What is Data Contract?
TL;DR
A versioned, code-defined agreement between data producers and consumers on schema, quality, SLA, and semantics. Prevents breaking changes, -70% data downtime, +50% cross-team data trust. Soda / dbt / Gable / PayPal. Core Data Mesh concept.
Data Contract: Definition & Explanation
A Data Contract is a formalized, version-controlled, auto-validated agreement between a producer (upstream team/service) and consumers (downstream analytics/ML) covering schema (column names/types), quality rules (null rate/range/uniqueness), SLA (freshness/availability), semantics, and change policy (breaking-change notification). It is the data equivalent of an API contract and a core concept of Data Mesh / Data Products. Context: microservices and org decentralization split data production from consumption across teams, so unannounced upstream schema changes (column drops/type changes) caused frequent downstream pipeline failures and dashboard breakage. Data contracts deliver -90% breaking changes, -70% data downtime, +50% cross-team trust, and proactive detection. How it works: (1) schema definition (YAML/JSON Schema/Protobuf); (2) quality rules (null caps/ranges/uniqueness/referential integrity); (3) SLA (frequency/freshness/availability); (4) explicit owner/consumer; (5) versioning + semantic versioning; (6) CI/CD validation (block merge on contract violation at PR); (7) breaking-change notification to all consumers; (8) catalog/registry integration. Leading implementations/tools: (1) Soda (SodaCL + data contract checks); (2) dbt (dbt Contracts — model contract enforcement); (3) Gable.ai (data contract specialist); (4) PayPal Data Contract Template (OSS, de facto); (5) Open Data Contract Standard (ODCS, Bitol/Linux Foundation); (6) Great Expectations (expectations as contracts); (7) Confluent Schema Registry (Kafka, Avro/Protobuf); (8) Monte Carlo/Bigeye (contract monitoring from observability); (9) DataHub/Collibra (catalog); (10) Recap/Memphis. Use cases: (I) breaking-change detection (column drop → CI fail); (II) cross-team handoff SLA; (III) Data Mesh/Data Product boundaries; (IV) Kafka event contracts (Schema Registry); (V) ML feature contracts (train-serve consistency); (VI) regulated data quality; (VII) vendor data exchange; (VIII) Reverse ETL preconditions; (IX) semantic layer consistency; (X) catalog discovery. 2026 trends: (★) Open Data Contract Standard (ODCS); (★) dbt Contracts/Soda Contract adoption; (★) shift-left (CI/CD validation at PR); (★) generative AI contract generation (infer from existing data); (★) Data Mesh growth; (★) Schema Registry + streaming contracts; (★) observability-linked monitoring; (★) catalog integration; (★) ML Feature Store contracts; (★) regulatory compliance contracts.