If you don't really know about something, your can't act on it. Data architectures and their model-developed data structures represent the meets and bound of an enterprise's policy.
When you are embarked on data modeling you're not modeling data, you're modeling policy.
If you are too narrow, the quantity of applications expand as the quanity of data models expand. Expands as well are all the data integration efforts. Some of these efforts consume massive resources so that there is interoperable policy across the enterprise.
Data models represent the critical first step in the creation of enterpirse-wide business information systems. Do you have one, harmoneous semantic model or a large quantity of parochial independent models. While the first is hard, expensive and time consuming, the second will lead only into total disaster.