Snowflake tags as schema-level key-value labels for governance

Snowflake defines a tag as a schema-level object assigned to another Snowflake object with an arbitrary string value, stored as a key-value pair. The tag name must be unique in the schema; values are always strings, with multi-value tags when more than one string is needed. An object can hold multiple tags, and one tag can cover different object types at once, with duplicated or distinct values such as a shared cost_center of sales or per-table values like engineering and finance.

After tags are defined and assigned, teams can query them to monitor object usage and support auditing and reporting. Snowflake-provided tags cover common governance cases. Highlights include defining a tag once and applying it widely, inheritance from parent securables to children (for example table to columns), configurable automatic propagation, replication of tags and assignments from primary to secondary databases, and either a tag_admin-centered model or team-applied tags with admin-controlled naming.

For shared meaning across systems, tags act as governed labels on warehouses, tables, and related objects rather than buried in ad hoc naming. Inheritance and propagation keep classification attached as objects relate in the hierarchy. Queryable tag assignments give data and AI teams a consistent handle for compliance reporting without inventing parallel classification schemes outside the platform.