Predictive modeling

Strategy research treats performance heterogeneity as a central puzzle, but most evidence on its sources is assembled only after outcomes have already diverged. This paper asks whether founders’ entry decisions reveal, before outcomes materialize, how widely those outcomes will later spread. Building on occupational-choice models and the options logic of entrepreneurial entry, I recover the payoff-dispersion threshold at which a founder is just indifferent between continued paid employment and founding, given the founder’s forgone wage, wealth, and the scale of the opportunity. I call this revealed-preference threshold entry-implied dispersion. Using Finnish linked employer–employee and business-registry data for 2001–2022 on the founders of limited-liability firms, I aggregate founder-level thresholds to three-digit industry-years and validate them against realized revenue-growth dispersion for two non-overlapping groups: the entering cohort of new firms and incumbent firms fixed before entry. Entry-implied dispersion predicts entrant-cohort dispersion immediately and incumbent dispersion with delay. A serial-founder test that removes each founder’s own prior entry hurdle attenuates the cohort-level relationship but leaves the delayed incumbent relationship intact, suggesting that the cohort result partly reflects persistent founder-level heterogeneity while the incumbent result reflects a broader industry-level signal. Two further checks sharpen this reading: an expected-payoff version of the same entry inputs shows no comparable link to realized dispersion, and entry-implied dispersion shows no comparable link to average realized returns. Together, the results suggest that a founding decision carries information about how unevenly a competitive arena’s future outcomes will be spread, and how much of that spread traces to the founders themselves versus the arena they enter.

Paper available upon request.

Organizations are adopting AI tools faster than they can tell which of their people will actually use them. This project asks whether structured executive interviews, transcribed and coded, predict which employees go on to adopt AI in their work, and whether leadership experience with transformation, technology, and scaling, observable from career histories alone, bears on an organization’s capacity to direct that adoption. The setting is an AI assessment platform used by private equity firms, in which structured interview transcripts and career records are linked to the organizations those executives lead. The aim is a measurement approach that treats AI adoption as a predictable consequence of what an interview reveals about a person and a team, rather than as a matter of tools or training alone.

Paper available upon request.

The company builds AI assessment software for private equity firms. Its product predicts executive performance from structured interview data and guides interviewers toward the follow-up questions that matter. The models sit on a matrix of outcomes and evidence, from career histories through short and long interview transcripts, each with its own validation protocol, and the product is in beta with clients.

I lead all of the modeling work: measurement design, model development, validation, and research and development.