```json { "table_name": "column_classifications", "table_type_tags": [ {"tag": "dimension", "confidence": 0.95} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["id"], "unique_per_row": true }, "time_axis": null, "reasoning": "The entity id has Uniqueness 1.000 (every row unique). The composite (classification_candidate_id, columns_catalog_id) with RPUV 14.96 and 1.19 respectively forms the natural grain — each row assigns a column_label to a specific column within a classification candidate. No usable time axis exists: _ingestion_timestamp is a provenance column (ignored per rules), and the high Monotonicity on id and columns_catalog_id reflects scan order, not semantic event ordering. Each row is unique information about one classification assignment → dimension." } ```