The Venture Capital Math Problem (continued)
My post yesterday is still generating comments. We've got about 180 comments so far and will certainly pass 200. It's an important topic so I'm thrilled that so many people feel compelled to engage in the discussion.
I've also received a number of private emails on the topic and several of them have included data which I did not have access to when I wrote the post. So I am going to do some follow up posts as I drill down in the data.
First up is the number of exits per year. The debate in my post was between 200 exits per year and 1000 exits per year. As you might expect, the answer is in between.
A friend sent me Thomson VentureXpert data going back to 1990. Here's the raw numbers:
Total Reported Exits Since 1990: 7,373
Total M&A Exits Since 1990: 4,392
Total IPO Exits Since 1990: 2,981
Whenever I look at venture capital data, I like to back out 1999 and 2000 because those years were not normal by any measure.
Total Reported Exits (less '99/'00): 6,204
Total M&A Exits (less '99/'00): 3,812
Total IPO Exits (less '99/'00): 2,392
If you take the data, after backing out 1999 and 2000, and calculate the annual numbers, they are:
Annual Average Reported Exits: 365
Annual Average M&A Exits: 224
Annual Average IPO Exits: 141
We know that these do not include all the exits as many funds who do poorly do not report and surely Thomson misses some of the exits from the funds that do report. But we also know that the 1000 exits per year number is way too optimistic. It's probably in the 400 to 500 range.
The next step is to figure out two important data points; the value of the biggest exit each year and the shape of the distribution curve of exits (is it power law, gaussian, poisson, etc). Now that I have the data and have access to the right kind of mathematicians (the readers of this blog), we can get somewhere.
Comments (Archived):
This is great data. Do you have average size (in dollars $) of each exit?
Would it be possible to get all the exit dollar values? Or at least the top/dominant few?Made some comments last night about the system/model (don’t think curve fitting is a good predictor for upper bound, it’s more a function of how much change “industry” or our economy can allow and the GDP)After the smoke clears and all of us go back to our respective day jobs, any chance we could get all the theories packaged into one report. Would be fun to see what the “Fred fans” came up with and have it as a handy reference. Maybe a handout to frighten new business grad students.This type of question/problem gives a great look at how much VC powered new business we can hope for (outside of the realm of BIG changes like entirely new markets which jump the shark). Also this would serve as a nice package of previous performance unbiased by individual fund supervisors. Ballpark numbers aren’t dangerous as long as you have to hang your hat on them. Probing data from many perspectives is a powerful way to characterize/understand it.
I have more data. I got three really solid data sets yesterday. What I lack is the time to sift through it right now. But I will
That’s what you have *2* junior guys for!!
They have their hands full with “real work”I try to do all the work that appears here by myself
Haha
FredHave been in touch with the guys at Thomson Reuters and they are kindly crunching some data on this. We should get it in a few days in teh internets can wait.
If 99 and 2000 were abnormal in the positive direction would that not be countered by the reduction in numbers post 2000, or particularly the abnormal numbers (if you have them) for 2008 / 2009? Maybe in the overall picture those numbers are required as they contain information from the preceding and following periods?
That’s the counter argument for sure.
Hi Fred,This is an interesting topic and your first post did a great job of quantifying the value of exits implied by the current levels of cash being raised by VC funds. I think the focus on the distribution of venture backed exits is interesting and a valuable exercise in its own right but, it distracts from the key point of the post. The key point for me from your first post was that exits totaling $150 billion pa will be required moving forward. Based on NVCA data over the last five years the average annual exit total was $50 billion (average disclosed M&A exit value * total no. of M&A exits + IPO exits). Do we expect to see the annual value of exits triple moving forward? I know you mentioned an annual exit value $100 billion, am I missing some data?Does the VC industry scale? I think as the size of the asset class increases we should expect to see the asset class’ returns decrease. This assumption is based on competition between VC’s for the best deals, and assuming the industry is functioning correctly, the expected return on the marginal investment should be below the sector average.Are $100+ million exits going unfunded? If so this appears to be a function of investment decisions rather than the availability of capital. Sure more cash being invested will increase the likelihood every $100 million exit being funded but the expected return of the asset class will decrease.Despite the potential for the sectors return to decrease the top performers should be able to maintain their returns.Sam
Excellent points, all of themThe math geek in me is attracted to the distribution of exits issueBut if 50bn is the number, then VC at 25bn/yr is a zero return businessAnd that’s the solution to the problem, plain and simple
The $50 billion pa figure for the last five years assumes the average value of the undisclosed exits is the same as the published exits. This assumption may not be valid.
