Can Artificial Software Pick the Best Start-ups?

By Lars Tvede, Strategic Partner

When we founded Nordic Eye, I was fully aware that artificial intelligence, or ‘AI’, could work extremely well in almost any sector, but I must admit that I didn’t include venture financing on the list. Wrong! It turns out that several companies have demonstrated great success in using AI to pick the most promising among thousands of unlisted companies that had received some kind of early venture financing.

One example is Quid Inc. In 2009, Businessweek magazine asked the CEO of Quid to use their software to pick 50 unheard-of companies with particularly great prospects.[i] He did so, and later it turned out that if a venture capital fund had made a portfolio by investing in just those 50 companies, it would have had the second-best-performance of all VC funds ever.

Yes, ever. And in 2017, he repeated the task, beginning with a database of 50,000 non-listed start-up companies and narrowing them down to 5,000 by screening for 1) young, low-valuation start-ups that had raised money quickly from a set of top performing VC firms, 2) higher-valuation companies that showed clear commercial traction already, and 3) companies operating in the most promising business segments, identified by studying what types of businesses the most successful VC companies had invested most money in recently.

The CEO then narrowed the list further to 200 companies by ranking how long it took between each funding round (best if a short time and preferably less than 9 months between each round), how many rounds were raised (better if many), and how much money was raised. And then he took it down to 50 by looking at, among other things, whether the founders had worked together before (a positive element) and which schools they had attended (good if top universities). Other applied indicators included momentum in press coverage (like such tracked in Google Trends), quality of new hires (LinkedIn) and whether early high-profile investors joined subsequent rounds.

Another research company, CBInsights, has invested nine years and over a million dollars since 2010 to develop an AI system called Mosaic, which tries to pick winners, and potential unicorns in particular, among unlisted start-ups by tracking indicators regarding their markets, financing and momentum. In 2015, New York Times asked them to find 50 future unicorns and from the list Mosaic came up with, 24 companies — or 48% — have indeed become unicorns since then.

Of course, the fact that a computer can pick likely winners doesn’t mean that any investor can get onboard these stellar companies early on. You often need a lot of meddle to get invited into the financing rounds of such companies. However, it’s clear that AI can help you decide on investments. Also when deciding to invest in secondaries of unlisted companies, such as ‘friends and family shares’ or employee shares. So, let me conclude by saying that we are following this space.