The Fallacy Of Bimodal Returns
There are a lot of "falsims" being bandied about in startup land these days. And one that really bothers me is the idea that returns on startup investing are "bimodal".
For those who don't talk in geek speak, bimodal means there are one of two possible outcomes. And in this case, those two outcomes are a total bust or a huge Google style win.
If you buy into that logic, then you want to be in every deal because if you are going to take a massive number of hits, you need to absolutely be in that Google style win or you are toast.
But startup returns are not bimodal. They exhibit more of a power law curve. There will certainly be one or two venture deals every year that generate 100x or more. And there will certainly be quite a few total busts. But there are a lot of outcomes in the middle of those two. And you can make a great return investing in startups without being in the 100x deal.
Here is the distribution of current returns in our 2004 fund. To be confidential, I am not listing company names and have "fudged" the top returning deal number so nobody plays a guessing game with that one.
A couple things about these returns. First, many of these returns are unrealized and carried at valuations forced upon us by our auditors under the auspices of "FAS 157". If we were working under a different accounting paradigm, we would be carrying many of these investments at much lower values. Second, this fund is only six years old. And so we are still carrying one third of our portfolio at cost. When this fund is fully realized, I am pretty sure there won't be a single investment that is worth exactly its cost.
So don't get too caught up in the total numbers here. The point is that startup returns are not bimodal in any way. They exhibit a power law curve. And you can make great returns playing in the middle of the curve.
I've spent my entire career playing the middle ground of this curve. With the exception of Geocities, which my partner Jerry led at Flatiron, I have never seen a 100x return. I suspect our first Union Square Ventures fund will change that. But from 1990 to 2005, a span of fifteen years, I built an excellent personal track record that helped me and Brad raise our first fund working exclusively in the middle of the return distribution curve.
And the way you do it is you keep your "busts" to less than a third of all of your deals and make sure you don't put a ton of capital into a bad investment. And you work the middle third to make sure you make a decent return on them. And you follow on agressively in your winners. You do that and you can post gross returns in the range of 50% annual returns gross before fees and carry.
The thing you most want to avoid is "doing every deal". You need to select good deals and avoid bad ones. That's what bothers me most about this "bimodal" argument. It suggests that you need to be in every deal so you can catch the one big winner. That's a bad strategy for everyone but the one person who can actually get into every deal. Because there is a good chance that you can't get into every deal. And there is also a good chance that the one deal you can't get into is the one that is going to be the home run.
So pick your deals carefully. Accept that you may not get into the next Google. Work the middle part of the return distribution curve. Recognize your bad decisions quickly. Work your deals hard. And follow your winners. And you'll do just fine.
If you want to read more on this topic, please check out this post by my friend Bryce on mostly the same topic. He wrote it last week and it was one of the inspirations for this post. I was also inspired by this post by Ron Garret.
It feels like they are some pretty big names on the other side of this argument. Are the people on the other side of this people like Paul Graham and Marc Andreessen? I know one person saying this is Mike “thunder lizard” Maples, though he argues that this bi-modality means he wants to very selectively go after thunder lizards, and not try to be in every deal.I’d like to know who the credible people are on the other side of this argument because what you are stating seems trivially obvious to me, and yet some seemingly smart people disagree so its always interesting to try to understand where the real points of disagreement are by looking in detail at both sides of the argument.
i don’t want to make this post into a pissing contest so I am not naming names
Lazy VC’s will go bimodal. That’s their names.
One of your best ever posts Fred. Chapeau!
i do best when i get fired up about somethingthis bimodal thing really bugs meto the core
Great post. From my side, it seemed more legitimate to heavily handicap the assumptions and look at six years. To me, that was looking from the standpoint of the ol’ ‘if it is going to happen, it will’ and be ready.It would drive you nuts if you had a gang of investors panting over the immediate 100x payoff instead of looking at what are the best options moving into the second, third and so on years.
If I remember correctly there was a post a while back about swinging for the fences (i.e. hitting the home run, gunning for the huge exit, backing a company who’s going to be huge) and some people, myself included, said I’d take a lot of singles instead of one home run any day. That sentiment seemed to get shot down. Yet, this seems to be that to a certain extent (provided there are some doubles and triples, too). Perhaps I’m remembering incorrectly or mixing two different examples (I’m no VC that’s for sure).
you want your cake and eat it too in VCi want the home runs toowhen they are there to be had
Good post, Fred. The bimodal fallacy is particularly detrimental in the context of a weak IPO market. It’s awfully hard to get above 10x in a particular portfolio company absent an IPO. Most successful exits will be in the $100-300 million range, which requires capital-efficient models, low post-money valuations, and (as you point out) limited losers and a few good winners to drive a fund into the 2-4x range net of fees. There was a 5 year period where things got silly, and there will always be a small handful of ridiculously large winners (e.g., Genentech, Amazon, eBay), but the top two quartiles of the 300-500 active venture funds will never be in those “once in a lifetime” opportunities and yet still solidly beat the S&P.
i agree Jeff but i am optimistic that we can get well north of 10x inour best deals with select M&A and secondary sales
being a consistent top quartile fund doesn´t require 10xs but don´t exlude them in advance
Doesn’t this also suggest that in venture investing, including (or especially?) early stage investing, price and valuation not only matter, but matter alot?
