Abstract
Machine data (MD), that is, data generated by machines, are increasingly gaining importance, potentially surpassing the value of the extensively discussed personal data. We present a theoretical analysis of the MD market, addressing challenges such as data fragmentation, ambiguous property rights, and the public-good nature of MD. We consider machine users producing data and data aggregators providing MD analytics services (e.g., with digital twins for real-time simulation and optimization). By analyzing machine learning algorithms, we identify critical properties for the value of MD analytics, Scale, Scope, and Synergy. We leverage these properties to explore market scenarios, including anonymous and secret contracting, competition among MD producers, and multiple competing aggregators. We identify significant inefficiencies and market failures, highlighting the need for nuanced policy interventions.