Fred,this is a ‘black swan’ issue. Extremely wild swings in exits are part of the game. By calculating them out, you miss everything that is important about VC as an asset class.LPs want exposure to this asset class exactly because of this situation. In what other asset class can you get a 100x on investment when you hit a real outlier?The average is not the point. It is the extremes at the positive end of the ‘value at risk chart’ that make VC as an asset class interesting.Let me put it another way: take Twitter out of your portfolio. Let’s see how happy your investors would be with your job then 😉 But seriously, by calculating VC the way in which you do it, you do everything that Taleb tells people not to do. Have another look at the black swan book, it has big implications for what you try to do.BTW: I twice calculated the money in vs. money for a parts of the UK VC ecosystem. Once for biotech, once for the Cambridge Cluster. In both cases, there was more money going in than money going out, if you excluded the outliers. It is the outliers that make this asset class.
JensIf the outliers ‘make the asset class’ then it isn’t an asset class but rather a bet, since you really need to model the predictability of that. In a way, the investments need to make sense without the blow out successes because those successes are actually highly concentrated towards a few firms.It is less about what Fred does and more about what his LPs do. For all his wisdom, Taleb doesn’t offer anything better than Fama-French, which if treated properly and a bit of perspective can still be a very useful guide.So I fundamentally think that the problem has arisen around incentives: excellent salaries and very little committed up-front by GPs. The theory tells you that GPs are heavily invested in their own funds. The practice is that ‘heavily invested’ may mean 1% of the capital comes from GPs (and that 1% might not represent the vast bulk of a GP’s assets excluding their primary home). In fact I have heard of funds lending their GPs the GPs own commitment to the find to be set against their future carry. Nice work ;)I don’t know the situation within Fred’s firm but it would be interesting to get a sense of what those levels of commitment are.
we have invested after tax dollars in our funds and own more than 1% of thembeyond that, I am not going to get into the specificsI think GPs having real skin in the game is important and I also think fees vs carry is an important factor
In investigating funds (across classes) it is always a key question for due diligence. How much of your own money is invested in the fund? And how much of the funds money comes from you? And does it really come from you or from some loan governed from notional future earnings. I have seen some research on the determinants of success of VC investments (and funds) by different variables I am looking to dig it out.
for sure!why doesn’t somebidy do this simple exercise:use whatever assumptions and distributions you like, but1) cut management fees in half (from 2% of committed capital down to 1%) 2) assume all that freed up capital returns the same as any other invested capitaland watch what happens to the overall returns as an additional $0.10 of every LP dollar gets put to workseriously, with all the math whizzes around these comments, somebody do the exercise. i put on my swami hat and predict the asset class suddenly starts to look like a reasonable risk
Great pointsthanks
You state: “Does the VC industry scale? I think as the size of the asset class increases we should expect to see the asset class’ returns decrease. This assumption is based on competition between VC’s for the best deals, and assuming the industry is functioning correctly, the expected return on the marginal investment should be below the sector average.”This was essentially my question? As VC funds increase in size and number – but deal flow remains stagnate – where does all that capital go? Is this industry going to follow the path of the banks – but instead of subprime lending – will it be subprime investing just to show that it is actively investing? What happens if this industry goes bust?I met a prominate serial entrepreneur about 11 years ago who was raising his first equity fund. He was approached by Deutsche Bank with a billion dollar committment. He turned it down stating that there was no way he could properly invest all of that money. Yet today, we see many times that being rasied. There comes a point of stauration. I just don’t want to have to bailout these firms as well.If the real problem (not looking at the specifics of the math) is too little deal flow and too much money chasing these deals – then why not work to fix this problem. If there is a bottleneck on deal flow – fix the bottleneck – think about paypal and how that was developed – instead of chasing limited deals – why not creat your own. If there is too much money in the market – reduce it – doing more with less would provide better returns anyways and really help the entrepreneur in the long run by forcing them to be more innovative, creative, and flexible.These are just my thoughts but given some of the number you are putting on this blog – it just seems this industry is heading down the wrong path – much like the banks did when they started manipulating lending policies just to bolster short-term returns.
