What venture capital can learn from data-science

By Graham Schwikkard on 30th October 2019

Equity Crowdfunding

With the majority of VCs losing money, SyndicateRoom CEO Graham Schwikkard asks whether they need a better strategy?

What venture capital can learn from data-science
Image source: Syndicate Room

Graham Schwikkard is CEO of crowdfunding platform SyndicateRoom.

I’m a self-confessed data addict. This means I come to venture capital investing from a different perspective than most VCs, who seem to be comfortable making enormous bets based on subjective assessments of the companies they invest in.

Personally, investing in a company based on whether I think it’ll do well in the future fills me with anxiety. But beyond that, this subjective approach to investing doesn’t work well for the majority of VCs.

These two charts illustrate the problem. The startup investment market grows in value consistently at between 23-28 per cent per year. 

 

Source: Syndicate Room

If you were able to invest £10,000 into all of the startups that raised seed or venture funding in 2011, your investment would now be worth around £74,000. And that’s taking into account all the failures (valued at zero), which we all know there are a lot of in startup land.

However, despite strong and consistent growth at a market level, the majority of VCs fail to replicate this performance in their portfolios. 

Source: Syndicate Room

Indeed, assuming you’d have the access to invest in all startups (which is a big assumption), you could beat all but the very best VCs simply by picking a large portfolio of investments at random. You’d just need to replicate what was going on in the overall market.

This underperformance versus the market occurs largely due to VCs picking concentrated portfolios of companies they consider to be “winners”. What causes the underperformance?

  1. Small portfolios are naturally more volatile than larger, diversified portfolios, hence the spread in IRR across different VCs.

  2. With very little information to go by when assessing seed-stage investments - and an almost infinite number of random variables affecting the future of a startup - it becomes impossible to forecast the outcome of an investment over a 10-year holding period. So, selection processes designed to pick winners are thinly veiled wishful thinking.

  3. Unlike in public markets, there is no infrastructure or systematic way to access all startup deals in an open market. At best, a VC sees the deals referred to them by their network.

There is no simple way to address the third point, but I will attempt to answer points one and two.

Diversification is a lot more powerful than you think

Let’s go back to our cohort of startups that raised funding in 2011 and pretend we were investing in them when they first started. How many investments should you make? My team and I ran monte carlo simulations looking at portfolios with 10, 20, 30, 40… up to 80 deals.

Source: Syndicate Room

Astoundingly - and perhaps contrary to conventional wisdom - the average returns of a portfolio increased with portfolio size while volatility decreased. The reason for increasing returns is quite simple: 

  • If 90 per cent of the returns are concentrated in the top 5 per cent of “blockbuster” deals

  • Then the larger the portfolio the greater chance of backing a blockbuster deal

Many VCs would argue that their access to deal flow and selection criteria would beat randomness, but the data, unfortunately, suggest otherwise for all but the VC elite.

Based on that finding, I’d argue that VCs really ought to be making at least 50 investments per year to maximise returns and minimise risk.

Access to deals

The elephant in the room is that my simulations have assumed perfect access to all startup deals, which is not replicable in reality.

So, how can you systematically access the best deals, or at least a sample that is representative of the whole market?

I would argue that it’s not possible to tell at seed stage what the “best” deals look like, but you can certainly identify which investors have a track record of backing them in the past. 

Obviously, past performance isn’t a reliable indicator of future results, though finding these “super nodes” in the startup networks might help identify deal flow which is likely to be at least representative of the market. Remember, in this case “average deal flow” still delivers growth of between 23-28 per ent per year and would beat most VCs.

To see how that would work, my team and I looked up on Companies House all the angel investors who were early backers of companies in a well-known list of fast-growing companies. The search identified 573 angels.

We then ran an analysis on each investor’s entire portfolio, applying additional filters including portfolio growth above 28 per cent per year.

This produced a short-list of 93 active angel investors who between them had backed pretty much every recent success story. Investing in all their deals between 2014-18 would have yielded portfolio growth of 42 per cent a year. Fewer than 7 per cent of all VCs achieve those growth rates.

So, I’d argue that if you’re an angel or VC, it’d be worth analysing who in your network is a top-performing investor and try to get in on as many of their deals as you can. 

And diversifying across deal flow from lots of top-performers is likely to work even better.

Graham Schwikkard is CEO of crowdfunding platform SyndicateRoom.

AltFi Toronto Summit 2019

AltFi is coming back to Toronto following on from a highly successful event last year. We'll again be bringing a high profile international showcase of speakers from the leading fintech and alternative finance leaders.

11th December 2019