A data warehouse is where most of a company's analysis usually happens.
It brings data from many systems into one place, in a consistent shape, so they can be compared and joined.
Without one, every analyst pulls from every system separately and gets slightly different numbers.
Data is organised into tables designed for analysis rather than for running applications.
A common design is the star schema: a central fact table of events surrounded by dimension tables that describe them.
For example, a sales fact table holds one row per sale, linked to dimension tables for product, store and date.
Facts are usually numbers to add up, while dimensions hold the labels used to group and filter.
History is kept, so trends over years can be studied even after the source systems change.
Modern warehouses separate storage from computing power, so each can grow on its own.
Many run as cloud services, charging for storage and for the processing each query uses.
Warehouses are built for reading, so loading happens in batches rather than row by row.
A single agreed definition of each metric, held in the warehouse, prevents conflicting reports.
The cost of a warehouse grows with careless queries, so scanning less data saves real money.
In practice, the warehouse is the source that dashboards, reports and most analyses read from.
The usual mistake is loading data without agreeing on definitions, which only moves the arguments into the warehouse.