Misconceptions About Building a Machine Learning Platform for Risk

Companies -- especially those who like to "build" rather than "buy" -- may be inclined to launch their own machine learning system instead of using a solution from an external provider. Building such a system internally, however, is not without its own set of challenges. This paper presents five factors businesses may overlook, including the difficulty in operationalizing machine learning models and creating feedback loops to enhance a model's self-learning capabilities. The document also aims to separate fact from fiction concerning the time needed to deploy a machine learning solution, challenges launching a model in production, and the role of metrics in assessing a model's performance and efficacy.