Fred- I can’t remember the source, but I have read that the median venture exit is $60MM. That is somewhat dated, but I’m sure you can find it or back into it from Thomson.
That seems low to me
This is probably true if you count all the “exits” where you get some money for selling the founder’s laptop for $500.
From the Boston Globe:The median acquisition price for venture-backed start-ups, meanwhile, tumbled to $63 million in the first half of 2008, from $93 million in 2007, according to data from Dow Jones Financial Information Services. The median price had been climbing steadily since 2002.
I think that’s a temporary dip
Perhaps, but since this is an analytic exercise, it probably doesn’t make sense to use a high water mark of 2007 or 2000, but a figure more normalized.
Fred- I think the best way to crack the math is to use as much LP data as you can get your hands on. That way you remove the uncertainty of unreported exits or at least get a better handle on the ratio of reported to unreported.Or better yet, let Noam Wasserman at Harvard or the Tuck Private Equity center crunch the numbers with their students.
Fred, I’m not sure your question is well formed.When you’re looking at a discrete data set, a binomial distribution with a low probability of success looks a lot like a typical power law. The models are totally different, but the curves look roughly the same. (Gaussian and Poisson are two types of limits of binomial distributions, so this applies to them as well.)If you’re trying to calculate total return to the industry, then the shape of the curve at the tail doesn’t really make much difference. The numbers are too small compared to the largest exits.So it’s probably impossible to tell the difference from the data, and even if you could, it wouldn’t much impact the result you care about.Generally speaking, you would expect a binomial (gaussian) from independent trials, and a power law from results that somehow built on each other. Probably you get a little of both. It’s quite possible that a blend of two models is most accurate.
The annual average if you include the 99/00 numbers are not off by much. I think that you should use them. As Ian Wilson points out below, those numbers are part of the return however irrational those returns may been. The reversion towards the mean post run up compensates for it automatically, doesn’t it? Am I making sense or do I need coffee?
You’ve been making perfect sense since we started this discussion yesterday morning
Can you share the yearly raw data set from Thomson VentureXpert (1990-current). Sometimes looking at the entire raw data set is better. Does it provide any more information than just a table? For ex: Does the report categorize exits by industry? That would be a neat way to compute returns by industry type.
i have the data but i am not sure its right to publish it on this blog or by making a spreadsheet publicly available. that data comes from a premium source that i don’t even subscribe to
Instead of working back from one big exit using a power law, why don’t you estimate the average exit and use a Gaussian (normal) distribution to work out the total size of the universe?My intuition is that it’s easier to make a decent estimate of the average exit, and the math around the normal distribution is a lot easier to understand.
But how do you know what an average exit is?Most VCs would say $100mm plusI’m not sure that’s accurate
Unfortunately I haven’t done stats for over 20 years, but my hunch is that your more likely to get a more accurate result by estimating the size of the hump of a normal distribution than estimating the peak of a power curve.I suspect the former technique is less sensitive to error.But we need some proper maths guys to confirm this.If it’s true, then you can just do a poll to find out the average exit size and weight it by the respondent’s authority.Crude, but possibly effective…
There is no need to try and estimate the peak of a power curve, The peak is for those exits that happen once in a million years (e.g. have a probability approaching 0)What you do need to estimate is two points in the graph, any two points.For example the number of exits between $100mm and $110mm per year and the number of exits between $200mm and $210mm.You need one more thing, an approximation (no need for a perfectly exact formula, as there is none) of the distribution of the exits. I does not matter if it is binomial or power curve, the data set is too noisy and too small to tell the difference anyway (a power curve is much easier to calculate)It is relatively easy to calculate the area under a power curve, or a normal distribution, I guess excel can do that in a few seconds.
