Data SLA (service-level agreement)

A data SLA is an explicit commitment, typically agreed between a data team and the consumers of a dataset, about measurable properties of that data: how fresh it will be (updated by 8am daily), how available (queryable 99.9% of the time), or how complete. It turns an informal expectation into something that can be measured and held to.

This differs from a data contract, which specifies the shape of the data itself, schema, field types, allowed values, whereas an SLA specifies operational commitments about delivering that data: timing, uptime and reliability rather than structure. In practice the two are often defined together and enforced by the same workflow orchestration tooling, which can alert a team the moment a pipeline is at risk of missing its SLA.

Data SLAs matter because they make reliability a shared, accountable target instead of a source of recurring disputes between producers and consumers of a dataset, and they let downstream teams decide how much to trust and depend on a given pipeline. The common pitfall is setting SLAs that are aspirational rather than achievable given the pipeline's actual data freshness characteristics, which erodes trust every time the target is missed, or defining an SLA with no monitoring behind it, so violations are only discovered by the consumer.