Data-Driven Firm Productivity: The Importance of Technological Architecture

Abstract

Large traditional corporations have been undergoing digital transformation since the late 2015, but investments in new generations of data analytics and machine learning yield largely uneven productivity benefits. We show that firms’ data architecture capabilities are complementary to the adoption of machine learning and data analytics, and are crucial to driving productivity gains from technological innovation. We use a detailed survey to derive separate indexes for three distinct architectural components. We report three findings: data platform and analytics/ML architecture are complementary to machine learning adoption; poor data platform capability is associated with fragmented ML adoption and lower productivity, but fully developed data architecture capabilities are associated with larger scope and variety of ML functionalities; cloud architecture is complementary to demand for data analytics skills. Empirical evidence also suggests that large U.S. corporations are early adopters of cloud and data analytics, and late adopters of machine learning technology.