investment.The only problem of this theoretically beautifull mechanism is the time lag. Its takes years to get the returns from a VC investment. This is likely to lead to over-reaction and big cycle in VC financing volume. And this is when you don’t take into account external events, like the actual financial crisis….Hope this helps.David

  71. Ed Burke

    “I think “back to the future” is the answer to most of the venture capital asset class problems. Less capital in the asset class, smaller fund sizes, smaller partnerships, smaller deals, and smaller exits. The math works as long as you don’t put too many zeros on the end of the numbers you are working with.”Well said. Enjoyed the article and blog comments. Even on the other side of the deal, many entrepeneurs do not want or need the amount of funding required by the venture firms to “make the math work”. In this industry, most of the time, less is more.

  72. Name

    i hvve no idea mate but whatever you say sounds really cool – like the stuff they have on TV in the business channels – no one gets it – but everyone is talking it!

  73. Name

    That really is such a great comment. Unfortunately I don’t think VCs will ever move to a model with less fees and more carry because deep down they know its almost impossible to make money from carry in this day and age. Let’s look at the math…Let’s say an early stage fund manages $250 million. To even *return* the fund, assuming 10% ownership at exit (most firms would be lucky to have this, although they own around 15-20% when they make their initial investment, by the time the company is ready to exit for any decent sized exit the number is usually far smaller) they need total exits of $2.5 billion.Remember that’s just to *return* the fund. To make any real money, as fred says, they need to return 3x that amount. That’s $7.5 billion in exits. For that one small $250 million fund.Assuming 100 companies in your portfolio you’re talking about each and every one exiting for an average of $75 million to get that kind of return. Put another way you’d need over 7 companies in your portfolio to exit for over $1 billion.Now I actually happen to think Fred is one of (is not the) smartest venture investor in this day and age, who gets more than most the realities of venture investing. His savvy investments means he may actually be able to pull this off.But how many other firms can say the same thing? I’d argue you can count them on one hand, and all the others have been around for a while and like Fred have a history of successful exits under their belt.LPs need to wake up and stop throwing money away into new and under-performing funds that don’t already have a history of great returns.

    1. fredwilson

      I think there are at least a dozen other firms out there that can pull it off. And most of them manage funds less than 250mm

  74. Name

    That really is such a great comment. Unfortunately I don’t think VCs will ever move to a model with less fees and more carry because deep down they know its almost impossible to make money from carry in this day and age. Let’s look at the math…Let’s say an early stage fund manages $250 million. To even *return* the fund, assuming 10% ownership at exit (most firms would be lucky to have this, although they own around 15-20% when they make their initial investment, by the time the company is ready to exit for any decent sized exit the number is usually far smaller) they need total exits of $2.5 billion.Remember that’s just to *return* the fund. To make any real money, as fred says, they need to return 3x that amount. That’s $7.5 billion in exits. For that one small $250 million fund.Assuming 100 companies in your portfolio you’re talking about each and every one exiting for an average of $75 million to get that kind of return. Put another way you’d need over 7 companies in your portfolio to exit for over $1 billion.Now I actually happen to think Fred is one of (is not the) smartest venture investor in this day and age, who gets more than most the realities of venture investing. His savvy investments means he may actually be able to pull this off.But how many other firms can say the same thing? I’d argue you can count them on one hand, and all the others have been around for a while and like Fred have a history of successful exits under their belt.LPs need to wake up and stop throwing money away into new and under-performing funds that don’t already have a history of great returns.

