A data lake stores raw data in its native format. No schema is enforced on write. Structured tables, JSON logs, images, and video all sit in the same repository. The schema is applied when the data is read, not when it is stored. That flexibility is the appeal. The data lake accepts anything. It does not require you to know how you will use the data before you store it.

The flexibility is also the problem. Without governance, a data lake becomes a data swamp. Nobody knows what is in it. Files have cryptic names. Metadata is missing. The data is duplicated across folders. Analysts cannot find what they need. Data scientists waste time cleaning and exploring instead of analyzing. The lake becomes a cost center, not an asset. The fix is governance: catalogs, metadata, access controls, and clear ownership. Modern data lakehouse architectures add table formats like Delta Lake and Iceberg that bring schema enforcement, ACID transactions, and query performance to the lake. The lake and the warehouse converge. The best of both: the flexibility of raw storage and the reliability of structured tables. The technology improved. The governance requirement remains.

Data lake characteristics

• Schema-on-read — apply structure when querying
• Raw storage — data kept in original format
• Scalable — handles petabytes on commodity storage
• Flexible — supports structured, semi-structured, and unstructured data
• Governance-dependent — without cataloging, it becomes a swamp

A data lake is a reservoir. Without a map, it is just a body of water.