That’s the whole point, the gaussian is a bad fit here. The left tail is truncated by the zero bound, too close to the peak. The right tail is where you have substantial fracion of aggregate returns; Small inaccuracies in the fit will yield large diffs in total return…
My belief is that you cannot exclude the 99/00 data from long term returns on venture capital. Such “boom” cycles are as much of the economic landscape as the current “bust.” Returns and exits do not follow a normal distribution but are non-Gaussian not just through outcomes but through time as well. Nassib Taleb talks about this in the Black Swan but despite his protestations it’s pretty well known in hedge fund math. Also, as an asset class, perhaps venture capital will tend to greater granularity. I think the math of web 2.0 investments might look radically different from IT infrastructure for example.
I feel that we are asking the wrong question when we ask “Is there a room for $25 billion in VC money?” The question to ask is”what do we need to do to make Venture Capital a scalable asset class so that we can put in more than $50 billion in 2015 and generate the same kind of returns?” And then the answers that we get will be significantly different: Do we need to teach a different set of skills in school so that more people think about starting VC fund-able business early on in life when they dont have too many responsibilities. Do we need to encourage experienced managers from established industries to rethink the whole business. Can VCs invest money in building the system to grow the entire pie. for example: bring in people to do the kind of things Fred and team are trying to do with education? Can we as a community build a support system and tools to grow the # of people who can start successful companies and there by create room for $50 billion in VC investment in 2015.
I like to think expansively the way you do but build my return assumptions with a more conservative mindsetThat way if my optimistic side is right, we win big, but if my conservative side is right, we still can give our LPs what they expect
I totally agree with that perspective.
The problem with the power law (and probably, poisson, etc) approach is going to be sensitivity. As mentioned, it is a black swan style problem. But the argument in Black Swan was that people didn’t model appropriately for those scenarios, for a multitude of reasons. One is psychological, but another is just the impact that extremely slight variations will make in your results. For instance, assume you believe the power to be on the order of 1.035. Changing the power by one thousandth dramatically affects your average outcome, as well as those top outcomes.I’ve made a little Google Doc with some charts of a hypothetical 2000+ companies and a power estimate – I’d encourage you to play around with it and save one for yourself; I’ve opened for editing now so that folks could save and add their own key metrics…we’ll see how well that works. http://spreadsheets.google….. Right now it’s not great because the power is too extreme, but I’d play with it as more data comes in.My belief without crunching a lot of data is that overall data is pretty close to power law…but there’s also an issue in the tail of the typical distributions. A typical power distribution would tell you that to get your top-end distribution right (ie, a handful of hundred million, maybe 1 billion dollar deal), there’s a slew of low-level results. If actually use a normal power distribution, then there are hundreds or thousands and then lop off the results below, say, $2M. While venture focuses on that top end, there are theoretically a massive number of companies that created some value. Now, once you consider the investment required to get them to that point, they are net negative…but how many of those, really?Now, I’d think that right now a big reason that long tail doesn’t accurately model reality is that there really is not a lot of exit opportunities for companies in the few hundred thousand range to maybe $1-2M range – yes, they certainly exist, but not in the numbers (ie, thousands of exits) that would make the rest of the curve work at the top end.My though exercise would be – is there a structural change that could enable value to be created in that tai? If you consider that much of the tail scales similarly as you “zoom in” to various parts (or scale back your power), is there a way to enable better, smaller exits for all involved, and that the right firm, with a larger number of small bets with a different desired outcome (ie, NOT trying to hit that black swan) could find returns in that tail?
Thanks for the google spreadsheetI’ll play around with it
So the distribution of returns isn’t quite a powerlaw, it is most definitely more poisson-like. The reason is that under any scenarios a Powerlaw can’t cope with the left hand extreme of venture returns which is that the most common return for a portfolio investment in cash multiple terms is between 1x and 2x. And 0x does happen but just not as commonly as 1x or 2x. So the shape is more poisson like.As I mentioned we analysed all this in depth in 2007 but several laptops on I am struggling to find the Monte Carlo models and datasets we built this in.
