Over the years, there have been several excellent analyses of the ideal fund size for a venture capital fund.
The work that stands out the most is a detailed study in 2010 by the folks at Silicon Valley Bank -in which they found that sub-$250M funds signifigantly outperform their larger brethren. This piece is a must-read.
They looked at 850 funds and found that only 5% of funds over $250M have ever distributed greater than >2.0x capital (so-called DPI) to their Limited Partners. That is a pathetic statistic for the venture industry.
The brain-trust at HBS (Felda & Josh Lerner) came up with similar data that shows that the optimal fund size is $200M or so (credit to Todd Hixon of New Atlantic for the graphic):
There is alot of wisdom in these data sets gathered over 30 years of venture returns.
…which is why it is so odd that 2011 was the year of the return of mega-fund.
Mega is Back – But Is This Time Different?
According to the NVCA, $18B was raised by VC funds in 2011 – a seemingly great headline number – up 32% over 2010.
But what belies that number is a jaw-dropping number of firms raising mega-funds in the past year or so:
1) Bessemer $1.6 billion
2) Accel (growth + early): $1.35 billion
3) Sequoia $1.3 billion
4) J.P. Morgan Digital Growth $1.2 billion
5) Khosla (growth + early) $1.05 billion
6) Greylock (growth + early) $1 billion
7) Kleiner Digital Growth $1 billion
Only three things can explain this:
1) LP’s have gone mad
2) Returns of large funds have fundamentally changed since the SVB and HBS data sets
3) We are in the midst of another period / trend in VC that might again end badly
Part of this rush to mega-funds is clearly driven by the strong later stage (but still largely paper) returns that many of these funds have and will realize from investments in Groupon, Facebook, Twitter, Zynga, and other recent digital media IPOs.
My full spreadsheet of IRRs and multiples from these later round investments is here. The returns in these late rounds are much better than you might think – except for that last round in Zynga which doesnt look so good right now.
So these firms did what any good investor would do – they leveraged (largely unrealized) momentum to raise mega funds when the wind was at their backs. Yet another grand venture experiment in progress – one for which we won’t know the outcome for another 10 years.
Back to Fund Size – New Numbers
Now back to fund size – the existing analyses mentioned above all look at funds across the history of venture capital. The problem with this approach is that in the 1980s and 1990s, almost all funds were sub-$250M - certainly all of the good ones.
That also means that the record breaking 20-30x (!) funds that Matrix, Sequoia, Benchmark, and Kleiner had in the mid-90s are part of this data above and probably move it quite a bit.
So what is the optimal fund size when focusing on funds from 2001-2011 vintages – looking solely at IRR – since some of these funds are still in the J-curve to look at DPI?
The results actually don’t change much – small is still beautiful – with a caveat that requires explaining.
Two fund size ranges emerge as the optimal based on this data set of several hundred funds.
$400-$450M AND $100-$150M – across all quartiles and the top quartile funds.
Bruce Booth wisely pointed out to me the following caution when looking at this data:
Since there are only 8 to 50 funds per sub-group in this data set, the outperformance (or underperformance) of one or two funds can move the needle for the group quite sigifigantly – and in the case of the $400-$450M range, Accel’s 2004 $400M fund that is currently rumored to be valued at 12.5x – largely due to Facebook’s 95% contribution to the TVPI of this fund – probably is moving the IRR needle quite a bit as well.
So with that caveat – I still stand by the rule that small is beautiful in venture – the recent mega-funds and the imperfect data above not withstanding.
Navin Chaddha from Mayfield put it best in a reccent interview in the WSJ:
“I believe that the VC asset class will continue to attract investment, but there will be a flight to quality, with both teams as well as track record playing significant roles. There will be the haves and the have-nots. Further, in the haves category, there will be a divergence between firms that opt to raise and deploy sub-$400 million funds into classic, early-stage companies and a few firms that will raise billion dollar-plus funds to continue a multi-stage, multi-sector (IT, digital media and health care) and multi-geography strategies all in one fund. “
But will the latter strategy actually pan out?
The historical data implies that it won’t.
Nice analysis.
Of course, its difficult to do with any precision because the dataset is really too small. To get enough numbers to be meaningful you have to group together apples and oranges: tech funds, mixed funds, healthcare funds; US funds and worldwide; 1980s, 90s, 2000s and so on. By grouping everything together you must hide some significant variation between these groups.
Thats shown up by the data on IRR split into fund die “chunks” of $50M – the pattern just can’t be real (with two large outliers at $100-$150M and $400-$450M. What that is is just noise, caused by some massive outliers. Thats what happens when the group sizes are smaller, and it precludes any sensitive analysis of the factors they control returns.
But the overall conclusion has to be right: small funds show greater returns.
The question you didn’t really address was why that is?
