Businesses generate more data than they can neatly model upfront: logs, clickstreams, images, sensor data, and exports from many different systems. A warehouse usually expects data to be shaped before it is loaded. That can be expensive, and it can leave out data that seems unimportant today but becomes useful later.

A data lake takes a more flexible approach. It stores raw and cleaned data in open formats, usually on object storage, so teams can use it later for analytics, machine learning, search, and compliance.

But flexibility needs control. Without owners, metadata, quality checks, and cleanup rules, a data lake slowly turns into a data dump that nobody trusts.

This chapter explains what data lakes are built for, how they differ from warehouses, how file layout affects performance, and how governance keeps a lake usable.