{
    "table_name": "bedtime_screentime_sleep_debt",
    "primary_key": "user_id",
    "candidates": [
        {
            "rank": 0,
            "partition_key": "user_id",
            "partition_key_label": "primary_identifier",
            "partition_key_reasoning": "user_id has 8,500 unique values across 8,500 rows (uniqueness 1.0, RPUV 1.00, 0% nulls). It is the entity identifier for this one-row-per-user profile table. The classification is dimension, so the primary_identifier label applies even though RPUV is 1.",
            "clustering_key": null,
            "clustering_key_label": null,
            "clustering_key_reasoning": "Dimension tables do not need internal sorting. There is no time axis; _ingestion_timestamp is a constant load-metadata column (1 unique value) and cannot serve as a clustering key.",
            "world_line": null,
            "session_column": null,
            "confidence": "HIGH",
            "notes": "Each row is a distinct user with demographic attributes (age, gender, occupation_type, chronotype) and sleep and screen-time measurements. user_id is the primary key. Low-cardinality columns such as occupation_type (RPUV 1700) or sleep_debt_category are attributes, not organizers. The _ingestion_*, _batch_id, _source_* columns are constant lineage metadata."
        }
    ]
}