Software metrics are a valuable tool in helping software engineers to develop large, complex software systems. However, it is vital that transparency and precision are maintained at all stages. We contend that without grammars we cannot define metrics rigorously, without transparent and powerful parsing tools we cannot collect data accurately and without flexible configurable visualisation we cannot exploit the full potential of our data. In this paper, we report the development of JST, a semantic analyser for Java, and show how it is incorporated into our pipeline-based approach to metrics collection and visualisation. We describe a new visualisation, class clusters, which not only demonstrate the data generated by our tools but also illustrate the value of 3D virtual worlds for visualising software metrics.