A data dictionary, at its simplest, is a list and description of every variable within a dataset, including information such as the units of measurement and what the variable represents.

For more complex datasets, including multi-level or larger database structures, the data dictionary also includes descriptions of the relationships between tables, and for categorical data with a pre-defined set of possible options (e.g. an "enum" datatype in SQL, data coming from a "select" question in a survey, or other defined data such as days of the week), the data dictionary should also include the list of all possible values.