Data warehouses and lakehouses are the big stores where a company keeps copies of all its records, apart from the databases its apps run on. There, analysts can add up and question millions of records together without slowing the apps. This zone of data engineering holds three products: the data warehouse, the data lakehouse and the real-time analytics database.
Why keep a separate store? An app's own database, described in Databases, is built to record one order or one payment at a time, very fast. A question such as "how much did each locality spend on groceries in March?" has to read every order of the month. If that question ran on the app's database, customers placing orders would have to wait. The two kinds of work are called OLTP and OLAP, and the page on OLTP and OLAP explains the difference.
Consider a grocery delivery company in Lucknow. Every night, its data pipelines copy the day's 2 lakh orders from the app's database into a data warehouse. The next morning, the finance team asks which localities ordered the most. The warehouse answers in seconds. It reads only the columns that the question needs, because it stores data column by column, as a columnar database does.
The company also keeps raw files that do not fit neatly into tables: riders' GPS readings, app click logs and photos of damaged items. These go into cheap object storage. A data lakehouse lays tables over those files. So the same files can be questioned like a warehouse and also used to train AI models. Open table formats are the public rules that make those tables readable by more than one piece of software.
Some questions cannot wait for the next morning. A real-time analytics database answers within a second over data that is still arriving, such as the orders of the last five minutes. The difference between loading data once a night and reading it as it arrives is explained in batch vs stream processing.
The choice of store is hard to reverse, because years of tables and queries get built around one product. This is called vendor lock-in. The data in these stores is described and checked by the tools in data governance, and it is read by business intelligence dashboards.