Sensors embedded in smart objects, smart machines, and smart buildings produce ever-growing streams of contextual data that convey information of interest about their operating environment. Although an increasing number of industries embrace the utilization of sensors in routine operations, no clear framework is available to guide designers who aim to leverage contextual data collected from these sensors to develop predictive systems. In this paper, we applied the Design Science Research methodology to develop and evaluate a general framework that helps designers to build predictive systems utilizing sensor data. Specifically, we developed a framework for designing context-aware predictive systems (CAPS). We then evaluated the framework through its application in MAN Diesel & Turbo, which served as a case company. The framework can be generalized into a class of demand-forecasting problems that rely on sensor-generated contextual data. The CAPS framework is unique and can help practitioners make better-informed decisions when designing context-aware predictive systems.