We often use data models to make broad sweeping statements about the business.

This forces us to generalize many attributes into entities into which they don't belong.

By breaking out the related entities into increasingly detailed attributive entities, the ambiguities and exceptions inherent in these generalizations can be identified early.

The goal is that the model be accurate for all possible cases. No exceptions!

Breaking fundamental entities out using increasingly detailed attributive entities promotes better understanding of the data, and expanded opportunities to use the data to support the business.

It also prevents us from building business systems at the wrong level of detail.

It turns out that by always connecting relationships to the bottom of our discrete detail "stack" we'll prevent most level-of-detail defects.

This means challenging any relationship that connects to the stack anywhere except at the bottom entity type.