It’s great that the problems with VC returns have forced a quantitative analysis of the asset class — as usual this blog is leading the way. I want to reference the qualitative idea discussed briefly yesterday, however, because I think its under-appreciated, namely that there is a huge difference between an great business and a business that is a great investment. There is no ‘scale problem’ with entrepreneurship — there will always be room for small businesses that effectively address market needs. There is a huge scale problem, however, with VC investing, because returns depend exclusively on liquidity events. Having lived in the early stages of the ecosystem for a long time, I think the most common mistake angels/seed investors make is to conflate the idea of an excellent, even cash-flow generating business, with one that has a realistic chance at an exit. Even more common is the mistake of investing in a company that may have a realistic chance at an exit but not at a multiple that provides a sensible risk-adjusted return! In other words, when you make a seed investment, you are entering in at the highest-risk stage and should expect to have your money locked up for 4-5 years; IMHO you need to receive at least 3x your money at the exit to have justified the risk and the illiquidity. I guess that brings up a related point that I don’t think is discussed enough in the seed/angel VC world: every investment needs to be compared against what could be made over the same period of time by investing much more liquid, much less risky, asset classes (bonds, publicly-traded equities, etc.).www.vc-brazil.com
“If the outliers ‘make the asset class’ then it isn’t an asset class but rather a bet, since you really need to model the predictability of that.”Bingo! You win. We are not dealing with a true “asset class” that you can model like stocks (dividends), bonds (interest), or real estate (rent). You are dealing with a lottery (fixed duration, variable payoff). Attaching the “asset class” moniker to venture capital was a marketing move designed to make the whole activity seem rational and quantifiable so that the pension system would buy into it.
Sorry to butt into this conversation, but I don’t quite understand the discussion; nevertheless, I really want to try wrapping my head around it.Is the basic issue that, unlike other things in which you can invest (securities, bonds, etc.), venture capital is inherently limited? So that, unlike other investments, there really is limit on your potential return?And here’s the dumber question… how is this related to finding a “best fit” to model of the average? To see if VC really doesn’t scale?
Brandingme — theoretically, VC is not inherently limited in the return it can provide. But one main difference between VC and publicly traded asset classes is that you can get in and out of your investments in the latter pretty much at will, i.e., there is liquidity. VC is probably the least liquid asset class. Fred has blogged about the need for a secondary market in VC investments in order to provide some liquidity but, until that takes root, the only way you get your money out of a VC investment is if there is a liquidity event, typically an IPO or a sale to another company. These are exceedingly rare, especially these days, making it ever more likely that, even if you invest in an excellent company, you will never get a dime back. I said that, to justify the risk and liquidity problem a seed investor needs 3x over 4-5 years; the truth is that 10x is more reasonable. David Rose of the New York Angels says he needs to believe in a 20x return or he won’t invest.vc-brazil.com
Thanks for the reply. Now I know why Fred advocates for a secondary market. How does this relate to the talk of VC not scaling, however?
