```json { "table_name": "availability", "table_type_tags": [ {"tag": "event", "confidence": 0.80}, {"tag": "dimension", "confidence": 0.25} ], "primary_type": "event", "insufficient_evidence": false, "entity": { "columns": ["beer_id", "place_id"], "unique_per_row": false }, "time_axis": { "columns": ["created_at"], "kind": "event_time" }, "reasoning": "avail_id is unique per row (Uniqueness 1.000), but the composite entity (beer_id, place_id) repeats across rows — beer_id RPUV 1.56, place_id RPUV 10.80. The event timestamp created_at (Uniqueness 0.980, Monotonicity 0.980) orders these repeats. A row exists because a user reported a beer's availability at a place — the action is the report itself, recorded by user_id. The boolean flags (is_in_bottle, is_on_tap, is_out) describe the reported state but do not accumulate into a tracked balance; replaying rows does not reconstruct a meaningful entity state beyond 'last reported status'. This is a log of occurrence reports → event. The unique avail_id per report gives a weak secondary dimension reading (each report is a cataloged thing), but the repeating composite entity with timestamped actions is the stronger structure." } ```