Applying use cases to design versus validate class diagrams - a controlled experiment using a professional modeling tool

Several processes have been proposed for the transition from functional requirements to an object-oriented design, but these processes have been subject to little empirical validation. A use case driven development process is often recommended when applying UML. Nevertheless, it has been reported that this process leads to problems, such as the developers missing some requirements and mistaking requirements for design. This paper describes a controlled experiment, width 53 students as subjects, conducted to investigate two alternative processes for applying a use case model in an object-oriented design process. One process was use case driven, while the other was a responsibility-driven process in which the use case model was applied as a means of validating the resulting class diagram. Half of the subjects used the modeling tool Tau UML Suite from Telelogic; the other half used pen and paper. The results show that the validation process led to class diagrams implementing more of the requirements. The use case driven process did, however, result in class diagrams with a better structure. The results also show that those who used the modeling tool spent more time on constructing class diagrams than did those who used pen and paper. We experienced that it requires much more effort to organize an experiment with a professional modeling tool than with only pen and paper.