Image from a blog post by Jonathan Serfaty of Telstra Ventures, “Using Data Science to Reshape the … [+]
Telstra Ventures
Telstra Ventures is a San Francisco-based venture capital firm that invests primarily in tech firms. As its name suggests, it was originally funded by the Australia-based telecom firm Telstra, though it has other institutional investors as well. When it was founded in 2011, like most venture capital firms it invested in companies on the basis of portfolio themes, early financial results, and detailed financial analysis.
Unlike most VC firms, however, today it has incorporated data science into its mix of investing criteria. Despite the fact that VC firms often invest in startups with a strong focus on AI and data science, relatively few use data and quantitative analysis of non-financial data as a guide to whether to invest. Some VC firms, including Andreessen Horowitz, Sequoia, Signalfire, Tribe Capital, and Google Ventures, have taken steps in this direction. Some focus more on data science as a way to identify investment opportunities, while others help their portfolio companies to build AI capabilities.
What possessed the leaders of Telstra Ventures to stop using only their gut and spreadsheets? They had something of an epiphany in 2017: if the firm believed that data science capabilities were important in the companies in its portfolio, why wasn’t it using those capabilities in its own decision-making? Their first step was to recruit Jonathan Serfaty, a data scientist who had previously worked at LinkedIn.
Mark Sherman, the firm’s Managing Director, was the primary advocate for hiring Serfaty, and he explained that it’s been a gradual process to adopt the new methods. “We’re layering in more data science over time,” he said. “Now it pervades our investment process as well as the value we add to portfolio companies.” He explained that data science wouldn’t work well for all types of venture investing. It doesn’t work well, he said, in early stage investing, where “there isn’t enough digital exhaust yet.” But mid-stage companies ($1 to $20 million in revenues), which is his firm’s focus, have customers, revenues, employees—in short, they generate some data. With later investments just prior to IPOs, there is even more data to analyze.
The Data and Models
Serfaty, who is now the firm’s Head of Data Science, had to start from scratch when he arrived. The firm had no data that could be used in training models. He began to crawl the web for public information that might be useful in models, and bought data from external providers. Eventually the firm’s repositories held data on millions of companies, but then the real challenge began. They needed to clean, integrate, and organize the data before it could be used to train models. Serfaty said that about 95% of the firm’s data science efforts, at least early in his tenure, went toward data preparation.
Now Telstra Ventures has not only a data store, but an automated pipeline that continually refreshes data on companies. They gather and analyze a variety of different metrics to assess a company’s theme, momentum, market, team, and any other factors that might correlate with success. They are always bringing on additional data as it seems relevant and potentially predictive.
Mark Sherman said that the firm has a clear philosophy that governs its use of data science. …….