That’s the second fallacy that is being bandied about. I plan to post aboutthat too
Fred,I agree with your original post. I think there is a lot of money to be made “in the middle”. I’ll be investing in ‘hardware deals’, so it is unlikely that I will ever see a 100x. But I will be very happy with a lot of 10x, a few 20x and very very few 0x.That said, I too believe aweissman comments that Px and Valuation matter a lot (as does timing of investment). You said you were going to post about this… Could you give me a hint now? Thanks.mjl
yes, when you are taking a lot of risk, you need to discount that risk appropriately or you will not make money in the long runa low price is the way you get paid for taking risk
Bryce wrote a great post about this the other day…here’s the link for anyone who hasn’t seen it.http://bryce.vc/post/138492…
that is such a good post, i am going to link to it
Yeah this is again an ex-post fallacy. Of course price matters. Simple math – the lower the price – the more you own. If you have chosen wisely – the greater your returns will be – as you own more of your winners. Returns are not about batting average or on base percentage – just getting into every winner and an equal number of losers does not generate great returns. Generating returns is all about slugging percentage: what % of my fund did I put into the home run and what % did I allocate to the strike out. It is there where price really matters. Where price doesn’t matter is where you have no conviction – and just want in on the deal. And if that’s the case – why bother?
i love the slugging average analogy harry
Great post and really looking forward to the followup about venture investing and its early influence on a start-up.
I could not agree more with your approach and your concerns about the invest-in-every-deal apprach. Being a VC is all about portfolio management and you lay this out in a very succinct manner.Theory is one thing, but actual results and track record are another. You have laid out your distribution curve in this post and it would be interesting to see the distribution curve laid out by a proponent of the “spray and pray” approach. I suspect the selective approach has a larger area under the curve.
Great post Fred. The bimodal thing seems to drive poor decisions in start-ups too. I look at some businesses which have established (against all odds) brands, substantial traffic and potential to generate revenue — and have been backed by tens of millions of dollars — and yet still don’t seem to be on the path to phenomenal success for their investors. Digg springs to mind.Entrepreneurs don’t want bimodal outcomes–the risk curve for an entrepreneur in a bimodal world (1 in a million chance you are Google/Amazon, 999,999 in a million that you are bust) is absurd odds, particularly if you take the convexity of utility. I am much happier with a 50% chance of a payout of $10m than a 0.5% chance of a payout of $10bn, because the top part of the later payout is cream that I can’t process.But that said a few things still seem to be systemically tricky in a world of 5x to 8x exits* Large funds with small numbers of partners still need big exits.* Too much capital in a sector can muddy the economics – forcing companies into ever greater investment (look at Winchester disk drives and, arguably, real time search), but if the exits aren’t there at the 100x level what happens to the eventual winner? And can it ever return decently to its backers?* As an entrepreneur, much of the language we hear from VCs speaks of the ‘being the next Google’–at least in public conferences or their slideshare presentations, rather than the realistic and practical conversation you’ve started here.bestAzeem
I especially agree with your last point. It’s important for us all to realize how much of the talk in VC/entrepreneur circles is bluster or extrapolation from personal experiences rather than axiomatic.
I have always pointed out that the 10 year slog to build a great company (not google) just a profitable technology company is never celebrated because its just not interesting from a press point of view.If you’re a VC I agree completely that you have to go into every deal believing its a 10x, and to get in on those deals you have to celebrate the ones that work.However, the most popular post ever here was about the entrepreneurs that push the ball up the hill for ten years.
Fred…this is one of your best posts.Dreaming big is great (and human nature) but on the investment as well as working in the startup trenches, the middle ground is where we mostly find ourselves. And we can win there.
Thanks Arnold. I know you’ve read a ton of my posts so that means a lot
This is certainly a candidate for BOAVC (Best of AVC).Disqus team: where is that feature where I can flag a post, Craigslist style? (best, worst, inappropriate, etc)I suppose for now, a blog owner could simply have a feature to list their posts by total Likes, but that’s not quite the same.
I’d go for that. That, and making sure the like button integrates with the facebook like button (the new seo)
Are you sure this is power law and not log-normal? I.e. what’s a log-log plot of this distribution look like?
great, important post. glad you wrote it.
If you put the same money into all these companies, which I am sure you didn’t, from a basic calculation it looks like you returned (will return) about 5x on the fund – congrats!
Feels like this is a reaction, at least in part, to some of the things said at the Angels in the Architecture panel the other day. I would have loved to see a more forceful discussion around this point there.Also, as a founder, I dearly hope you’re right. The expectations forced by a bimodal return distribution are out of control and can seriously warp critical decision making. If you swing for the fences every time, you’re going to strike out a whole lot. So while I’d love to create the next google, it’s good to hear that there are opinions in the industry that value a Tony Gwynn.
Does this mean that you would invest in companies that can’t be a huge success (becaouse they are too niche or whatever other reason) but are likely to have a nice return? or you look for potential huge successes that can give you a nice return or a burst if they go wrong?
We seek to make 10x on everything we doMost of the time it doesn’t happen
It would be an interesting world if we applied that expectation of returns system to adopting new tech. If it doesn’t make me 10x more effective I shouldn’t even consider it.
