```json { "table_name": "report_insights", "primary_key": "jobs_id,index", "candidates": [ { "rank": 0, "partition_key": "jobs_id", "partition_key_label": "E", "partition_key_reasoning": "jobs_id is the entity column identified by the classifier. RPUV=11.91 is excellent for partitioning (between 2 and 100). Top 3 coverage is 12.17%, well below 50%. Semantically, this groups all insight messages generated during a single analysis job execution together.", "clustering_key": "index", "clustering_key_label": "S", "clustering_key_reasoning": "index is the sequence axis identified by the classifier. Monotonicity=0.917 indicates strong sequential ordering. Each job produces an ordered sequence of insight messages (table_count, column_range, other) that should be read in order. Sorting by index within each jobs_id partition preserves the natural message sequence.", "world_line": "jobs_id", "session_column": null, "confidence": "HIGH", "notes": "This is an event table tracking insight messages generated per analysis job. Each job (jobs_id) produces a sequence of insights (ordered by index). The composite primary key (jobs_id, index) uniquely identifies each insight message. Partitioning by jobs_id groups all insights from the same analysis run, while clustering by index maintains their generation order." } ] } ```