Essentially, because at some point more money is put into the VC asset class than can ever be returned, at least not returned at the kind of multiples that VC investing needs to justify itself. There just aren’t that many grand-slam companies and/or exit possibilities.For example, assume that LPs invest $200 billion in VC funds in year 1, is it reasonable to assume that there will be liquidity events in the next 3-5 years that return $1 trillion (5x gross returns)? Probably not. Again, there just aren’t that many grand-slam companies and/or exit possibilities. On the other hand, if $200 million is invested in VC funds in year 1, it is reasonable to assume that there could be $1 billion returned (5x gross returns) over 3-5 years? Probably. In sum, VC doesn’t scale because at some point there is too much money chasing too few good deals and returns in the VC asset class overall plummet.(There are probably 50 people on this blog that could do a better job of explaining this than I can but there you go…)vc-brazil.com
Well, you did good, this makes perfect sense. Thanks for giving me a jumping off point as I read through all the other comments 🙂
Hi to everyone, new to the group but interested in offering some perspective from an emerged market.Entrepreneurship has and will continue to provide new solutions to the inefficient systems in place we use on a daily basis and it better be promoted at all levels.Regarding scalability, shouldn’t VC’s be looking at new markets in search for oxygen in these troubled times? The industry is well underdeveloped and opportunities to cherry pick the low hanging fruit are overwhelming.My approach as a new participant in the VC industry is an entrepreneur. I have done it for almost 20 years in a series of deals and it has worked with far better results than those being shared here.I agree that VC’s should align their interest with the LP, entrepreneur, and stake holders to be able to generate and distribute value, so bootstrapping your VC operation is the best example you can give; fees should be based on bottom-up budgets not on fixed %’s. In my experience sometimes great ideas fail because too much money is given at any time; there should be a bare to the bone use of resources philosophy to make the business burn only as much needed but close to what you are able to generate.If you guys question the number of quality startups in the US, what do you leave to our emerging markets. Good ideas with experienced entrepreneurs are out there but one has to go and dig them out. Here, money is not after the deals, the deals are after money because the PE/VC industry is none existent. This scenario allows the entrepreneur to achieve great things with minimal resources. Like first concentrating in nailing the concept to then scale it. What I have found to validate with some LP’s is that it works well when the GP creates its own entrepreneurial thesis for the market and builds around it to capture the value of the sum of the proposals.Exits? That is the art part in this industry.I enjoy reading your posts today and look forward to participate in future conversations.
VC is a local businessWe need to see local VCs emerge and help them get funded
using the data you gather from the math answer, it will be amazing if you are able to somehow show a mathematical correlation of IPO exits into the NASDAQ to further justify the correctness of your blog here http://www.avc.com/a_vc/200…
I hadn’t thought of that
Off-the-wall thought: we’re concentrating a lot on the _value_ of exits (and hence the overall total ROI that the VC asset class has). We’ve noted that a lot of the exit values might well be in a “long tail”, which would not necessarily be included in, say, the top 200 deals.What about the ROI of individual investments? That distribution would provide insight into whether VCs generally do achieve 2x to 3x returns on any size of investment (and hence they’re “good” for a company, whatever the size of the investment). If VCs’ investments’ ROIs follow a Gaussian distribution, what’s its mean? Moreover, has it changed over time, as the total amount being invested has grown? My guess is that given the old rule of “3 out of 10 fail, 4 do OK, and 3 work well” (or similar) that it’s nothing like Gaussian, because return rates on a few deals have to be huge to make the mean 2x or 3x.If the distribution is unchanged over time, that provides some hope concerning the scalability of the asset class. This is irrespective of the _size_ of the deals being conducted.Also, if one could compare the distributions for different VCs, one could see which ones were good to have invest in your company: if a VC achieves 3x returns consistently across the board (if only!) they’re far more preferable from an entrepreneur’s point of view to those that rely on huge returns very occasionally, even if both funds have the same ROI overall.However, I’ve no idea if this data is available…! Does anyone else think it would be useful?Separately: can whoever supplied the Thomson data (and hence has a subscription) produce a scatter graph of exit values and post it somewhere? Just an image would do, no raw data. That would give us a vague feeling for whether we’re anywhere near correct on distribution shape.