That’s exactly the question I had…It seems many investors (angels and VC’s) only look for businesses that could he HUGE, perhaps because they know they won’t be?Is it like shooting for the moon – if you miss you’ll land in the stars?Because I think there has to be plenty of counter-examples where great journeys began with one small step at a time…
It’s interesting that many natural and healthy ecosystems follow power law curves for populations. There if one species grows disproportionally it can have severe consequences for the habitat and other species that dwell in it. This contrasts with business where one startup’s extreme success can have benefits for other nascent businesses. I wonder if USV’s success in the profitable middle is at all related to population models. I’ve brought up chaos theory before when we talked about the distribution of returns and regions of stability. Some years are just awesome to invest in because the market gobbles up companies. In other times no matter how fantastic the business, it’s fate is sealed by bigger macro trends.Look forward to your related post on valuation. What valuation is best for any startup? Not the highest now, but the right ballpark valuation now that optimizes your startup’s chances for survival and growth.
Just wanted to add, USVs success is related to systematic application of driving principles of investment. There you can’t predict extreme outliers but reliably can minimize investing in poor performers, and grow the rewarding middle.
paradoxical that a VC can built a great reputation and returns playing deliberately in the middle of the curve but an entrepreneur still has to pitch in the curve’s upper echelons to be taken seriously?
I don´t think playing in the middle of the curve is a deliberate decision. What is deliberate is the 1/3 strategy: 1/3 quick write offs + 1/3 money back + 1/3 heavy follow ons
I think what people are missing is that the 1/3 strategy enters after the funding decision has been made. It does not influence in the funding decision itself.
are you consciously aware of this distribution when making investment decisions? i.e. are you sitting there going, “ok, this feels like a middle of the curve investment, i like it. this other one, however, feels like it will end up on one of the extremes, i think i’ll pass.”is it there a class of opportunities that you *do* treat as bimodal – hit or miss – type investments?thx.
We don’t try to build this curve. We make every investment with the beliefthat it will be a 10xThe curve is what happens to us
Fair enough, but you have to at least expect – when making the investment – that some of your companies are more or less likely to end up in the middle of the curve.ex: At the time of your initial investment, and leaving the team aside:Twitter = swinging for the fences on a 3-2 count.It’s a home-run or you’re fuckin’ out.But Etsy = swinging for the fences on the 1-1 pitch… maybe even with a runner on base already. If you miss, you’ll take a strike and be down in the count – but not out, or maybe you just fly-out and advance the runner…
the king of the sports analogy – reece pachecoi like your mountain man beard by the wayi barely recognized you the other day at Feld’s thingi think you should add some dreadlocks and go for the ronny turiaf look
haha! i went to the Celtics v. Knicks game last night and my sister jokedthat i should be “that guy” (Ronny) for Halloween.
Not only do I think bimodal returns are not true. I think google sized returns are a largely unpredictable outlier.I once worked for a company. The internal feelings were if we executed like hell the company might be worth $2B in 4 years. 24 months later the company was worth about $100B. The company has since come back to earth, it generates a lot of revenue and now is worth a lot, but much closer to 2B than 100B.No one predicted that. Anyone who says they did is full of it.
On the scale from bust to 100x what is the minimum return that is attractive? Obviously you don’t know this going in, but is there a benchmark that, if achieved, you think “we’d do that again”?
10x going inOnce we are in, the acceptable return changes based on the cards that we get
Great post Fred. A few thoughts …Isn’t bi-modal a curve with two peaks? You tend to see dual peaks in pricing schemes where you can maximize revenue at a higher volume lower price or higher price lower volume with a trough in the middle. When I hear the drum beat of 100x or bust I look at this as a “binary” outcome – Hit or Fail. Either way, certainly a foolish way to look at life. Championships are built on singles and doubles, blocking and tackling … it all comes down to execution. From a portfolio perspective, it is getting all of your mid tier companies to perform well that makes the fund, which is your point and I think its great to call it out.With a lot of buzz around super angels and exiting at smaller valuations, do you think that the trend is more towards the big or more towards the small?
So it’s trimodal?(based on the tiered description you provided.)
This is the Long Tail of Investing. Investors make big money at the head of the tail, but it is the value created in the mid tail and the end of the tail that dictates the overall return.
Nice analysis. Thanks. One question: how would this change if you only did initial rounds and not follow ons? (Generally hear that most VCs make best returns from doubling down on winners but haven’t seen data on this).Personally I think there are times to be valuation sensitive and times not to. When I was working in VC we passed on 3 huge winners due to small differences in price. One smart VC told me that when you are betting on the leader in an interesting space you should be less valuation sensitive. Also when companies look like they will be hugely volatile in their outcomes. Obviously in general valuation does matter though.
I don’t like to lose deals over price if we want to win them. It reallycomes down to conviction. When you have it, you should win the dealWe may run the first round only numbers. Could be interesting
You and other VC’s don’t like to loose deals.But IMHO I look at this wholly from a gambling perspective.I look at VC’s as gamblers who who gamble with the odds in their favor. Rarely do you come across a VC who gambled with the odds stacked up against them. So you try and spread the risk around and based on your power curve. Investing in the web services sector allows you and others to spread the risk more effectively as the capital requirements are lowered. I am not sure this model works for VC’s involved in the areas wherein capital requirements are higher say in semiconductors, life sciences and sustainable energy fields.It would be interesting to see the distribution curve for those VC’s.
