The knowledge of future Signal Phase and Timing information (SPaT) of traffic lights ahead enables a vast number of driving assistance functions, such as Green Light Optimal Speed Control (GLOSA), Red Light Duration Advisory (RLDA) or Traffic Signal Adaptive routing (TSA routing). The purpose of TSA routing is to reduce the travel time by choosing a route that is possibly longer than the shortest one but has less red lights. Whereas GLOSA and RLDA are quite easy to implement from a scientific point of view, TSA routing presents a certain challenge: first of all, TSA routing necessitates predictions on future signal states on a wider look in the future than GLOSA and RLDA, a possible reason why this field of research still seems rather unexplored. Second, there are still many unresolved issues, such as the inadequacy of graphs for TSA-routing, or proper traffic load estimations. In this paper, we present a fully functioning model for TSA routing on the basis of our forgoing research on the prediction of future signal states and address the question of practical usability by evaluating our model under realistic conditions. We analyze, among other things, the impact of partial knowledge on traffic light's future signal states and the impact of different traffic loads on TSA routes by means of a test field in Munich, Germany. We describe necessary modifications of the underlying transportation network's graph structure and shortest path routing algorithm to allow routing under consideration of future signal states of traffic lights. We show that, albeit there are many erratic aspects in traffic and signal states, TSA routing still reaches a significant travel time gain over usual routing in our test field.