  75. Guest

    This is the correct perspective to look at VC, and unfortunately very few institutions have despite many of us screaming for over ten years. We took the ‘if you can’t beat them join them’ approach with Initium VC, but we failed to raise our $250 million fund in 2002, despite a very strong team with far better track record than all but a few, and for example I found Google and many others long before those who invested.However, I would like to point out one macro issue that most are missing in this debate, which is global. Not only has VC been institutionalized, complete with the foolishly applied asset allocation model, but most of the institutions are as biased as individuals. For example one chairman of a global tech giant, myself, and one of the premier VC partners who ever lived joked in an email a couple of years ago that most venture firms are really investing in real estate– that is, most LPs have far more invested in real estate where the firms are located than they do in VC. It isn’t transparent in many cases, and often denied, but there is no question at all that the majority of LPs are activist and strategic, so it’s not even close to being a free market.Now consider the changes globally in the past decade. Twenty years ago the U.S. had a near monopoly on the entire seed to exit ecosystem in technology. In the past decade venture funds have emerged all over the world at the same time that local research has improved, entrepreneurial cultures have improved, and IPO markets in many cases have surpassed the U.S.So in much the same way as the housing bubble and investment in general, during the historic events of the rise of Asia and the creation of the EU- complete with seed to exit ecosystems in ventures that would otherwise in previous decades have likely emerged in the U.S., when investment in the U.S. should have been declining in many asset classes, instead we formed bubbles.The tragic aspect of all this has been that the over investment in the asset class acclimatized customers with unsustainable subsidies, creating the largest price war in history– particularly on the Web– but in many areas, so it has become far more difficult for entrepreneurs to grow organically as partners with customers who are willing to pay for products and services rendered– which has historically been an essential part of sustainable global economics. Of course that’s the point of price wars– way too much money in the wrong hands who have interests that are misaligned towards the broader economy– that’s the macro problem we’ve seen on Wall Street and in the valley.Another related issue is the competency of so many VC partners and analysts– what a difference between the 80s and 90s — in the 80s almost all were deeply experienced in every aspect of building companies to include finance. By the late 90s we had hundreds who had never built a real business, but rather built false companies and profited from financial engineering (and I’m being kind here).We will obviously continue to see a correction, but particularly in the U.S. we’ve seen a perm change (short of yet another devastating bubble- doubt we could survive in current form.) We’ve got to return to product engineering and company engineering– moving far away from financial engineering– otherwise our culture will continue to move towards fraud and away from fundamental value creation, which is what built this country.Thanks for the work- MM

    1. fredwilson

      Great point about the global oppty. If we are to expand again, it had better be internationally

  76. Carlos T

    Regarding your comment: “….each VC investor owns, on average 20% of each portfolio company”. Then later you mention a higher percentage, “…..If you use my 3x gross and on average 50% ownership by VCs”,My question is, which is it? What percentage of company does a VC own in exchange for capital? 15%? 20% 50%? I look forward to a reply.

    1. fredwilson

      20% is the assumed amount each VC owns50% is the assumed amount all the VCs together ownusually there is a later stage investor who invests for less than 20%i’m assuming two early stage VCs at 20% each and one late stage at 10%

  77. Sean Glass

    Fred,This is a great post. I’ve been noodling with this same problem myself recently. I think that your basic assumptions are correct. Part of what I’ve spent time thinking about is that there are artificial limitations that reduce returns for funds as well. For instance, if you invest in a company later in your fund cycle, you may be forced to look for an exit before the ideal time. This can lead to forcing the company into a recap or liquidity event before full value would be realized. This can also lower the ultimate return. Also, there is certainly a reinvestment risk. If you have a company that you get an exit from 4 years into the fund, you will reduce further capital calls perhaps, but the full amount of the fund is ultimately not being invested – this can lead to the opposite issue of waiting too long to move to an exit.Azeem, your comments are spot on. It definitely seems like there is a big distinction between funds where the partnership is dedicated to building businesses and knows it will build wealth through the carry, and funds which have been raised for partners to live off of the management fee. This is certainly less of an issue with small early stage funds (2% of a 10-$40MM fund is a good amount of money, but not a HUGE amount). Your idea for top tier funds to reduce the management fee makes a lot of sense. One top tier fund I know in SV actually insists that at least 10% of the money in the fund comes from the partners themselves and that the partner sponsoring a deal participates individually as well. I think this certainly creates more alignment between the GP’s and LP’s.Top LP’s (think University endowments such as Yale which led the way into the VC and PE asset classes) won’t invest in first time funds. Period, end of story.One of the great separators in Venture Capital (at least according to the top VC’s I know) is being able to pick out top tier teams. As there aren’t great quantitative screens for human capital (What is a tier A team vs. tier B team when you’re considering investing in 2 CS students in a garage at Stanford or a Harvard dropout named Bill or Mark?), great VC’s tend to have both market foresight and the ability to pick great people to back. Once someone has shown evidence of this (i.e. through successful angel investment or a first fund), then the I would posit the likelihood of future success is higher.I still have to wonder if pure luck has some role ultimately in if a particular fund is a success or not. Since a single fund doesn’t invest in a smooth curve of companies, but rather has a very small sample, there is some chance that even a very well managed fund may miss out on the one “home run”. Do this – build an overly simple model – single, double, triple, home run and strike out (to use an American baseball analogy). If you want to be fancier, you can use the power distribution for returns to give you an idea of how many exits you would have at each point (single = 1x double = 3x triple = 10x home run =40x). Then calculate the IRR net of fees and carry. Now what happens if you have 1 less home run? With 5% homeruns, 5% triples, 15% doubles, 15% triples, and 60% outs with 5% of the fund invested per deal, your IRR works out at just over 14%. Not bad, but not great. Now let’s say you don’t have that one 40x deal, but rather it’s a ten bagger. If nothing else changes, the IRR drops to just over 4%. Now you can smooth this out by investing less than 5% per deal, and certainly you do better by doubling down in companies that are raising money and are looking successful and cutting out companies that you see are going down, but fundamentally my point is that if one or two exits make the fund, then perhaps timing, luck etc could play a larger role than we might like to admit? I’d be interested in hearing Fred and other current VC’s opinion on this, and also how this may be more or less applicable depending on the stage of the fund, size of investments, etc.

