# A Tale of Two Squirrels: The Not So Simple Math on Venture Portfolio Size

By now most VCs are familiar with Dave McClure’s large portfolio theory. In short, he believes that at seed stage, it doesn’t make sense to have a portfolio with fewer than 50–100 companies, because venture returns depend on outliers and you need a big enough portfolio to consistently capture them.

In the post, he outlines a range of typical outcomes for a large batch of seed-stage investments. You can see some variation of this trend in most published venture returns data such as Crunchbase or PitchBook.

These are large ranges (because there’s a lot of randomness in startups), and depending on where you end up in these ranges, you could make or lose a lot of money. Most investors prefer a bit more certainty.

Thankfully, statisticians have invented something called a Monte Carlo analysis, popularized by Nate Silver of 538 fame, to simulate the impact of this randomness by simulating a large range of possible outcomes.

My friend Yannick Roux (@yanroux, Blog), a London-based VC, kindly built a Monte Carlo simulation in Excel to help me model the range of possible outcomes for venture portfolios.

### The “Blind Squirrel” Portfolio

We have an expression “Even a blind squirrel finds a nut every once in a while.”

In other words, any VC with decent deal flow and a reasonable selection process, if they write enough checks, should eventually pick a winner. I’m not saying that’s a good way to invest, but let’s do the math.

Working with Yannick’s model, I plugged in some assumptions from the *middle *of the ranges above. This represents the “average” venture investor, hence with outcomes that fall in the middle of these ranges.

Then the Monte Carlo engine quickly ran through 10,000 simulated portfolios and listed the outcomes.

#### I repeated this five times.

Changing *only *the portfolio size each time, and *leaving all other variables constant* (such as fund size average investment per company per outcome)*.*

#### These are the results:

As you can see, the results for the three largest portfolios are almost identical, but the results for the 20- and 50-company portfolio are worse. That’s because, in this model, we’re only expecting big (e.g. >50X returns) winners to occur 1% of the time.

In a portfolio of 20 companies, 1% of 20 is, more often than not, zero. But in a portfolio of 200+ companies, you could pretty reliably see a couple 50X outcomes in each iteration of the portfolio.

Here’s a frequency distribution showing the breakdown of return multiples 10,000 simulated portfolios of 20 companies vs. 200 companies. It’s a bit easier to visualise this way.

**We’re Not Average! Enter the Super Squirrel.**

The “blind squirrel” portfolio was designed to match the outcomes of the venture universe in-general.

These are the middle of our ranges — and a median return of 3.18X before fees and after a 10-year lock-up isn’t terrible.

But we should hope that a well-known venture fund with a recognized brand and a large team of experienced partners would attract better than average quality companies, and be better than average at picking and supporting winners.

So I re-ran the model with input assumptions towards the higher end of our ranges, a different picture emerged: 20 companies is still not a great portfolio. But 200 companies can get you better than 4X before fees.

Now these are much better returns. And in this model, the impact of portfolio size becomes much more pronounced. That’s because payoffs in venture are asymmetrical, meaning the impact of the losers (e.g. you lose 1X your investment) remains the same regardless of how amazing you are.

The impact of thewinnersis exaggerated for Super Squirrel VCs, because there are more bigger winners in Super Squirrel’s portfolio.

**What about the 50 company portfolio?**

As you saw above, the 50 company portfolio doesn’t do badly. The top quartile returns more than 6.34X, which is better than the 100 company portfolio.

#### But it carries a lot more risk, and you can see that in the shape of the curves:

Notice that second gray hump on the right? That squirrel looks more like a camel! (a bi-modal, or *Bactrian* camel at that) That’s because your chance of hitting a “big winner” (50X — 100X) is about 1%.

In a 50 company portfolio, that will happen about half of the time. So the fund outcomes in the hump on the right have that *one* big winner in them, and the ones on the left don’t.

But in those great outcomes, it’s really down to that *one big winner.* If I re-run the Super Squirrel model and *remove* the top performing company in each scenario, then that whole second hump goes away.

Notice below, the top quartile return for the 50 company fund drops by 49%, but the top quartile return for the 200 company fund only loses 20%.

#### Now imagine you’re the manager of the 50 company fund. You’re six years in and you have that one company — late stage, growing fast, looking good.

