Existing registries organize functionally similar services into groups without considering past service-usage from the consumers' perspective, a.k.a. pragmatics. Pragmatics can help registries to calculate service similarity more effectively and improve organization schemes. However, pragmatics are not available beforehand and their highly accumulated number over time creates time and space efficiency challenges. To be responsive to dynamically arrived pragmatics, we propose a self-adaptive service-organization approach that follows an iterative life-cycle for autonomously evolving organization schemes. To face the emerging efficiency challenges, our approach adopts a greedy online algorithm for the evolution of organization schemes that considers and stores only the top-k pragmatics of each service. We evaluate our approach on benchmark services and the results show that the effectiveness and efficiency of our approach are higher than those of a state-of-the-art service-organization approach, while a low total number of pragmatics is greedily stored by our approach.