Some interesting points. I’ll throw in my $0.02, without any data:I’d expect the distribution of venture returns to look log-normal versus a power law. They look fairly similar for the “average” deals, but I think that the scarcity of buyers at the very top of the market will mean that there isn’t scale invariance, which is important for the power law to hold true.Put a different way, since there are fewer buyers that can play at the high-end of the market, the price distribution isn’t likely to look the same at the high- and low-ends of the market. It might if you included the post-public run up, but as you note most venture investors have cashed out before that point.My guess is that the VC “math-problem” is actually driven by the fact that there’s too much money chasing the outlying big deals, which are more infrequent than most investors think and offer a less-optimized price. I’d think investors would be better suited looking for a way to manage and profit from the fat part of the distribution: the small- and medium-sized exits. But I understand all the reasons why this is a tough model for VCs
thanks Tomgreat points
Fred, ever since your original post there’s been something about the logic of your initial argument that’s been bugging me, but has been annoyingly out of reach…Finally, my sub conscience spoke to me, so here it is: as a person who spent a lot of time crunching flow of funds numbers for the stock markets, one thing you learn pretty quickly is that the money to fund a prolonged market rally just appears (as if by magic). Even when you think that all the players have put their last penny into the markets, new funds can just keep on coming in for years after you called the top.This implies that there is no practical limit to how high markets can go (from a flows, not *valuation* standpoint). The same is surely true of the IPO and M&A markets. If there is an appetite for deals then the money to finance them will come from somewhere.So I would argue that the current sorry state of the exit market reflects a lack of *appetite* for what’s on offer, not a lack of capacity. And as such does not represent a physical cap on the size of the VC industry.Speaking on (slightly) less solid ground I would further argue that your original thesis doesn’t so much reflect on the state of VC industry, but on the business models that VCs are currently funding.For internet stocks in particular, once the market realized that simply buying traction was a risky business, exits started to dry up. When web companies find a way of generating some good old cash flows then the appetite for acquisitions and IPOs will surely return.
i totally agree with that last pointthe value of a business is the present value of future cash flows
Supply and Demand?A theory:The VC industry’s aggregate returns are starting to behave like a mature industry with excess capacity. The excess capacity was created when money flooded into this once red-hot asset class, causing returns to suffer greatly. Unfortunately the flood of money created “sticky” excess capacity– too many VCs, continuing to invest and raise too many dollars. Redeploying money and people (VCs) didn’t really happen.The return to equilibrium and average risk-adjusted returns is a long one, since:– Funds lives are long (therefore accountability to the investor is very muted in the short-term)– Data on real returns takes a long time to season, since investments take a long time to season — Data on real returns is hard to measure (opaqueness, black swans, etc)– New VC capacity like their carries and careers (and why not?)Having said all of that, the VC industry has a strong ability to create its own demand– Say’s Law. Great VCs, investing in great ideas with great managment teams, will create their own markets. The World economies can easily absorb $25 billion of new investment per year with strong returns, IF the money is invested productively. The excess capacity argument would not apply if most of the money went to work int he right places.The problem with the flood of new money was (and is) the VC industry’s ability to absorb it quickly. The money is only spent productively when there are enough seasoned VCs to continue to choose the time proven:– right idea– right time– right teamThis aint so easy. So, capacity needed to grow more organically for returns to stay high.To test this theory, I suspect returns from the Top VC Firms have stayed high; perhaps they are lower than historic levels, but are they still higher than alternative asset classes (risk adjusted)?Thoughts?
I agree with most of this analysis. Though I can’t say that I’m convinced we can put $25bn per year to work productively even if we had an industry full of experienced and capable VCs
Thanks.Well, it would certainly take a very long time to “grow” that many good VCs…so that point of the debate may be moot anyway.Do you know of data to test the thesis about the returns for the top VCs?
Venturxpert from thomson reuters
But this does show how to fix the problem. Moving forward – the real world answer to the venture capital math problem starts with demanding capital efficiency. I commented on Azeem’s answer here: http://azeemazhar.com/?p=25… – Someone replied with an example of Google, but the reply didn’t apply to what I was really trying to say. Google is the model for capital efficiency. Look at this http://news.cnet.com/8301-1… – capital efficiency at play. If you look at their search engine’s code you can see that they are attempting to save dollars on each nanobyte. Now take Twitter. I am the first to say I LOVE TWITTER. But $56mm in funding and the first thing we hear is that they are an overpriced Apple tech company very well known for a Fail Whale. Sure you can’t put a price on innovation, and I see Twitter’s game changing future ahead, yet I see Sergery Brin and Larry Page both smacking their foreheads! I even read an article in Rolling Stone a few years back about a few upcoming tech stars who would use their venture capital funds for nights out. I even recall reading about prostitution! WTF! The venture capital industry needs to find companies and innovative individuals that cost less to IPO and M&A. That would fix your numbers moving forward. At least choose companies with clear paths to exits with no break in their internet business models. This is not right! http://twitter.com/copeters…And while I am on this kick. I understand capital efficiency and am seeking seed money for master programmers, graphic designers, and refurbished Dells to build patentable internet technologies. (sorry had to do it.) one of them is the internet version of Walmart. Not what you are thinking, what I am SEEING and engineering. http://categorykiller.infoThanks for the reads Fred. I am learning so much about this industry from you.
