Bringing more context to founder conviction
At nvp capital, we have honed our founder framework over the years around three core pillars: zero-to-one experience, excellence, and domain expertise. These guide how we get to know founders and build conviction about the people we want to back.
At the beginning of this year, we wanted to bring more precision to our process. How do we translate these broad qualities into specific, measurable attributes? How can data sharpen our understanding of what matters, and change how and where we source?
The raw data about founders is everywhere: career histories, educational backgrounds, research papers, company records, product launches, and accounts of what people had built. But that information was fragmented across sources, making it difficult to see patterns at scale. What could we learn by bringing it together and evaluating what experiences actually matter?
New tools and models put those questions within reach, just as AI was changing how people build companies. We set out to examine our assumptions, uncover patterns, and use the evidence to refine how we find and evaluate founders.
Building the research
Finding the right collaborator was an important part of the project. We were introduced to Rohan Sontakke, a UC Berkeley student immersed in the university’s entrepreneurship community who brought technical fluency with emerging AI tools and curiosity about our research questions. We were fortunate to bring him on as nvp capital’s data science fellow.
Together, we refined what we wanted to learn, selected the cohorts, and developed the approach. The research covered more than 2,000 founders across 1,200+ companies, collecting 130+ attributes per founder and analyzing approximately 850 variables, including signals derived from their experiences and accomplishments.
To understand a recent generation of founders building in the AI era, we compared those whose companies had reached a substantial funding milestone with a cohort that had raised less. Fundraising gave us an observable measure of early progress while longer-term outcomes were still unfolding, rather than proof of lasting business success.
Rohan built a research pipeline combining structured data providers, web searches, and language models. It connected fragmented information into founder profiles, which were manually audited. Models also helped turn open-ended descriptions of work and accomplishments into comparable attributes.
The work came with the usual challenges of collecting so much data so fast, but the result was worth it.
Making the framework more concrete
The analysis examined which experiences were common across the dataset and which differentiated the cohorts. We studied attributes individually, in combination, and within sectors.
Some experiences that sounded compelling were widespread in both groups. Others became more informative when considered alongside a founder’s role, prior company, or sector. Rohan and I then mapped those findings to more specific attributes within our three pillars, testing our assumptions and sharpening how we defined each one.
“The most interesting part was when the data and our pre-existing intuition diverged,” Rohan said, “and deciding how much to trust each.”
Those findings also informed the scoring approach: different attributes and combinations carry different weight, reflecting their associations with the fundraising milestone in the study cohort. The resulting scores give us a consistent basis for comparison and a clearer view of the experiences worth digging into.
Context matters
One of the most useful lessons was how much context sits underneath a familiar label.
“Repeat founder” tells us someone has started a company before. It leaves open what they personally built, what happened to the business, and what they learned. Those distinctions mattered in the research and have sharpened how we evaluate second- and third-time founders.
Similarly, being early at a startup is only one place to look for zero-to-one experience. A researcher may have developed a new technical approach, or someone at a large organization may have launched a major initiative. Looking closely at what they created and owned helps us recognize experiences that broad career labels can miss— and build conviction where others might pass.
For research backgrounds, we look beyond a degree or institution to the field, published work, recognition, and practical technical contributions. Sector also mattered: an experience associated with the fundraising milestone in one field could be less informative in another.
That specificity helps us ask better questions and sharpen our sourcing: which research communities, programs, teams, and career paths should we spend more time getting to know?
Putting it to work
The project produced a research report and a founder scoring tool. The report informs our framework; the tool helps our small team prioritize research and prepare more focused founder conversations.
The detail behind the score matters: what someone created, what they owned, where they demonstrated exceptional ability, and what we still need to learn. For example, understanding a founder’s distinctive insight into an industry still requires spending time with them.
Our aim is to arrive with context and make our time with founders more substantive. By making our beliefs explicit, testing them against evidence, and refining them as we learn, we strengthen the foundation for our own independent conviction. That’s what nvp capital has always aimed for: doing the work to see founders clearly, before anyone else does.
