Below are the common phases applied, for a real world requirement(data) map to DBMS specific schema.
Conceptual models ER/EER/CODASYL/Hierarchical bridge the gap, in mapping the real world data(requirement) to DBMS specific schema.
In RDBMS world, object-relational impedance mismatch is the cause for applying these phases in database design.
Using RDBMS, reduces the agility of the application, because once the database is designed, it takes a lot to modify the DB schema again, based on new requirement.
But using, Document database(say MongoDB), factor of object mapping is intuitive, that minimizes impedance mismatch issue.
So, embedding & referencing of entities in JSON, would suffice for modeling real world data(requirement) to a document database.
Conceptual models(ER/EER/..) would not look like a necessity, for modeling document database.
Modeling data for document database(say MongoDB), are these phases still valid?