Let me offer a couple of suggestions to start with:
In the early 1990s, I always remember Kevin Kinsella telling me about the early Avalon funds from the 1980s that had such high returns. With the early funds, capital was severely limiting so the investors had to be VERY careful with it. Every start-up was as leanly financed as possible. Everyone knew that every dollar had to count. The results were impressive. But as the fame of the fund grew, and more and more people wanted to be exposed to such high returns, capital was no longer limiting: now it was top ideas to invest in that became limited. To get so much capital to work, the bite sizes had to become larger. Instead of the investor telling the entrepreneur to use as little as capital as possible, they wanted to push as much at them as they could. And so returns fell. Bottom line: if you are aiming for 5x multiple, its a lot easier to reduce the cost by $1 than add $5 to the selling price.
The other factor at play is the kind of investments you can make (and maybe have to make) if you have a large fund to invest. In healthcare (where I come from) this means investing in later stage, capital intensive programmes if you have to get a lot of capital to work. Twenty years ago, these were perceived as safer than early stage (cheap) plays, because most of the technical risk had been removed. Not today. Commercial risk (that is, that the technology works but will anyone buy the resulting product?) dominates these late stage plays, and its a big problem. The problem is magnified by buy-outs – if large companies snap up early stage companies with world-beating technology, then the ones that stay private to consume the capital of the mega-VC funds are really the “silver medal companies” – the ones that weren’t good enough to be a buy-out target. And these “silver medal companies” consume vast quantities of capital only to deliver to the marketplace a product that disappoints in terms of sales, and therefore disappoints in terms of returns to investors http://www.tcpinnovations.com/drugbaron/?p=152
Small funds can’t afford to invest in these later stage opportunities. They have to go for the lean small early stage bet that sizzles with promise and potential, and which ultimately yields the best returns. By contrast, the returns of the mega-funds must be dominated by the large, flabby investments in “silver medalists” that flatter to deceive.
Very interesting analysis! Good job of backing up your points with actual data!
I’m afraid you (or the SVB folks) have fallen into a statistical trap I’ll call the Switch Hitter problem. Say you have two batters: Batter A bats 0.500 left handed and 0.200 right handed. Batter B bats 0.400 left handed and 0.100 right handed. Clearly Batter A must have the better overall batting average, right? Well, no. It depends on mix. If Batter A bats 80% right handed and Batter B bats 80% left handed, their respective averages are 0.260 and 0.340. The problem is mix in the sample set.
You’re looking across a dataset that has an enormous amount of volatility among vintage year averages at the same time that the mix among fund sizes also moves a great deal. If you don’t control for these together you can get a very misleading result. Indeed, your analysis defines “large funds” as >$250M. But I believe there are no such funds at all prior to 1991 so nearly half the vintages in your sample don’t have a single data point in one of your buckets.
A simple test. Take a look at page 7 of this Cambridge Associates data.
https://www.cambridgeassociates.com/pdf/Venture%20Capital%20Index.pdf
If what you’re saying is true, that larger funds systemically underperform smaller funds after controlling for vintage year, then the capital-weighted average IRRs (what CA calls the “Pooled Return”) should generally be lower than the unweighted arithmetic mean IRRs from the funds in those vintage years (since larger funds will be given greater weights in the former).
But in 87% of the vintage years you reference the reverse is true: the capital weighted returns are higher meaning that, on average, larger funds must actually be outperforming smaller funds in that vintage year! The average across all the years (weighting each vintage year equally, thereby controlling mix changes) is over 800 bps!
Lerner and Hardymon’s analysis is more relevant as it does attempt to control for vintage year, but even they are prisoners of history. There were no >$1B funds for most of the sample. They didn’t really appear until 1999 when returns for the whole industry (small funds too) began to tank. So now returns are lower… is that because $1B got raised or because there’s too much money overall? Had someone raised a $1B in a great vintage year like 1993 would it really have underperformed?
That’s not what the data since 2000 actually suggest… the $1B+ funds are doing just fine relative to their vintage year peers. They just don’t get the benefit of getting to average in blockbuster IRRs from the mid 1990s vintages since no such funds were in the historical dataset. Accel is #2 on your list… are they really lagging with Facebook and Groupon in the portfolio? And Felda Hardymon’s own BVP is right at the top of the list… do you really think he interprets his own analysis as a condemnation of $1B+ funds?
The most important takeaway, though, is all of this is rounding error compared to things like a firm’s track record (which is highly predictive of future return in VC, unlike in other asset classes). So the industry should stop focusing on it so much. What really matters is aggregate capital going into the industry, and that has continued to fall even as it has collected in fewer, larger funds.
Excellent feedback Tom.
And I’ll add to Tom’s reply that $1B Venture firms:
1) Few deploy money to monolithically to Venture (sequoia and PIPES).
2) Deploy money freely across their investment networks (thereby behaving like an index), and thus mitigate risk across those funds actively.
3) Deploy money in the same fragmented fashion as subprime VCs, meaning their risk to money ratio has turned subprime as well, skewing large funds differently.
So, as the old saying goes, not everything that can be counted can be counted on….but nice try