In data modeling, a logical data model is the representation of business data into a data model that can be the basis for the physical database implementation. It identifies a data "periodic table" which will be the basis for the business organization’s functions, processes and task to be performed. Data modelers design a logical data model in order to be able to establish a data processing environment where the basic data is captured only once, stored, and then shared to data consumers who are authorized by the company for generating statistical reports or the public who may want information.

At its most basic level, a logical data model defines things about which data is kept such as people, places events. These are technically known in database term as entities. The world relationship is another database term used in a logical data model to mean the relationship or connection between the entities. Finally, the term attribute refers to the characteristics of the entities.

There are certain rules to follow in using attributes in logical data modeling. The rules below are as follows:

1. An attribute should posses a unique name and the same meaning must be consistently applied to the name

2. An entity may have one or more than one attributes. Every attribute is owned by exactly one entity in a key based or fully attributed model. This is also referred to as the Single Owner Rule.

3. Any number of migrated attributes may be owned by an entity as long as the migrated attribute is a part of the primary key of a related parent entity. In principle, the primary key must always migrate while a non key attribute must never migrate.

4. An entity instance can not have more than one value for an attribute which is associated with the said entity. This is also known as the No-Repeat Rule or First Normal Form Rule.

5. An attribute which is not part of a primary key can be null or meaning not known. In the past, this was known as the No-Null Rule but is no longer required now. The data modelers in the past refused to take a non key attribute which could be set to null .

6. Models should no not constrain two distinctly named attributes in which the names are synonymous. Two names are said to be synonymous if both as alias for one another, whether directly or indirectly. Also, they are said to be synonymous if there is a third name for which both names are aliases.

Attributes may be multi valued, composite of derived. A multi valued attribute can have more than one value for at least one of the entity’s instance. As an example, a software whose entity is called application may have a multi value attribute called platform because different instances of the same application may run on different platforms. To illustrate further, the application may be a document processor which can run on Microsoft, Apple and Unix platforms.

A composite attribute may contain two or more attributes. An address can be a composite attribute consisting of street address, city, region, state and so on.

A derived attribute is an attribute whose value is taken from other data and may be a result of a formula. A person’s age can be an attribute derived from another attribute which is the birthday. Derived attributes are very common in business data warehouses where atomic data are aggregated heavily to form the report about the company profile.