Of the 22 columns, 1 is an identifier, 7 are categorical dimensions, 10 are numeric metrics, and only 1 is a true date/time. The remaining 4 are ingestion metadata. The counts come from the column profile and from the distinct-count checks in steps 2 and 3, which confirmed 8,500 rows and 8,500 distinct `user_id` values. **Identifier (1)** - `user_id` is unique per row (8,500 distinct in 8,500 rows, USR-00001 to USR-08500). It is the key, so don't aggregate it. - The profiler's "~7,297 distinct" was only an approximation. The exact check in steps 2 and 3 shows it is fully unique. **Categorical dimensions (7)** - `gender` has 3 values. - `occupation_type` has 5 values. - `chronotype` has 3 values. - `primary_bedtime_app` has 6 values. - `sleep_debt_category` has 4 values. It is derived from `total_sleep_hours`, so don't treat it as independent of that column. - `blue_light_filter_active` has 2 values. It is stored as a BIGINT 0/1 flag, so it is categorical despite the numeric type. **Numeric metrics (10)** - `age` ranges from 18 to 65. - `bedtime_phone_minutes` ranges from 1 to 180. It is a duration, not a timestamp. - `screen_brightness_pct` ranges from 10 to 100. - `caffeine_post_5pm_mg` ranges from 0 to 250. - `physical_activity_min` ranges from 0 to 112. - `sleep_latency_min` ranges from 6.0 to 123.3. - `total_sleep_hours` ranges from 3.2 to 9.8. - `deep_sleep_pct` ranges from 8.1 to 28.0. - `rem_sleep_pct` ranges from 9.6 to 27.0. - `morning_alarm_snoozes` ranges from 0 to 7. It is a small discrete count and could also be treated as ordinal. - `next_day_fatigue_score` ranges from 1.0 to 10.0. `bedtime_phone_minutes` is also personally identifying, so report it only in aggregate. **Dates/times (1)** - `_ingestion_timestamp` is the only real timestamp. It holds a single value (2026-10-01 03:08:19.556483) across all rows, so it is load metadata and can't support any trend analysis. - The dataset has no event date or time of any kind. `bedtime_phone_minutes` is tagged "temporal" in the card, but that is a mislabel. **Technical/lineage columns (4)** - `_batch_id`, `_source_file` and `_source_system` each have one distinct value, so they are constant metadata with no analytical value. - These sit alongside `_ingestion_timestamp`, which is counted above under dates/times.