Companies providing mobility solutions, such as Uber, Lyft and a host of short-term car rental companies, can generate value from information technology to track their vehicles location. Real-time decision making to act on this information is critical for the success of one-way mobility companies. We show how these companies can predict vehicle demand with high accuracy for vehicles across time and urban areas. We validate this model by tracking the movement and transactions of 1,100 vehicles from the carsharing service Car2Go in Berlin. With our model they could preposition vehicles to increase service levels with a smaller fleet. The accuracy of the model to predict demand area and times is a key contribution to urban mobility systems. Prepositioning vehicles based on expected demand and supply as modeled in our paper are vital to the business models of emerging transportation network companies like Uber and will be indispensable for autonomous vehicles.