Don’t be fooled by how much money twitter has raised. Focus on how little money they have spent in three years.
I was surprised at the comment at the end of the last post that the VC industry would go ‘back to the future’, because I thought you created the future. I don’t think there’s any way to get back to the future anymore than to have record labels, neighborhood book stores, video stores, classified advertising or having a Normal Rockwell come to life.I really enjoyed your talk at Columbia B school that you posted not too long ago where you talked about going from generalist to web centric in focus. I think that trend of specificity will continue, not go back to the way it was. The limit on creating software/web driven solution used to include capital, now it’s not as much. I bet the VC model will continue where there are higher capital needs and fragment along those lines. If you keep focusing on Web services, eventually you’ll know what you’re looking for so well you’ll be able to build it more efficiently than waiting for someone to come along who has traction, experience and no competition for investment dollars.One way to create scarcity in web/software deals though would be to go where people don’t want to go, which is concept businesses with first time entrepreneurs. For good reason too. The next way to create scarcity might be to use what you know to create businesses that creates new solutions for existing industries because VC as a class are smart and well connected. That can translate to sales early to key customers across many different industries for web-centric solutions that need to exist for existing demand.Maybe some version of an incubator will end up working by being more predictable and include larger %’s of ownership. Perhaps more than ½ right out of the box. When you do your next posting on this topic, perhaps you can use the exit assumptions as fixed and vary the % of ownership assumptions. That’s another way to skin this cat.I bet your instinct has been steering you and Brad this way (toward a leveragable position and process) since way back when you raised only $125m from a feeling about how your business scales. You also commented on having too many partners in a previous firm recently.All this math is subterfuge. To me it’s obviously a power distribution to me and if there is too much money chasing too few qualified deals, that need less and less money, the model falls apart, on average. It’s a business model issue. The math is a trailing indicator.Something new will come though, just like newspapers, video rental, books, whatever.
By back to the future, I meant the way the VC business is organized and structured, not what we invest in
I was commenting on the same, but I wasn’t clear. I was questioning the ability to go back to that structure because the environment is different now and that perhaps a new structure to this part of the industry will emerge.
I am sure there will be new structures emerging like Y Combinator but they are net additive to the existing VC business
So here is another way to look at it. Assume you have $1,000 in the fund and do ten deals at $100 each. Then, of those ten, two are great exits (stars), two are just good and so on as per the table below. The stars obviously dominate the math, and you can see that to get a 3x return in five years (with no dilution) the stars have to return 8.9x.Assumptions:Time to Exit = 5years% Dilution Upto Liquidity = 0.00%Number of Investments @ $100.00 each = 10Annual Return for Star = 55% or 8.9x(probably won’t show up as nice table but….)Dispose As Investment ROI/yr % Return$=============================================Star $200 55% $1,789 or 8.9xGood $200 30% $743 or 3.7xOK $200 15% $402 or 2xAlive $200 0% $200 or 1xWrite Off $200 -100% $0 or 0x==================================================TOTALS $1,000 $3,134 or 3.1xThis means that a $20M investment for 50% of the company has to yield at exit of $356MM. Therein lies the problem. Except for the outliers (Google, eBay, etc) the exits just cannot support the returns expected, even assuming 1 in 5 venture deals have great exits.Yet another qualitative thought. As business models and types of company have changed to require less money, why do funds get bigger and bigger requiring larger and larger investments. The venture industry needs to be innovative just as it expects its investment portfolio companies to be.
Its a lot easier to make these numbers work if the investments are smaller amounts