Overfetching

Overfetching occurs when a data retrieval operation returns more data than is actually needed by the consumer. It is a common issue in APIs, web services, databases, and client-server applications, where payloads include fields, records, or relations that the caller will not use.

Causes include selecting broad data shapes (for example, using SELECT * in SQL), returning entire resource representations in REST endpoints, or ORM configurations that load full objects with many attributes or related entities. In GraphQL, overfetching is less common because clients request only the fields they need, but it can still occur if the schema is poorly designed or nested data is fetched unintentionally.

Impacts of overfetching include higher bandwidth usage, increased latency, extra processing time on servers and clients, greater memory consumption, and worse battery life on mobile devices. It can also complicate caching and increase the risk of exposing sensitive data.

Mitigation strategies include using sparse fieldsets or field filtering in APIs, designing endpoints to return only necessary data, and using projections in SQL to select specific columns. GraphQL helps by letting clients specify fields, though careful query design and server-side validation are still important. Other approaches include backend-for-frontend patterns, pagination or cursor-based pagination, and selective eager loading with explicit joins to avoid fetching unused relations.

Overfetching is often contrasted with underfetching, where too little data requires additional requests. Balancing both concerns is a common performance optimization in API and database design.