Awesome suggestion, Chris.I ran the first round numbers this morning, and our returns look (generally) like a power law curve as well.I also looked at the UTIMCO to get a better handle on what returns in the venture and PE industries look like, and that also gets you to a power law curve.More info, including graphs, here: http://bit.ly/9AUKIJ
Thanks, that was extremely helpful
nifty! appreciate the charts in response to Chris’ curiosity.Now if only one could tell where companies would fall on that xaxis at the beginning 😉
Definitely. Figuring out where each investment is on the curve is the $1,000,000 question for venture fund managers, I think. Figuring out where each fund lies is the $1,000,000 question for LPs.
great convo on VC return distribution on AVC today. we ran some numbers in response to @cdixon’s question on the return distribution of initial rounds. here are the results http://bit.ly/a7gnl2
Not surprising. Anyone who has a return distribution like USV is going to have an initial distribution like the one in this link. You need winners at first to have a chance to feed the winners – and you are likely to get the best deal – or the greatest % ownership in the initial round. It is a rare big winner that allows you another bite at the apple – with a big down round before taking off again.
Christina, great work. So Fred/Christina, what is the conclusion?The full graph is shifted to the right (better overall returns) vs. the initial-round only graph?Followons help, but not that much?(BTW Christina, your twitter handle is really hard to find, not on your home page, about me, or LinkedIn… maybe that’s on purpose. found it on google tho)
Without knowing the $ amounts invested per deal it is hard to come to any concrete conclusion – it is the missing factor here – but my guess is that the basic conclusion is that follow ons certainly help a lot in the overall return profile of a fund – but getting the initial investment correct is by far and away the most important thing.
follow ons help a lot because that is where we get into the capital allocation game
Thought this is a neat exhibit of what Google is doing with Caffeine. If you do a google search for ‘startups bimodal investing,’ 8 of the top 10 results are either Avc.com and this post or this post on a syndicated site and the post is only 2 and 1/2 hours old.http://www.google.com/search?q=startups+bimodal+investing
I know its not a maths blog but you (or they) don’t mean bimodal, that is where your graph has two peaks with a dip in the middle.
Fred, is it fair to say that in the web/tech world you either make it big with very large popular traction (i.e., Foursquare, Twitter, in reverse order) or otherwise you had better have a solid business model with revenues and all those other old-fashioned things? If so, I wonder if the middle of the curve that you’re describing is more likely to be of that second sort… And if that’s the case, I wonder if your post touches on a whole variety of other issues related to the state of the industry.
i would argue that eventually you need a solid business modelif you don’t then you are playing with fire
You should link up the interim valuations section with your Valuation Blues post (http://www.avc.com/a_vc/200… to further the understanding.
yup, that is how the valuations in the distribution curve are arrived ati am still not happy about FAS157i think it sucks
Agreed. Next fund going to opt out of that? Or do you think investors would never go for anything but clean GAAP audit?
i think we’ll stick with gaapbut you got me thinking
I think instead of putting each company on the X-axis if you would have put number of companies against returns in multiples (1x to 30x) then it would have form the curve you are talking.Number of companies in Y axisreturns on the X axis( 0x to 30x).It would have followed the standard (Gaussian or may be little distorted) of the world.
Awesome post. I’ve been saying a MUCH less eloquent version of this for the past year.Our best returns in our 2000 fund are in companies where we doubled down. The largest returned $320 million. We worked our “middle outcomes” really hard and it paid off hugely. We are firm believers in trying to be selective about deals both because we don’t want to be in too many losers and because we know there is a HUGE cost associated with helping your middle outcome companies succeed. According to Prequin this fund is the best performing fund in the country for 2000.I question the strategy of “being in every hot deal” because “hot” is defined as the entry point. Given TechCruch, blogs, Twitter and other marketing hype machines a lot appears to be hot that will turn out to be not. If you’re in too many deals I don’t know how you can seriously help the middle or bottom outcome companies unless you rely on others. And I can’t imagine a strategy of relying on others for my success.
this would be a fun topic to riff on for a while Markget you, me, and someone like minded (Feld?) and turn a camera on andtalk about this with real anecdotes for twenty or thirty minutes
Why like-minded? Get someone who disagrees with you (Ron Conway? Paul Graham?)That’ll be more fun.
Yes please. And I will happily host on Mediaite!But also – you wrote of deals, not people. And it’s the people behind the startup who make or break with their judgment, team management, ability to pivot and psychotic workaholism. So I would think that an element of this discussion would have to be how investors can improve their odds by working with good people who know how to take – and mitigate – risk. (Illuminate Ventures released a whitepaper addressing that factor: http://www.illuminate.com/w… – yes okay it’s *that* whitepaper.) But either way on a more general level – while it’s hard to predict what will be the next Google, it’s easier to predict if a company will succeed when you know it’s being led by a rock star. So I’d guess that would factor into the above discussion, and I look forward to watching it when it’s been duly recorded!
Yes, please hound them and do it…Disqus should have a feature where you One-Click to “Challenge X-commentator to an instant video debate”. X-Person accepts and a 5 min video is recorded and thumbnail re-inserted back inside the comments.
Love to. Maybe when I’m in NY and dial in Brad on Skype?
SUCH a great post, Fred. Thanks for writing this. It seems so few firms, particularly the more recently created ones, really follow this prescription for success.
they willyou learn the venture business in two ways:1) get mentored by experienced VCs2) learn the hard wayi did both as have youeither way, you learn this stuff eventually
What about this graph* make weight=return instead of weight=1* make a series of frequency bandsNow it looks like tri-modal:https://spreadsheets.google…
william made the same observation in another commenti’ve always said 1/3 will fail, 1/3 will succeed but you’ll wish youdidn’t make the investment, and 1/3 will outperformare those the three modes?
“1/3rd will succeed but you’ll wish you didn’t make the investment”If you take the mantra “it is what it is” and ‘I have a greater god than money” then even for you this is a sad commentary.You’ve pushed the ball that is society up the hill, but only have seeds and regret.
