Beyond Data Accumulation: Data Traps and Collaborative Meta-Knowledge

Joint work with Francesco Castellaneta, Filippo Dal Lago & Davide La Torre

Abstract

The resource-based view predicts that firms in data-intensive settings gain advantage by accumulating proprietary data and training predictive models internally. We identify a boundary condition: when prior strategic choices narrow the experiential domain captured in proprietary data, accumulating more data does not necessarily broaden learning and can instead result in a data trap.

We introduce collaborative meta-knowledge—predictive regularities encoded in trained model parameters that travel across firm boundaries without transferring underlying records—and theorize how data-local exchange can provide access to external learned patterns.

Using information from Private Placement Memorandums (PPMs) collected by Limited Partners (LPs) across multiple, often competing, private equity funds, we construct an ex post counterfactual model-exchange simulation for 89 firms. Externally trained models improve predictive performance for 82 of 89 firms (92%), with substantial average gains.

Because the simulation abstracts from search, governance, and coordination frictions, these estimates should be interpreted as upper bounds. The results qualify a strongly proprietary-data-centric view of advantage: in prediction-intensive competition, performance may depend not only on data ownership, but also on the portability of learned patterns and firms’ ability to deploy them.

About the speaker

Samuele Murtinu is a Full Professor of International Business and Head of Section (Entrepreneurship) at the Utrecht University School of Economics. His recent research mainly focuses on (i) the organizational impact of AI, (ii) human-machine interactions, and (iii) different types of innovation financing. Samuele Murtinu has acted as independent evaluator for the Italian Ministry of Economic Development, several governmental agencies, universities, private organizations and industry associations.

Language

English

This is a free seminar. Registration is mandatory