Word Usage Graphs (WUGs) are an approach of representing relations between word usage pairs, where each word usage is considered as a node and the weighted undirected edge between such a pair represents its semantic proximity. This shifts problems of Computational Linguistics into the graph problem space. There is only little research into how such WUGs can be annotated efficiently and effectively. Therefore, we build a simulation to test a broad range of sampling, clustering and stopping procedures with respect to their impact on finding good solutions. We show that it is possible to simulate graphs which share characteristics close to the observed WUGs. Based on this we are able to scrutinize various annotation procedures and are able to extract their advantages and disadvantages for the annotation process.