Not sure what to make of this comment – if I had to guess at what Fred means by the “1/3 will succeed but you wish you didn’t make the investment” comment I would say this: as an investor, you always want more of your winners and fewer of you sub-optimal investments – that is ultimately how you are judged. So my guess is that as you look at all the hard work you put into an investment – and you end up with a 2X return after 7 years of ownership for an average 10.4% return to your investors before fees and incentives – you will look at that as a successful investment, but not one that met your standard in terms of time spent or return generated relative to your overall portfolio.
Harry, its kind of like the Moulton Barn. Sometimes you have to appreciate something for its beauty rather than purely how much its worth. When it was built there wasn’t a worry for how much it was worth….it was an endeavor to build something……and it endures.
True enough – but now you are bifurcating the purely financial (which iswhat LP’s pay Fred for) from the overall. I’m betting that if Etsy turns outto be a slight winner – Fred and team will be very proud of of them and whatthey have accomplished – and at the same time – they will wish they hadallocated that extra capital into Twitter or Zynga.
I certainly don’t want to get in a pissing war, but its not purely financial. If Fred didn’t have a great reputation by doing things that weren’t based purely on the financial he wouldn’t get the best deals…so long term yes its best for LPs but short term no.My point is the only negative comment was reserved for those in the middle…..that’s why people think its bimodal.
Actually, I think people think it is bi-modal for a completely differentreason. Companies like Google and Facebook, Zynga, Groupon exhibit anextreme example of survivorship bias. IF you look at each company that wasventure funded individually – of course there are massive winners – and intoday’s world they get written about, blogged about, made movies aboutetc… So people look and see massive riches.They also see very high profile failures – like pets.com or webvan – andthose get written about.what never gets mentioned is the thousands of companies in the middle who goabout life quietly. some go bust – some do quite well – but no one knowsabout them. People don’t do speeches about how they hit a triple – they talkabout “thunder lizards” and so the perception becomes reality.BTW: I don’t believe that Fred thinks completely financially – and I am surehis reputation is earned by doing more for companies than we know as readers- however, I do think he wrote this post from a complete financialstandpoint. But I get your point completely.
On this we agree completely…..100%.
that’s it harrythanks for clarifyingi’ll add one more thingthe middle 1/3 is where most of the work is done in the post investment management part of our job
Regarding the spray and prey approach … I am not sure that this is what the Super Angels of the world are doing. I am not sitting with them individually, but it appears that they ‘are’ being selective, that they are “trying” to choose companies that have a good idea, a smart leader, a market that appears to be large, with a solution that addresses a real problem. This isn’t every start up. By making many small bets, these investors are creating an index fund of sorts, rather than trying to pick out just the big winners. And if these small businesses sell for $5 MM or $10 MM then everyone wins. It makes it easier for there to be more winning deals, not the binary outcome that some have tried to pin on Start up outcomes.Where the model seems to make the most sense is when they get to double down on the high flyers. By being early, they can see the numbers from the inside and get a “working knowledge” of the entrepreneur which is probably the best vantage point to decide whether or not to put in a follow on investment.
following your winners is so key
I agree Alex. Curious at the how the curve differs depending on investing stage which I think is a fundamental piece of this argument. Need someone doing this to get companies off the ground so you have something to look at don’t you?
Not sure if this in direct response to the talk Ron Conway gave at Stanford the other day (see link: http://bit.ly/b4FoZw), but for early-stage deals don’t you have to believe that you are betting on the winner? Therefore, even though in actuality returns aren’t bimodal, shouldn’t your investment thesis be more centralized around whether or not this is the winner in the space versus is this the right price?
Conway’s reference to a binary outcome for these companies was not about fail and return nothing … he was talking about the winners in a specific space. Google has won search. Facebook has won Social Networking. In the context of these victories, I think this more similar to the winner of the super bowl. In 2009, the Colts had the best record in football, but they were not the winner. The Saints were the winner. That is not to say that the Colts didn’t have a successful season, they did. But there was only one winner.Conway and others are looking for the companies that are going to win in their category. But 3 years into a deal, they are going to work to maximize the return on the middle investments too. That is what drives the overall return for the portfolio.
it is way more complicated than thatron’s view is biased because he is “the only one” that i refer to inmy blog post
Also, are you saying that USV returns are representative of the entire VC industry?(I wonder if CrunchBase has enough data to generate a bigger sample? Of course it’s probably biased (meaning non-representative) in its contents, too.)
no i am not saying that
Awesome post. For some danish guy just getting interested in the VC area, this is really great.In the summer I took a little introduction course on the venture capital topic. We were actually told about the bimodal returns; either you make great returns or everything is lost.Sure, its a nice simple model for the understanding purposes, however realising its not like that, even better. Thanks 🙂
Fred, I love this post. I have been a bit puzzled by all this talk about doing every deal because it just doesn’t make sense from the other areas I’ve studied in investing.I’d love to see a post where you talk about how you avoid those bad deals.
I love your posts Fred. You and gabrielweinberg.com are the only two blogs I follow religiously and I also had the great pleasure of seeing you in Toronto. But I have to disagree with this post. (I did a full writeup on my blog, but it would seem self interested to linkbait it here). Basically my argument boils down to this: If you convert the returns to annualized percentages your own data set is bimodal (assuming some value for bankrupt companies due to tax writeoffs). Paul Graham is using hyperbole because obviously you don’t NEED a Google, but he is basically right that software based startups are usually a you-made-it vs you-didn’t trade off.
