Software analysis process - which order of activities, is preferred? An experimental comparison using FOOM methodology

Data modeling and functional modeling are two main activities of the analysis process. Traditional development methodologies usually emphasize functional modeling via dataflow diagrams (DFDs), while object-oriented (OO) methodologies emphasize data modeling via class diagrams. UML includes various techniques for both data (structure) and functional (behavior) modeling which can be utilized in various ways. In fact, different methodologies utilize data modeling and process modeling techniques in different ways and orders. This study is concerned with the ordering of modeling activities in the analysis stage. Our main question is if it is better to create first a functional model and then a data model, or vice versa? We conduct a comparative experiment in which the two opposing orders are examined We use the FOOM methodology as a platform for the experiment as it enables the production of both a functional model (hierarchical OO-DFDs) and a data model (an initial class diagram), which are synchronized. The results of the experiment reveal that an analysis process that begins with data modeling provides better products than one that begins with functional modeling.