**The most important metric is `total_sleep_hours`.** It is the central outcome in this dataset. `sleep_debt_category` is derived from it, and it moves with latency, fatigue, snoozes, caffeine and bedtime phone use. This is my judgement from the dataset's structure, not a computed ranking. **Highest records: 9.8 hours** - The maximum is tied across 17 users, for example USR-01351, USR-01363, USR-01396, USR-01715 and USR-02117. All five are in "Optimal Recovery". - On average they fall asleep in about 22.4 minutes. - Their fatigue score is 1.0, the lowest possible value in the data. - They average 0.12 alarm snoozes and about 13.8 mg of caffeine after 5pm. - Deep sleep is 23.7% and REM is 18.4%. - Average bedtime phone use is about 22 minutes. This is an aggregate, since individual phone-minute values are person-identifying. **Lowest records: 3.2 hours** - The minimum is tied across a much larger group of 242 users, for example USR-00001, USR-00003, USR-00025, USR-00070 and USR-00086. The five examples all work as Healthcare / Shift Workers, and four of the five are Night Owls. - On average they take about 70 minutes to fall asleep, roughly 3 times longer than the top group. - Their fatigue score averages 9.53 out of 10. - They average 6.05 alarm snoozes and about 38.1 mg of caffeine after 5pm. - Deep sleep is 19.9% and REM is 18.7%. - Average bedtime phone use is about 124 minutes, roughly 5.5 times the top group. **Takeaways** - The two ends of the sleep range differ sharply on behaviour and next-day outcomes. Short sleepers have longer latency, heavier phone use, more caffeine and far higher fatigue. REM share is essentially the same in both groups, and deep sleep is only a few points lower in short sleepers. - The lowest value is not mapped to a single category. The 3.2-hour users are split between "Moderate Debt" and "Severe Sleep Debt", while all 17 at 9.8 hours are "Optimal Recovery". So `sleep_debt_category` is not purely a function of sleep hours, and other fields must contribute to it. - These are observational associations. They do not show that cutting phone use or caffeine would raise sleep hours. The queries behind these figures were run on `raw.bedtime_screentime_sleep_debt`. The rendered result blocks show the individual extreme records and the group averages.