Challenges in the Model Development Process: Discussions with Data Scientists

Abstract

Businesses are increasingly seeking out analytics to improve decision-making processes, although often with hesitations. Decision makers often do not have the sophisticated analytical skills needed to fully understand the analytics process. Contrastingly, data scientists may lack the business acumen needed to fully grasp the business context of the decision. In this research, we consider the perspective of the data scientist through a series of interviews to draw out challenges in the analytics process. We use principal-agent theory as a lens to shape our understanding of the conflict that arises due to goal misalignment and information asymmetry between the principal and agent. Findings are presented in the CRISP-DM process and a future research agenda is proposed.