**The most important numeric metric is `total_sleep_hours`.** I chose it on judgement rather than a computed importance score, for three reasons: - It is the outcome the dataset is built around. - `sleep_debt_category` is derived from it. - The other measures (latency, deep and REM %, fatigue, snoozes, caffeine, phone minutes) behave like drivers or consequences of it. **Highest records: 9.8 hours.** - Five of the 17 users at this value are shown in the first result. All are in the "Optimal Recovery" category, and four are Morning Larks. - They have short sleep latency (about 16–34 min), a next-day fatigue score of 1.0, and mostly zero post-5pm caffeine. - Across all 17 users the averages are 22.4 min latency, 1.0 fatigue, 0.12 alarm snoozes, 13.8 mg caffeine and 22.3 min of bedtime phone use. **Lowest records: 3.2 hours.** - Five of the 242 users at this value are shown. Four are Night Owls and all five are Healthcare / Shift Workers. - Fatigue is high (8.0–10.0) and latency is long (about 48–84 min). - Across all 242 users the averages are 70.0 min latency, 9.53 fatigue, 6.05 snoozes, 38.1 mg caffeine and 123.5 min of bedtime phone use. **Comparing the two ends:** - The short-sleep group has roughly 3x the sleep latency, about 9.5x the fatigue, about 50x the snoozes and about 5.5x the bedtime phone minutes of the long-sleep group. - Deep sleep % is 19.9 vs 23.7. REM % is nearly identical (18.7 vs 18.4). - This is observational data, so these are associations, not causal effects. **Caveats:** - **The floor looks artificial.** 242 users sit exactly at 3.2 hours, versus 17 at the 9.8 maximum. This suggests the value was clipped at a minimum, so the lowest record is a floor rather than a true extreme. - **The category is not a pure function of hours.** Users with the same 3.2 hours fall into two categories, "Moderate Debt" and "Severe Sleep Debt". The user-level rows show this directly: USR-00001 and USR-00003 are Severe, while USR-00025 and USR-00086 are Moderate. Another input, possibly fatigue or latency, must also feed the category.