Thanks for posting, Zach. It’s a quality post, so I’ll link it here: http://zachaysan.tumblr.com…
Zach, that is a great post and you have a great blog … you should allow comments.
yes!ideally disqus comments
Paul Graham contradicts himself in that he also says the most important thing is to get to “ramen profitable”.Many times the market doesn’t move as fast as you think. Everybody worries about missing the market but sometimes its not about missing it, its about waiting for it to catch up.
It’s knowing when the market is ready and how it moves. It seems to be a partial guessing game.
I don’t think he is contradicting himself. He is saying if you get to ramen profitable, you have a better chance of avoiding the bimodal (his words) outcome that most start ups face of either making it or not. I think his word choice is wrong, it should be “binary outcome.” Here is his quote:”A startup that reaches ramen profitability may be more likely to succeed than not. Which is pretty exciting, considering the bimodal distribution of outcomes in startups: you either fail or make a lot of money.”http://www.paulgraham.com/r…
Well read Alex…I should have re-read his post as its hard to win an argument with somebody as well written as Paul.However I’m writing from the perspective of just an entrepreneur, not an investor. This is where I think people get confused, when they only read from the vc investor point of view.Many peoples flowcharts look like this:Idea –>Raise VC Money –> WinIdea –>Don’t Raise VC Money –> FailYou can pull out the VC line and again this is what people are saying is reality.I assure you this is not reality in the majority case. However, if you read my thread below Fred’s you can grow and have a great company but from a VC perspective be a fail .Edited because my flowchart didn’t work.
Yours is an interesting post Zach. You should not worry too much about linking to it directly.Fred has many constraints when he writes, and one of these – like other normal people – is modesty.What he won’t say is this: “If you *really* know what you’re doing you can still make good returns without a Google”. That doesn’t necessarily imply that returns are not, at least a bit, bimodal-ish (as you correctly point out, in software and web services it’s all about your marginal returns). Nor does it imply that a 100 bagger is not extremely welcome in any portfolio (as Fred himself confirms in the comments).It simply means you still can outperform your asset class without owning a wonderstock.It’s the difference between playing the lottery and playing the horses.
hi Zach. i went and read your post (via hacker news)a couple quibbles1) you can’t convert to IRRs without knowing the year of the investment. some were made in 2004, some were made in 2005, etc, etc. we are still investing into that fund2) if you throw out the top two investments, you will still have a terrific fund3) all the zero IRRs will turn into something else. and i bet at least one or two will turn into >50% IRRs
Falsim?! What the heck is that? It’s not a word, its not English……
Hi, Fred. I think the points the “bimodal returns” people make are often bunk.However, your graph doesn’t demonstrate your point. You need to graph on a frequency distribution. If you use a histogram with particular bucketing, this will indeed look bimodal. With other bucketing, it won’t.That doesn’t matter, though. Businesses aren’t some natural phenomenon. People make them. VCs help make them. Our VC investor was in our office just this week and it was great. He asked questions. He encouraged us. He pointed things out. He held us accountable. Not to him, really, but to our own vision.A VC who gets into every deal is saying that the only thing they bring to the table is cash. Which might be true, but I’m not sure why they would admit that. As an entrepreneur, I’m much more interested in the smarts than the money.
I tend to disagree but you may be right. So as usual, let me use History through KP’s first fund: http://www.startup-book.com…Without Genentech and Tandem, KP would still have made a 2.8x multiple. I tend to disagree becauseif you do not have a fund returner in your portfolio, it is tough; if you only have 2x returns and not 10x then it is not very attractive for your LPs, and not attractive for yourself with the carried. Still KP would have done a 2.8x multiple, not so bad, so OK you are right it is not bimodal, but it is not so attractive still. But where I am still skeptical, it is about the IRR you get on 10 years with a 2.8x multiple…
What I love about the chart linked here is that their largest investment in terms of $ in – Tandem Computer – turned out to be their 2nd largest winner. Clearly they followed on with almost 20% of their fund ultimately invested in that name. That’s good investing.
Your chart clearly illustrates the bi-modal nature of returns. A few companies generate a high return, lots of companies generate almost no returns. There are almost no companies in the middle. If the distribution of returns were more uniform, you’d see more companies generating mid-level returns.This is another flavor of complaint about VCs pushing companies to go for the home run. The perception (whether true or not) is that VCs would rather the company go for 100x return and die if it fails, than generate mediocre return on investment.I do think the situation has gotten better since after the dot-com boom.
by the time we are all said and done on this fund, there probably wont’ be any zero IRR dealsand i’l bet that the middle 1/3 of the fund, if you take out the best 1/3 and the worst 1/3 will be pretty damn good
Actually, this post shows the displined nature of investing- I can’t help but to think of Buddah’s 3rd way/middle path.Interesting, 20% of these companies hit the 10x or above threshold. Another 20% hits between 5-10x.By aiming for the middle you actually are getting pretty even returns. Eating the cake is nice- but not if it is all you eat, it seems. By aiming for the middle, you actually got a healthy balance.**This law may work for your diet too. No guarantees.
Careful — there’s a difference between “aiming for the middle” in terms of your overall portfolio and doing it on individual deals.All statistical measures are moving targets. From what I understand, USV is always aiming for >10x on any individual deal, and that’s what produces the mean return of ~9x. If you start funding companies you believe are 5x-category in addition to the >10x-category ones, you have to think that your mean will be depressed.
