Data Architecture, Machine Learning and Firm Productivity

Abstract

As enterprise IT systems increasingly incorporate data-driven technologies, it is crucial to understand complementary organizational practices that allow firms to unleash productivity benefits from adoption. This study uses survey and prediction methods to measure the data architecture of 225 large corporations and finds that data fabric capability complements enterprise ML software adoption. While corporations without a coherent data fabric suffer productivity losses from enterprise ML software investments, such investments lead to significant productivity gains among corporations with fully-developed data fabric capability. When data fabric capability is particularly low, enterprise ML software standardization significantly improves firm productivity.