I am not a purist when it comes to data modeling (logical and physical). This may be one of my flaws (there are so few - haha). I do believe that logical modeling and physical modeling are two different arts with two different skill sets.

Logical modeling requires a person to understand and depict the business rules in a modeling tool, using things such as super/subtype relationships, etc. Logical data modeling represents the organization data. Definitions of entities and attributes are a must for any good logical data model, as well an understanding of how business is conducted. I have seen some of the best logical data modelers who have never seen a database or the DDL (data definition language) that creates the database objects.

Physical modeling takes into account how the data is inserted, updated or deleted - as well as how the data is used (performance and usage). Tuning of the insert, update or delete and the queries is a must, and sometimes we get pretty creative on when (or when not) to build indexes. I call this an "indexing strategy." Every data group needs one for each database you are designing. I have seen some of the best physical data modelers and database designers who have never seen a data modeling tool.

When you are lucky enough to have both in one person – hold on to them!