How Telstra Ventures uses data science to improve venture capital investing

The venture capital industry has played a vital role in rapidly growing, cutting-edge technologies. Yet, it has been a laggard when it comes to adopting new technology itself. 

About five years ago, Mark Sherman, managing director at Telstra Ventures, set out to change that by building up his data science team. Telstra Ventures hired Jonathan Serfaty, a former LinkedIn engineer, as its head of data science. Serfaty had been working on LinkedIn’s lead prospecting pipeline, which mapped well to the deal pipeline VCs use. 

It took a few years to get things off the ground, but Telstra Ventures is already starting to see some impressive results: 

• Telstra Ventures is now sourcing 15% of new deals from data science recommendations and data science tools have informed 100% of all deals since 2020. 
• 57% of the deals sourced through data science raised an additional round within a year, compared to 33% for deals sourced the old-fashioned way. 
• Data science source deals saw an increase of four times in reported valuation, compared to a 2.4 times increase for deals sourced using traditional channels.

The new approach is still in its infancy, but it shows tremendous promise. Sherman expects the company could source as much as half of their new deals using the latest data science techniques within five years. This approach works because Telstra Ventures focuses on companies that have already been doing enough business to generate a trail of data. 

“This would not work as well if you were trying to do the same with pre-seed and seed funding because there is not as much digital exhaust,” Sherman said. 

What to model

Creating a digital model of a startup in an emerging market is a bit more complex than modeling a public firm in an established market, Serfaty told VentureBeat. He has invested significant resources in tools for crawling the Internet for public information and curated the appropriate mix of third-party data services. 

They have developed metrics to characterize the ways companies engage with customers, their growth rate and the connection between players in a market. Serfaty said, “There is so much information that is hidden and unknowable. We are looking for proxies that are at least directionally good enough to be useful.” 

Many of these models take advantage of graph data modeling techniques Serfaty worked with to improve lead prioritization for the sales team at LinkedIn. He told VentureBeat, “We measured a lot of signals from incoming accounts and leads to figure out how to prioritize leads for the sales team. This is a similar problem to what we are doing here.”

Improving the venture capital pipeline

A venture capital deal pipeline has three key elements: sourcing, benchmarking and value-add. Sourcing is the process of sniffing for momentum within a market segment. Benchmarking is comparative financial analytics to understand a company’s strengths and prospects. Value-add involves finding ways to improve the prospects or value of companies. Telstra Ventures has developed data science tools to improve all three of these processes. 

With sourcing, the traditional venture capital approach is to rely on inbound or outbound lead generation. An inbound process might involve becoming well-known in a domain that attracts startups in that area. An outbound approach involves researching the market and working the network to find businesses in a specific area. 

The data science effort helps identify and prioritize candidates for outreach. This takes advantage of several proxies that correlate with various measures of success, but that are easier to measure for startups. This is 15% of companies with the outsized returns mentioned above. 

Data science teams also help investors assess companies identified through other channels before proceeding further.

“Data science touches every investment we make, whether inbound or outbound,”  Sherman said.  

Telstra Ventures also makes extensive use of the new data science tools in the benchmarking phase. Although VC firms have always done analytics, the latest data science workflow takes things to a new level. For example, the data science team has developed tools for generating over two hundred KPIs that can help compare the performance of different firms. 

According to Sherman, ten years ago, most decisions were based on intuition. Now, by comparing this much richer set of metrics, his team has a much higher confidence interval in making investment decisions. 

The data science workflow also helps Telstra Ventures improve the value-add phase by identifying specific weaknesses to mitigate and opportunities to pursue. 

Telstra Ventures specializes in assisting companies in cultivating more revenue-bearing relationships. Serfaty’s team developed various graph analytics tools to identify and prioritize prequalified prospects and determine the appropriate contact to get the ball rolling. 

It took some time for Telstra Ventures’ team to figure out how these new data science tools could fit into their workflow. Now the investors are starting to suggest adjustments for better models and new metrics to track, said Serfaty. 

For example, the investors have asked for network insights to help understand how they are connected to a company and who they should reach out to for an introduction, as well as tools to help search and map out sectors for thematic research. 

“Additionally, as the VC landscape evolves, we’ve gotten suggestions on how we can monitor and evaluate Web3 companies,” said Serfaty.