In contrast to relational databases, in MongoDB we have to base our modeling on our application-specific data access patterns. Finding out the questions that our users will have is paramount to designing our entities. In contrast to an RDBMS, data duplication and denormalization are used far more frequently and with solid reason.

The document model that MongoDB uses means that every document can hold substantially more or less information than the next one, even within the same collection. Coupled with rich and detailed queries being possible in MongoDB in the embedded document level, this means that we are free to design our documents in any way that we want. When we know our data access patterns we can estimate which fields need to be embedded and which can be split out to different collections.