  78. fredwilson

    that’s a great question. my initial instinct is that “institutional venture capital” doesn’t scale but entrepreneurship does. i suspect other funding sources, like angels, friends and families, small business loans, etc can provide a lot more capital and need a lot less aggregate return.

  79. Tyler

    An important distinction should be made among the type of entrepreneurs and small businesses. Most entrepreneurs and small business, particularly the one talked about by “public official”, are replicative entrepreneurs and SB’s. E.g- dry cleaners, restaurant, home builder, life insurance sales etc. VC’s invest in innovative entrepreneurs- people and sb’s that are trying to fundamentally change/ improve the way things are done. Obviously the upside is far greater.Under that assumption,it is okay to encourage entrepreneurship even if VC investment in “innovative startups” will ultimately decrease. Small business growth is very positive for the overall picture of the economic environment, if only for the simple fact that it keeps America’s culture of entreprenuership alive, thus luring the promising young innovators in that path.

  80. fredwilson

    none. crunchbase is the best and its free

  81. Jeff Clavier

    I was about to make the same comment.The issue is the business model of venture capital, and the fact that there is not enough exits/liquidity out there to sustain the level of investment in the asset class.Entrepreneurship does scale, if the goal is to build a sustainable business from day one. I meet an increasing amount of companies that start generating modest revenues from day one, enough to sustain an operation running with reduced salaries and a shoestring budget. As these revenues scale, so does their ability to invest in their business that can then throw some or a lot of cashflows.The big problem we have as VCs is that we need the companies we invest in to reach some measure of scale in a limited time (the lifetime of a fund) so we can bag our multiple through a liquidity we can distribute to our own investors. And there is no way for us to force that, especially in this environment where IPOs, trade sales, and buyouts are not really an option.

  82. Eric Olson

    > I don’t have a conclusion, I’m just saying it is interesting to think about and ponder the root causes of limited VC industry scalability..Absolutely agree. It is certainly a topic that deserves a lot more attention and a lot more discussion. Looking forward to keeping the discussion going!

  83. jeremystein

    i dont necessarily agree that new entrepreneurs are less qualified than previous entrants. frankly, im not sure what you mean by qualified. there is no doubt in my mind that entrepreneurship will be the next wave of employment. the majority of entrepreneurs will be freelancers (people leveraging skills that they already have). that being said, you can exclude them from this conversation because they won’t seek venture funding.i think that there are a lot of quality people who will leave broken industries and start new companies. there are companies doing *billions* in revenue that are *hurting*. smart people in failing industries might have good ideas on how to fix that.all first time entrepreneurs, quality or not, face a steep learning curve. lets focus on quality and for the sake of this comment, quality will be defined as going after the right market at the right time. although timing increases your chance of success, the learning curve comes into play. in an emotionally variant environment, i believe that there is luck involved with the learning process. some members of this group will be lucky enough to learn as they go, and others won’t. i think you will find that a group of new entrepreneurs will have a similar distribution of returns to that of second time entrepreneurs (this includes people who have failed on their first try).

