I pulled a 15-row sample, shown in the result block above. The sample is stratified across the four sleep-debt categories and the main segments. I left out `bedtime_phone_minutes` because the dataset card flags it as person-identifying, along with the ingestion metadata columns. **What the sample covers** - **Occupation:** Corporate 9-to-5 appears most often, with Healthcare / Shift Worker, Freelance / Creative and Remote Tech also present. - **Chronotype:** Night Owl, Intermediate and Morning Lark all appear. - **Bedtime app:** TikTok / Reels, Instagram / Reddit, YouTube, Messaging / Chat, Streaming and News / Reading all appear. - **Gender:** Female, Male and Non-Binary all appear. - **Sleep debt:** Mild Deficit, Moderate Debt and Optimal Recovery appear. Severe Sleep Debt does not, which is not surprising since it is only about 7.6% of the data. **How the sample compares with the full dataset** The full dataset has 8,500 users, one row per `user_id`. It splits into these sleep-debt categories: | Category | Users | Share of dataset | |---|---|---| | Moderate Debt | 4,462 | 52.5% | | Mild Deficit | 2,004 | 23.6% | | Optimal Recovery | 1,387 | 16.3% | | Severe Sleep Debt | 647 | 7.6% | Across these categories, average sleep hours fall from 8.35 (Optimal Recovery) to 3.76 (Severe). Over the same range, snoozes rise from 0.81 to 5.88 and fatigue from 1.08 to 9.58. **Patterns in the sample rows** - Rows with short sleep have higher fatigue and more snoozes. For example, USR-03199 slept 4.86 hours, had a fatigue score of 5.8 and 5 snoozes, and is in Moderate Debt. - Rows with 7 to 8 or more hours sleep show low fatigue and 0 to 1 snoozes. USR-02102 slept 8.04 hours, scored 1.0 on fatigue and is Optimal Recovery. - `blue_light_filter_active` is stored as 0 or 1. Many rows have `caffeine_post_5pm_mg` of 0, while others reach 113 mg. Fifteen rows is only a glimpse. Any conclusions should come from the aggregate tables, not from this sample.