Data cleaning is a self explainatory term. Most of the data warehouses in the world source data from multiple systems - systems that were created long before data warehousing was well understood, and hence without the vision to consolidate the same in a single repository of information. In such a scenario, the possiblities of the following are there:
1. Missing information for a column from one of the data sources;
2. Inconsistent information among different data sources;
3. Orphan records;
4. Outliar data points;
5. Different data types for the same information among various data sources, leading to improper conversion;
6. Data breaching business rules

In order to ensure that the data warehouse is not infected by any of these discrepencies, it is important to cleanse the data using a set of business rules, before it makes its way into the data warehouse.