From all the conferences The Recursive goes to, and all the investors we talk to, one thing comes up more and more in every conversation — investing is no longer a gut feeling. VCs are increasingly turning to statistical modeling and data-driven systems to sharpen their decision-making.
Across the industry, firms are building frameworks that merge machine learning with interpretability. The result is a new kind of investing: transparent, explainable, and grounded in evidence.
Ahead of her talk at Venture Intelligence Day in Prague, we met with Natasa Ralevic, Data Scientist at Angelini Ventures, to discuss how statistical modeling and transparent analytics are transforming the way investors identify opportunities and make strategic decisions.
Venture Intelligence Day, hosted by Vestberry, is connecting data-driven VC professionals to share insights, build community, and expand their networks.
Integrating statistical modeling into real-world investment
Natasa Ralevic is a mathematician with a specialization in statistics, dedicated to systems integration and digitalization. In her role at Angelini Ventures, Natasa supports the team’s investment analysis, market research, and portfolio company initiatives.
“In my work, the integration of statistical modeling into the investment process happens in three key phases,” she explains. “First, we define the problem. Before deploying any model, I collaborate with the investment team to clarify strategic questions, like, for example, which early-stage digital health companies are most likely to scale from Series A to Series B within 24 months.”
Once alignment with business priorities is established, the focus shifts to data collection, cleaning, and enrichment. Building structured pipelines that integrate internal and external sources ensures the data foundation is reliable.
Finally comes model development, validation, and deployment. Interpretability is crucial in this context. “We design models that generate feature importance and partial dependence plots, then embed outputs into Power BI dashboards,” Natasa explains.
“The idea is to make modeling results actionable for the investment team, so it’s not a black box, but a transparent, data-driven layer in the decision-making flow.”
Investor Score
One of Natasa’s standout projects at Angelini Ventures is the Investor Score, a quantitative framework designed to make investment evaluation more data-driven and transparent.
“The goal was to move beyond purely qualitative assessments,” she says. “We built a scoring system grounded in historical data that reflects investing potential in the healthcare sector.”
Through extensive feature engineering, her team created an index aggregating multiple dimensions into a single, interpretable score, allowing objective benchmarking of startups and co-investors.
“The insights proved immediately actionable,” Natasa notes. “We used the Investor Score to identify and target high-potential startups and co-investors during industry conferences, making our networking and sourcing efforts far more strategic.”
Instead of hiding complexity, these systems are designed to reveal everything that can influence decision-making. Models highlight which factors, such as sector growth rates, founder networks, or prior investment performance, most influence outcomes. Dashboards visualize those insights through rankings, scenarios, and comparisons that any investor can understand at a glance.
“I favor simpler models when possible,” she says. “They’re easier to explain, validate, and maintain.” She uses explainability tools like feature-importance rankings and partial-dependence plots to show which factors drive predictions and to visualize “what-if” scenarios.
Just as importantly, she emphasizes how results are communicated. “We translate complex probabilities into practical insights, like ranking companies by quartile, so that investment professionals can act confidently.”
Ultimately, humans remain in the loop. “Models inform decisions but never replace human judgment. Our analysts review, question, and sometimes override results. That dialogue builds trust and continuous improvement.”
Understanding what really drives success
Natasa believes that people and networks often signal future performance more accurately than balance sheets alone.
“Investor and partner strength, team composition, and board experience can be powerful predictors,” she explains. “Understanding who’s backing and guiding a company tells you a lot about its strategic potential.”
When asked about underused analytical methods in venture investing, Natasa shares that is graph analytics.
“Startups don’t exist in isolation, they’re part of complex ecosystems of founders, investors, and partners,” she notes. “Graph analytics can reveal which companies are embedded in high-value networks and identify the most influential nodes.”
She envisions a future where automated data pipelines merge internal and external sources, market signals, clinical data, historical investment records, into unified, intelligent systems.
“That’s how we’ll achieve real-time, data-driven understanding of market dynamics and emerging trends,” she says.
