A data warehouse, on the other hand, is designed for totally different job. It is designed to give people information rather than data. An example of information would be that the profitability of our company has increased by 3% per month over the last six months.
In order to provide good quality information for a company (in other words, in order to build a data warehouse) we have to overcome two main problems:
The first: is the data is scattered across multiple transactional systems, so we have to extract it from those systems, transform it into a standardized format and finally load it into a central repository called a data warehouse. The second is that we have to reorganize it so that it is presented to the users in an understandable way. As a side issue we need to ensure that the information system is very responsive — in many cases we try to get response times down below two seconds.
The term "data warehouse," in its broadest sense, describes both the centralized repository and the system for delivering information — which is another way of saying that they are woefully sub-optimal for carrying out detailed and complex analysis.
This was first published in August 2010