Data Mart

A data mart provides data for a defined business subject or group of consumers. A finance mart might serve daily revenue by currency, while a support mart serves case-level detail. Its defining feature is its focused purpose, not a fixed table size or a requirement that all rows be aggregates.

Specify what one row means, which records count, units, correction rules, and the time through which input is complete. A total labeled revenue is ambiguous if one team includes refunds and another excludes them. Reuse governed definitions while allowing different presentations.

A mart can be implemented with tables or views and may draw from a warehouse, lake, or other sources. A department-specific dataset is not automatically isolated from shared governance. Access permissions, freshness, and ownership still need explicit design.

Build and validate a replacement before making it the published version. If the candidate omits unknown amounts, successful SQL execution is insufficient evidence of completeness. Keep the previously accepted output when the publication check fails, and expose its age to consumers.

Reference: AWS: data marts. See Designing End-to-End Data Architecture for the worked example.