How AI Is Guiding Venture Capital To Startups

Increasing scope, efficiency and productivity. A question is whether this will change the venture capital cliche about investing not in companies but in people. JL

Maija Palmer reports in the Financial Times:

One of the biggest challenges for venture capital companies is finding investment targets. Machine learning and predictive analytics are starting to transform how an investor puts a portfolio together. As well as giving opportunities to a different set of start-ups, machine learning is likely to change the structure of the VC industry. Data scientists and engineers are vital to the business. “What used to be a handcrafted job has become significantly scalable.You become 10 times more productive.”

One of the biggest challenges for venture capital companies is finding interesting investment targets before anyone else. It is often a laborious, travel-intensive job. But machine learning and predictive analytics are starting to transform how an investor puts a portfolio together.

“My job used to be about getting on a plane once a week and going to a different European city to try to find people who were doing interesting things,” says Roberto Bonanzinga, co-founder of InReach Ventures and previously a partner at Balderton Capital, a UK-based VC firm which invests primarily in early-stage European technology companies.

It was inefficient, he says. “I would look at 50 companies a day, maybe 1,500 a month, and of those maybe 100 would get to the next level. We would do one deal a month.”

Often, Mr Bonanzinga adds, promising companies would fall under the radar because they were not actively looking for money. Unless start-up founders were well connected, or in a tech hub such as London or Silicon Valley, investors would have little chance of discovering them.

Mr Bonanzinga thought he could combine internet data and machine learning to do a better job of ferreting out prospects. It took two years and £5m in investment for InReach Ventures to create the software, which has so far trawled through 95,000 European start-ups, picking out 2,000 that Mr Bonanzinga might be interested in.

The software determines this based on the people they are hiring, the products they are developing and the traffic on their website, among other things. For example, InReach identified Oberlo, a Lithuanian start-up, as an investment target because it was advertising for engineers to solve a particular type of ecommerce problem. “We did a deal before any other VC firm in Europe even knew they existed,” Mr Bonanzinga says.

“What used to be a handcrafted job has become significantly scalable,” Mr Bonanzinga says. “You become 10 times more productive.” InReach Ventures has invested in seven companies so far, and while it is still too early to say how well the portfolio will perform, Mr Bonanzinga has already had one exit, selling Oberlo to Shopify, a Canadian ecommerce company, just 12 months after investing.

SignalFire, based in San Francisco, was one of the first VC companies to move to this kind of data-driven model. Founder Chris Farmer started applying data to venture capital around 2007, at a previous company, using basic algorithms to track factors such as how well products were performing in Apple’s App Store. He grew interested in building a more complex system that could track companies in a more comprehensive way and thus founded SignalFire in 2013.

It took eight years and tens of millions of dollars to build what Mr Farmer calls a “mini-Google”. The software now tracks 8m start-ups across the world, drawing on sources from sales data to academic publications to financial filings. Companies that are outperforming or doing something notable are flagged up on a dashboard. SignalFire can then deploy some of its $375m under management.

Ten years ago, Mr Farmer says, the project would have been impossible. “These kinds of data storage and processing capabilities weren’t available. We needed the kind of computer power and storage that was only available in the bigger consumer internet companies.”

Now, however, even smaller companies can crunch data at a large scale, thanks to database tools such as Hadoop and Apache Spark, and the possibility of renting cheap server capacity from Amazon Web Services.

Like Mr Bonanzinga, Mr Farmer says the system is helping him discover companies he would not have seen before. “There is a broader geographic scope to where our portfolio companies come from. We backed a company from Romania that we never would have seen otherwise. We’ve passed on some very well-connected founders and we have gone with some first-time founders. It doesn’t entirely eliminate bias but it does make it more of a meritocracy. It makes you take a second look,” he says. It is still too early to judge how well the portfolio will perform, however.

Aaron Joyce, co-founder of Aibl Tech, a Stockholm-based start-up, has benefited from the data-driven approach. Aibl, which helps companies analyse customer data, was discovered by Mr Bonanzinga only a few months after it had been started.

“We had started approaching a few local investors for funding, but we were not having much luck getting through, just talking to the junior guys. Then Roberto [Bonanzinga] emailed us. I was suspicious at the beginning, here was this partner contacting me. I would never have thought we had a chance of raising money with someone of his calibre,” says Mr Joyce. “It is a much more meritocratic way of investing. It’s not who you know or where you went to school.”

Andreas Thorstensson, partner at Stockholm-based EQT Ventures, says around 30 per cent of his investment decisions now come through a data-analysis platform he has built, called Motherbrain, which monitors around 2m companies on a daily basis. Mr Thorstensson says he regularly passes on investing the kinds of entrepreneurs that might typically have got funding before. “The data doesn’t lie,” he says.

As well as giving opportunities to a different set of start-ups, machine learning is likely to change the structure of the VC industry. SignalFire is built like a tech firm — data scientists and engineers are vital to the business and own shares in the company. “We don’t treat them like the IT department. They are the heart of the business,” Mr Farmer says.

Mr Bonanzinga, meanwhile, is the only investment partner at InReach, which has software and a team of data scientists doing the work that would have been done by an army of associates at a traditional VC firm.

The cost of running and maintaining the data platforms is considerable. Mr Farmer says he spends over $10m a year, while Mr Bonanzinga plans to spend at least £1m annually. “It will change how venture capital firms spend their fees. Traditionally most of it went on salaries. In the future it will go to data and computer scientists and into buying in good data sources,” says Mr Bonanzinga.

EQT Venture’s Mr Thorstensson says, however, that he does not believe machines will significantly decrease jobs across the industry. “It will enable us to spend more quality time with the companies we are investing in, instead of going to general events and doing the tedious jobs,” he says. “AI is good for filtering out the noise, but the decision to invest or not will always be about instinct at the end.”