The most common way of storing data is based on the notion of a relational model—an idea introduced by Dr. Edgar Codd in the early 1970s. Here, an entity is stored as a tuple (or row) of attributes (or columns). A database is simply a set of rows, all of which have the same set of columns (or schema). Tables are defined using a static data schema; relations between entities are modeled by foreign keys or relationship tables; rows from different tables can be referenced using foreign keys.

There are many ways of representing the conceptual data model presented earlier. However, not all representations are efficient for all use cases. To enable figuring out an optimal relational structure, Dr. Codd expressed a series of progressively more restrictive constraints on the structure of data. With each level of constraint/rules, the amount of redundancy in the representation will be reduced. This process of introducing constraints and refactoring the structure to reduce redundancy is called normalization, and each level is called a normal form. Let's look at the various forms.