Schema-on-Read is a data‑management approach in which data is stored in its raw, unstructured, or semi‑structured form, and the schema (structure, data types, and relationships) is applied only at the time the data is read or queried. Instead of forcing data to conform to a predefined model when it is stored, the structure is interpreted dynamically based on how the data will be used.

Schema‑on‑Read is fundamental to data lakes, clickstream analytics, IoT pipelines, JSON/XML ingestion, semi‑structured logs, and exploratory or machine‑learning workloads where flexibility is essential.