Aim for 10% but assume most are going to be in the 5x range- the question is how to get to 10x. It’s more about how investing modesly may help you pick up unrealized gems (it means more information about doubling down, and with whom)
Other fascinating issue- this truthfully explains the fascination with second market. If you can pay to play, you actually can look deeper into the valuations and have it react as a full public market and try making money at one thinks it is a peak price.
I think the proper statistical term is that the distribution is skewed as opposed to bimodal.Not much argument about that… so what does it mean?I was surprised to see PG argue it means you should worry more about selection than valuation.For one thing, being more selective is the opposite of the shotgun approach YC takes, funding a lot of startups very early. Being selective is more like the rifle VC approach.For another, at any valuation, if you double the valuation, you reduce your returns by more than half. Put another way, higher valuations force you to be more selective. To get the same return you would get by halving the valuation, you would have to lop off more than half the distribution.Any investing activity is about managing risk, getting in if and only if you have an edge, and doing your homework.(Edited to try to turn it into a coherent thought)
Poker doesn’t have to be very skewed, depending on the stack sizes and the particular game being played.
You edited your post and now my nitpick refers to nothing and makes me look silly. On the other hand, it was pedantic — so I do deserve it — but since I play poker for a living, I tend to chime in whenever my subject comes up. +ev bets.
sorry! didn’t see the comment or would have left the reference… if one is sensitive about looking coherent, should edit first, post later lol.In general the poker distribution is not just shifted but skewed in favor of the better players who are winning big pots and losing small ones, no? And no-limit hold’em, a few big pots make all the difference, limit games less so, so a good player’s distribution is less skewed?It’s a complicated game, if you’re playing against players who are weak because they lay down too often, you can be +EV by bluffing more often to pick up small pots and occasionally losing a big one.But in most of my sessions the big pots make all the +/- EV, partly because in my case the EV is close to 0 lol.
In any no-limit hold’em game worth playing in, returns will be pretty skewed on a per hand basis. But alot of that is a function of (lack of) player skill and stack sizes.That’s why it’s a nitpick. If you’re game selecting well (think of that as a topdown approach) then your returns will be pretty skewed on a per hand (bottom up investing) basis. But it isn’t _necessarily_ so.
great comment as usual druce
ty, I try to post my cranky antisocial comments pseudonymously LOL 🙂
The key to understanding YC is that PG gets to invest in all those companies at sub-$500K premoney valuations.If you provide $20,000 for 5% of a company, that’s a $400K premoney.YC then turns around and markets those companies to angels and VCs for anywhere from $4-10 million premoney.That’s a 10X markup from investment to Demo Day. Given those parameters, is it any surprise that YC argues that valuation doesn’t matter?
Late to the party again here today – but all the wiser for it – as the comment section has easily as much to say as the initial post.These conversations get easily confusion. The distribution of returns shown in the post – and the subsequent analysis posted by Christina here (http://bit.ly.9AUKIJ) are all ex post data points. No one invests or thinks about investing to a power curve – or a bi-modal distribution or anything of the sort. Investors make investments thinking about risk and return – in Fred’s case he lays it out pretty clearly – he believes that he will make 10X on every $ he puts out.Statistically speaking, when you have 30 data points – or 40 or whatever the number (Brad Feld replied to one of my comments saying that in his new fund he believes they will make about 7 investments per year as a team – so over a 6 year investment period about 40 investments) These numbers are far too small to be truly statistically significant – and thus will show a great deal of leptokurtosis – or fat tails in the distribution of returns. Some will be 40X returns and some will be goose eggs – and plenty – if you are smart about generating liquidity when you can after realizing it will not be a 10X or above return will be in the mid ground.The key, as an investor – is to try and keep the zeros to a minimum – and to quickly realize the mid ground of investments – focusing all of your time and capital toward the big winners.I’m quite sure that if you take a look at not just the return per investment but the actual $ returned per investment in Fred’s graph – which is really all that matters from the standpoint of an LP or GP – the returns will be more skewed – because a smart investor you will cut off the losers and ride the winners hard.Chris Dixon’s question in the comment strand is an excellent one – basically: absent questions of valuation, are you picking winners? The corollary is then whether you are allocating capital efficiently (i.e.l the most capital to the biggest winner) and managing risk correctly (cutting off sub optimal investments and putting all of that excess capital into winners).I think that is why Ron Conway can argue that companies either win or they lose – and price does not matter all that much – because in the end – when you look at a large portfolio of investments over a long period of time – it is those winners that have been feed a greater % of your capital that really do provide all of the returns.So if you look at a typical Gaussian curve (bell curve) distribution of returns in a portfolio – and say the average portfolio in VC has yielded roughly 0% return over the past 10 years ( I think this is close to accurate) you are going to see fat tails (leptokurtosis) – as some investments are big winners and some are complete zeros.What Fred is putting in simple terms – and what most of the readers seem to grasp well is the fact that you while you are always going to have fat tails in venture (particularly with so few investments per fund – hey this isn’t high frequency trading) you can move the curve to a more positive average position (slide it to the right on the returns axis) by limiting your losses – and you can make the tails far fatter (hopefully to the right) by feeding more capital into the winners.So returns are certainly not bi-modal (and I think what people mean is binary – they are either winners or losers) – but when graphed in terms of contribution to total return – many, if not most, of the investments could well have been zeros – leading one to think of them as binary.
