Abstract
The benefits to data analytics and machine learning have been distributed unevenly across firms around the world. Research on IT productivity points to intangible capital as a key driver of value creation from innovation in computing. We argue that a crucial component of intangible capital is organization-wide technological architecture, which is idiosyncratic and difficult to measure. We use a novel survey instrument to quantify large corporations’ data architecture capabilities by their closeness to “best practices” of frontier digital companies. Using the prevalence of third-party maintenance as a proxy for legacy servers before 2016 and an instrument for data architecture coherence, we find that improving data architecture coherence increases machine learning capabilities. Legacy servers reduce data architecture coherence particularly at corporations with complex software systems, consistent with the hypothesis that costs of digital transformation are greater when workers need to develop more complicated co-invention processes to interact with technical systems.