```json { "table_name": "availability", "primary_key": "avail_id", "candidates": [ { "rank": 0, "partition_key": "beer_id", "partition_key_label": "E", "partition_key_reasoning": "RPUV=1.56 indicates each beer has ~2 availability records on average. With 75,415 unique beers, this provides excellent distribution. Top 3 coverage is only 0.11%, showing no concentration. Semantically, beer_id is the primary entity whose availability timeline we track across different places.", "clustering_key": "created_at", "clustering_key_label": "S", "clustering_key_reasoning": "Monotonicity=0.980 confirms strong temporal ordering. This is the event_time axis identified by the classifier. Sorting by created_at within each beer_id groups all availability updates for that beer chronologically.", "world_line": "beer_id", "session_column": null, "confidence": "HIGH", "notes": "This is the primary perspective: tracking each beer's availability history across venues over time. The event records when/where a beer becomes available." }, { "rank": 1, "partition_key": "place_id", "partition_key_label": "E", "partition_key_reasoning": "RPUV=10.80 means each place has ~11 availability records on average. With 10,495 unique places, this is a viable alternative partition. Top 3 coverage is 3.97%, showing reasonable distribution. Null percentage of 3.52% is acceptable. Semantically, place_id represents the venue whose inventory timeline we track.", "clustering_key": "created_at", "clustering_key_label": "S", "clustering_key_reasoning": "Same temporal ordering axis. Sorting by created_at within each place_id groups all availability updates for that venue chronologically, showing what beers became available when.", "world_line": "place_id", "session_column": null, "confidence": "MEDIUM", "notes": "Alternative perspective: tracking each venue's inventory changes over time. Useful for place-centric analytics (what's on tap at this bar). Lower confidence due to 3.52% nulls and higher RPUV than beer_id." }, { "rank": 2, "partition_key": "user_id", "partition_key_label": "E", "partition_key_reasoning": "RPUV=114.42 means each user has ~114 availability reports on average. With 1,025 unique users, this creates larger partitions. Top 3 coverage is 8.00%, showing some concentration (user 76677 has 5,325 records). Semantically, user_id represents the reporter whose contribution timeline we track.", "clustering_key": "created_at", "clustering_key_label": "S", "clustering_key_reasoning": "Same temporal ordering axis. Sorting by created_at within each user_id groups all availability reports by that user chronologically, showing their reporting activity over time.", "world_line": "user_id", "session_column": null, "confidence": "LOW", "notes": "Third perspective: tracking each user's reporting activity. Useful for user contribution analytics but less natural than beer or place perspectives. Lower confidence due to higher RPUV and top-3 concentration." } ] } ```