  84. jeremystein

    The right market at the right time does increase the *chances* of success. There was an HBS study conducted on this topic and I believe it was in this article: Sadly it is behind a walled garden.The best and brightest aren’t always the ones to succeed.Anyone who plays golf knows that “nice miss” is a common phrase. If I am hitting my approach shot to the green for the first time, I might hit the ball 15 yards to the left, hit a tree, and when the ball finally lands it might end up a few feet from the hole. I might step back and say that was easy. Every time I wanted to land a few feet from the hole I might aim for the tree on the left. I might be able to hit it again, but eventually there are going to be devastating results.The same holds true for entrepreneurship, whether it was my first choice of career path or not. You can launch a product successfully but you can’t be sure that it was due to great execution or whether you met the perfect situation. If it was the latter, you might have trouble the next time around. How can you really be sure in the short term? Having great advisors/networks/etc helps, but they aren’t running the day to day business.How do you define experienced (does this include failed companies, several failed companies?)

  85. TU

    You’re right, the startup is not impacted negatively when stupid ideas are routinely funded. In fact euphoria for any asset class should increase valuations so all startups seeking capital benefitBut the impact on investors can clearly be very negative. And if the phenomenon is widespread, then at some point there will be a correction and then there are consequences for those who seek liquidity….seed capital, exit etc.

  86. jeremystein

    I’m not arguing which I would rather invest in.Just stating that the distributions would probably look the same. Each groupwill have huge successes, breakevens, and failures. Only a few people in theexperienced group (serial entrepreneurs) have a significantly higher chanceof success.

  87. mfeinstein

    I agree. I think that one source of the limit of scalability of venture capital is that, as Jeff points out, the model is based on building businesses that achieve scale. And, our economy can only absorb so much new technology at once (my theory — no proof provided). As the economy grows, more technology can be absorbed. So, the venture business can grow slowly, perhaps slightly faster than GDP (or shrink slightly more slowly). Again, my theory, and no proof provided.If I ever go back to school, I’d study this to see if I am right…

  88. azeemazhar

    Jeff,You are right–there is a problem inherent in the VC model which is the skew towards get big fast.But there is also a recurrent (often implicit) belief that fund durations are too long for GPs (who could otherwise be working at an I-bank) and for LPs (who hate the lock-ups). That is the point about all the J-curve arguments that are made, isn’t it?So I guess the question could be rephrased as: Given it can take a long, indeterminate time to build a sustainable business what are the right funding models that make sense?In some sense the conversation has got skewed. VCs hang on to their best deals or push them to exit in lock-step with roadshows for their next fund. A more active market in above-the-board secondaries could help a lot because you as Softtech might say: “Our focus is on getting companies to $50m in valuation, and then we’ll sell our stake on in a secondary”. Your return on an IRR basis could still be healthy. Factoring in for liquidity risk it could be excellent.It might also provide a focus for increasing specialisation (and hence better outcomes) within VC. Wouldn’t it be better for you if you could recycle Softtech Capital (and your GP time) back into your sweet spot rather than leaving both tied up in companies as they go through slow maturation to IPO?

  89. fredwilson

    Yes and SOX is not as big of a problem as people make it out to be. It was hard for a while to be SOX compliant but we’ve now got accountants and finance people who understand it. It makes being public even more expensive but its not the reason we don’t have a lot of IPOsThe reason we don’t have a lot of IPOs is its hard to build a company that can perform well as a public companyVCs and entrepreneurs have been badly burned taking companies public that should not be public. And the public market investors have tooSo the hurdle is higher. Maybe too high for most of this decadeMy gut tells me we’ll see more venture backed IPOs in the next five years

  90. fredwilson

    I hate overfunded sectors and copy cat investors

  91. fredwilson

    We are back in school mikeI’m learning all about power laws and poisson distributions today

  92. Facebook User

    The scale invariance of growth as a functino of capital investment is deeply interwined with the question of whether a poisson or a power law applies to these exits over time.