Data normalization exists in multiple levels, according to multiple rules. Normalization reduces data redundancy and increases data integrity. Informally, the cutoff for "well-normalized" is between 3NF and 4NF. Eventually, 6th Normal Form is what you'd call a whole host of Key-Value pairs, where the key is a Primary Key (this is known as a "Columnar data store").

For well-Normalized data... put simply:

Every non-key attribute depends on the key (1st normal form) the whole key (2nd normal form) and nothing but the key (3rd normal form)

• Colorado State Paper

Unnormalized Data

• anything goes

First Normal Form (1NF)

• Cells contain only one value

Second Normal Form (2NF)

• Cells contain only one value
• All non-Prime fields are functionally dependent on the whole primary key

Third Normal Form (3NF)

• Cells contain only one value
• All non-Prime fields are functionally depending on the whole primary key
• All non-Prime fields are independent of each other

Source

Relational Database Schema Design Overview

Database normalization

Related Notes