The routing plan of a delivery fleet is usually developed a priori and provides an optimal (or near optimal) way of serving customers by taking into account certain constraints, such as delivery time windows. However, such plans may not cope adequately with the dynamics of a city logistics environment, in which unexpected events (e.g. traffic congestions) often occur during delivery execution. We present the development and evaluation of a real-time fleet management system that handles such unforeseen events. The system monitors the delivery vehicles in real time, detects deviations from the initial distribution plan, and adjusts the schedule accordingly by suggesting effective rerouting strategies. The system has been tested in simulation environment and in real-life cases and the results show that delivery performance is enhanced significantly and customer satisfaction is improved.