Great post, and it gives some hope to angels who don’t have capes. Mike Maples recently said that if you are not in one of the top 15 deals each year (the “Thunder Lizards”), then you don’t have a business model. I would like to believe that is not true, since I probably will not ever see one of those deals. The USV portfolio does have Thunder Lizards in it, though, so you have to trim the big winners out to see if “ordinary” investors have a shot at making money. That is not clear from the data, but if you drop the top 2 or 3 deals, it looks like you are going to be scrapping to make even 3x over time. There is a lot of talk about whether angel investing is financially sensible for most, or whether it is just akin to buying a vineyard–where you have fun as you lose money. For that matter, is there a long term sustainable business model for institutional VCs not in the top quartile? I hope the answer is yes, as disruptive innovation makes the world a better place, and strong early stage capital markets are a key component of the disruption ecosystem.
I don’t mean to be negative. Maybe this is just a case of me misunderstanding your data.I ran a spreadsheet on your numbers above. The top 5 average 13, the bottom 5 average .6 and even the middle 11 (11? Did you know that company D was listed twice?) only averaged 2.2…So if you argue that it’s not ‘bimodal’, is it safe to say that it’s tri-modal: “The Very Very Good, the Bad, and the Ugly”…But if you remove the top 5 and average the bottom 16, you get 1.68 – which looks pretty bimodal to me. Your good companies carry the weight of your bad companies.Clearly you can’t (shouldn’t) “do every deal”, but I think knowing that it’s bimodal is a good thing. The difference here is in the definition of “bimodal”. In your case, bimodal is not “on or off”, it is “Big Win or Mediocre Win”. It’s still a Win, either way. Congrats!Take care. mjl
Where is JLM- He went bimodal on us?
Been in San Francisco kissing frogs. Just got back. It’s cold in San Francisco.However, I am bi-modal. LOLUhhh, I think I am.
lmho I knew this comment would have you pop your head above the water 🙂
ROFL …That sounds like quite a complicated life.
I’m not sure your interpretation of the Power Law is correct and in fact, you may be making the exact point of a bimodal return.Pareto distributions are a form of a Power Laws which that 20 percent of the instances generate 80 percent of the value, which is quite bimodal in terms of return.The classification of the distribution depends on the specific exponent of the curve, which to compute you need to take the logarithm of each axis and then compute the slope. Further, according to Wikipedia, “although power-law relations are attractive for many theoretical reasons, demonstrating that data do indeed follow a power-law relation requires more than simply fitting such a model to the data.”I encourage you to rethink whether you have enough data or expertise to draw any such conclusion.
forget the words power law and bimodalthat’s not really the issuethe issue is whether or not you have to swing for the fences in VC to winmy experience with the fund I graphed and every other fund i worked on in the past 25 years is that you can do very well without any 10xs in your portfolio
Interesting … I think that all VCs are always swinging for the fences. The analogy that seems to apply here is should you swing at every pitch, or should you wait for a fast ball on the inside middle of the plate that is going 88 miles an hour. If you are patient, and wait for the right pitch, in the right spot, (product market fit with the right team) you will be rewarded.
Well, if we forget the words power law and bimodal, then I suggest you change the title of your post something closer to your point about swinging for the fences and less about something statistical with empirical underpinnings.It’s irresponsible to promulgate a fallacious meme by saying it’s a power law and a bimodal outcome. You should issue a retraction.
a fallacious meme?you are kidding me, right?
Tortoises, slow as they may be, live an awfully long time!
Binary Investments, The Middle Kingdom, And Super Exits http://goo.gl/fb/C8e6l
One additional point needs to be added: success on middling and even loser investments often translates into participation in blockbuster investments.I’ve analogized from VC work to my work as a contingent fee lawyer before (see http://bit.ly/ndBId ). In my practice, there is little better “marketing” or “personal branding” you can do than simply doing a good job on a case, and I imagine the same is true on your deals. You never know where the next blockbuster is coming from; by doing a great job on non-blockbusters, you dramatically increase the likelihood of being invited to the blockbusters.For me, I’ve lost count of the number of times I simply worked hard and got a decent result out of a middling case referred to me by an attorney, after which that same attorney referred me a far more lucrative case. I bet the same is true in VC deals and the relationship between founders and investors; having a reputation for doing a good job — a reputation formed primarily on these middling and loser deals, since nobody has blockbusters all the time — goes a long way.
Fred great great post. This is why as an Angel I always insist on pro-rata rights – so I can “bet heavily on the winners.” A very prominent VC has told me that an Angel should always waive pro-rata rights on an up round which for the life of me I cannot figure out why I would do that. If I do get into an angel round, where I took the most risk, and the company progresses, I absolutely want to bet more cash if I’ve already negotiated and have the right to do so. Am I missing something here on that logic?Also why I don’t do a ton of deals, but more like 2-3 a year where I can be heavily involved. Obviously with Appolicious that means I am doing very very few right now given my time commitments.
the pro-rata right is an optionit is worth a lot in the right dealyou are so right to insist on it
Wowwwwwwwwww….What is it bimodal or trimodal……and I dont understand how you’ve fired it up fredwilson..
So what makes for hot companies. And is it better to have lots of light or generate lots of heat? (if your company is the flame, so to speak)
you could have lots of customers and not have lots of press. You workseamlessly, and you do some quiet function rather than one that becomesmedia centric. (actually, I always liked that approach- be useful, not amedia circus)
I think people confuse “hot” with “unavailable”.
bang on Charlie!