Science requires tools, and computer science is no different. In a typical research context however, it is not known upfront how a tool should work. Researching the tool's design is part of the investigation process. Various designs have to be prototyped and experimented with. This paper focuses on the research process of interactive visualization tools. We present how to improve development, so that a novel tool can be tested and modified at (almost) the same time. We present the Mondrian framework, which supports on-the-fly prototyping of interactive visualizations. As an example, we present the research process of the Feature Dependency Browser, a visualization tool which we developed to allow software engineers inspect runtime dependencies between features.