- What if they “only” sell for $200M and you get crushed under a stack of liquidation preferences? (Like: your bag of winter nuts.)
- What if Amazon goes after them? (And wants the nuts, too.)
- What if a similar company tries to IPO and it’s a disaster?
- What if the Wunderkind founder gets hit by a bus? (Crossing the road.)
- Or suppose that company does well and you decide to raise another fund. (Harvesting your nuts from the same tree.)
- Then you’ve got to convince your LPs that lightning will strike twice, and you’ll find another big winner again in your next fund.

You explain that even though nearly half your returns from your last fund came from a single company, you’re sure you can pull that rabbit out of that hat again. (Or more nuts from the same bag — you get the idea.)

#### These questions will haunt your dreams.

### But We’re Not Squirrels!

It’s true, most VCs will tell you their investments are *not random*. They will claim they are able to access and carefully select the best companies in which to invest.

As an LP in a 20 company fund, all you need to do is pick a fund manager who is consistently able to attract and consistently select the top 5% of seed stage startups.

But remember, if you have someone who can consistently *select* the top 5% of *publicly-traded *equities year after year, you have Charlie Munger of Berkshire Hathaway.

### That’s not a simple task!

And it’s theoretically easier to identify good companies in public markets, where you have decades of historical data, competitive data and armies of analysts poring over every available scrap of information.

So the person who can consistently pick the top 5% of seed-stage startups is *much smarter than Charlie Munger. *(When you meet that person, please please please send her my way!)

### But what about Sequoia Capital? Kleiner Perkins? Andreessen Horowitz?

Concentrated portfolios have been the venture game for the last few decades:

Most institutional investors allocating into venture capital (representing at best a single digit percent of their asset allocation) have been fighting for allocations into a very small number of top-decile fund managers.

#### These are typically based on Sand Hill Road.

How do we explain all those famous funds with concentrated portfolios that have done so well? It’s true, a few fund managers have done a great job of landing their outsized share of big winners fund after fund. So this must be possible.

We believe, the main difference is that these people are investing in later stages (Series A onwards). At later stages, a more concentrated portfolio might make more sense, as a higher proportion of your investments should be “winners” and fewer will go to zero.

In that case, your ability as a fund manager depends less on your ability to “select” winners and more on your ability to get into the best deals.

That said, although companies in later stages may be 10X further along in traction and the likelihood of success may have improved somewhat vs. the prior stage, their pre-money valuations may have increased much more. (Our typical entry point on valuation for seed-stage is about $2.5M pre-money, whereas a Series A might start at $15-$20M pre-money and a Series B might be at $40M-$50M pre-money).

### Finally

Entering at higher valuations means you need to exit at higher valuations to see a comparable multiple. For example, to get an Amazing (50X) outcome on an investment at $50M pre-money requires getting more than $2.5B exit valuation, whereas to get such an outcome on an investment at $2.5M pre-money requires getting only a $125M exit valuation (before dilution to simplify the math).

The net of all of this is that, in our opinion, later-stage investing may have a worse risk-adjusted return profile than seed-stage investments, especially for fund managers who do not have the same kind of branding and deal access as the Legends of Sand Hill Road.

### How Big Should My Portfolio Be?

We believe, if you’re 1) investing* at seed stage*, and 2) you are an *average* investor (in terms of deal flow & selection experience), and 3) *your main goal is maximizing financial returns*, you’d want a minimum of 100 companies to get a decent shot at a 3X gross return.

If you’re a *really good *investor, 50 companies might be enough. But if your one big winner doesn’t deliver hugely… that’s the risk. So, in our opinion, if you want consistent outperformance and *unicorn failure insurance* you should aim for 200–500 companies.

### This is Not Revolutionary

I’m not the first person in the history of finance to suggest that diversification might be a good thing. And 500 Startups isn’t the first early-stage fund to favor a large portfolio. (That was Y Combinator, or Ron Conway before them). But we keep having this debate for some reason.

So I wanted to unpack the math a bit.

### Acknowledgements

None of this math would have been possible without the portfolio Monte Carlo simulation engine developed by Yannick Roux, who also reviewed and improved drafts of the post.

Plus great inspiration from @twentyminutevc in his great discussion with Josh Breinlinger and the ensuing tweetstorm.

Many thanks to Dave McClure, Aman Verjee and Eddie Thai for all the feedback on drafts & constantly prodding the math. (And Yiying Liu for photoshopping the Patagonia vests on to the venture squirrels — priceless!)If you learned anything new from this post, it was truly from the shoulders of giants on which I stand.

Legal Notice:

The statements here are merely my opinion. As the author, I believe most of what I wrote — but not all of it. It’s funny, not legal.