In the simplest case, a multidimensional cube can be built based on a single table as data source. However, this doesn't happen often.

In real life, where access to all possible information of interest is crucial, data coming from multiple external sources have to be extracted, filtered, merged, and stored in a central repository, called a data warehouse. Its architecture is based upon separate dimension tables and fact tables joined to each other by foreign keys. Correlation of dimension and fact tables can be described either by the star schema or snowflake schema.

A star schema, the simplest style of data warehouse schema, consists of one (or maybe a few) fact tables directly referencing to any number of dimensional tables. The following diagram shows a classic star schema. In a RadarCube library, each dimension is represented by a single flat hierarchy.

A snowflake schema is arranging tables in such a way that fact tables in its center are connected to dimension ones, either directly or through other dimension tables. This architecture is applied for elaborate dimensions with multiple levels of tables and where child tables have multiple parent tables. An example of such hierarchy separated into three tables is given in the picture below. In RadarCube, these hierarchies, created on the basis of their own dimension tables, usually form one multilevel (or user-defined) hierarchy.

 An addition to the names, a dimension table often contains some extra information about its members (for example, the table Customers can have not only name and last name columns, but also address, phone numbers, email, etc.). These additional columns are called hierarchy members attributes.

There are two versions of RadarCube: Desktop (or Direct) uses tables/views/stored procedures of a relational database as data source. An OLAP Cube structure itself is then described by the Cube Editor of the TOLAPCube component.

 The other version (TMDCube) allows an access maintained by the MS Analysis server of versions 2000 or higher to the OLAP cubes built by the server.