What is Schema-on-Read

Definition

Schema-on-read is an approach where data is stored in raw form without a predefined structure and a schema is applied only when the data is read for a specific use, giving maximum flexibility to interpret the same data differently for different needs.

« Back to Glossary Index

Stores data without committing to a structure up front, maximising flexibility
Lets the same raw data be interpreted differently for different use cases
Speeds ingestion since no schema must be enforced at write time
Suits exploratory and evolving analytics on diverse, raw data

Real World Example

A data lake stores raw JSON logs as-is; one team later reads them as session events while another reads the same files for security analysis, each applying its own schema at query time.

FAQs

How does schema-on-read work?

Data is stored raw and a structure is imposed only when it is read, so interpretation happens at query time rather than at write time.

What are the benefits of schema-on-read?

Flexibility to store anything, fast ingestion, and the ability to apply different schemas to the same raw data for different needs.

What is the trade-off versus schema-on-write?

It pushes structure and validation effort to read time, which can mean slower or less consistent queries and weaker enforced quality.