There are two main reasons for separating logical and physical database design:

When designing a data base system, it is often advantageous to overview the data, even when the database tables will not be normalized.

When you separate the logical and physical design stages, the logical design can be normalized to third normal form. Normalization insures the minimization of data redundancy. The physical data model is created from the logical data model and then denormalized during the physical design stage. The designer then knows that every redundant field in the physical model must have an update procedure in the system code.

Distributed systems are one database system logically, and multiple database systems physically. Separation of logical and physical models during design allows both aspects of the system to be treated effectively.

One way to design distributed systems is to create a Project Data Model made up of multiple subject area ERDs, each of which represents a different database in the distributed system. The ER Model diagram displays all subject areas in one diagram, so it is easy to visualize and report on the relationships between the various databases that are included in the system.

You can then generate one physical model from the ER Model diagram, and one from each of the subject area diagrams. The components of the physical model maintain their logical source. For example, the source of a table in a physical data model is an entity in an ERD; the source of a column in a PDM table is an attribute in a ERD entity.