A Kafka topic without schema governance is a black box: producers publish whatever format they choose, consumers must guess the message structure, and a schema change by the producer silently breaks every downstream consumer. The Confluent Schema Registry solves this by providing a centralised, versioned repository of message schemas. Every message produced carries a schema ID in its header; every consumer looks up that ID from the registry to deserialise the message. Compatibility rules enforced by the registry prevent breaking changes from being registered without explicit approval.
Apache Avro is the most commonly used serialisation format with Kafka Schema Registry. Avro uses a binary encoding that is compact and fast to parse, and its schema definition uses JSON to describe field names, types, defaults, and documentation. Avro schemas support evolution — adding optional fields with defaults, removing non-required fields — in a way that maintains backwards and forwards compatibility between schema versions. This evolution capability combined with registry enforcement makes Avro the standard choice for Kafka pipelines where schema changes will occur over time.