The 15-row sample above is rendered from the first query. It covers every column except the technical ingestion fields and `bedtime_phone_minutes`, which I left out because it is person-identifying and should only be aggregated. The table has one row per user (`USR-xxxxx`). **What the sample shows** - **Demographics:** ages run from 18 to 56 and the rows include Female, Male and Non-Binary users. - **Segments:** the sample mixes occupations (Corporate 9-to-5, Healthcare / Shift Worker, Freelance / Creative, Remote Tech) and chronotypes (Night Owl, Intermediate, Morning Lark). It also covers several bedtime apps, such as TikTok / Reels, YouTube, Instagram / Reddit, Messaging / Chat, Streaming and News / Reading. - **Sleep:** `total_sleep_hours` ranges from 4.45 to 8.04 in these rows. The shorter sleepers tend to have more alarm snoozes and higher `next_day_fatigue_score`. For example, USR-05967 slept 4.45 hours, snoozed 4 times and scored 5.5 on fatigue. USR-02102 slept 8.04 hours, snoozed once and scored 1.0. - **Category label:** `sleep_debt_category` is derived from sleep hours and is already in the data. **How well it represents the full 8,500 rows** Across the whole dataset (query over the gold table by sleep-debt category), the categories are Moderate Debt 52.5% (4,462 users), Mild Deficit 23.6% (2,004), Optimal Recovery 16.3% (1,387) and Severe Sleep Debt 7.6% (647). The sample leans toward Mild Deficit, with 8 of 15 rows, against 23.6% overall. It has 5 Moderate Debt rows, 1 Optimal Recovery and no Severe Sleep Debt. So it is a good picture of the columns and value formats but not a proportional one. A stratified sample by `sleep_debt_category` would be needed if you want one that matches the category mix. Severe cases look quite different from the rest: on average 3.76 hours of sleep, 5.88 snoozes and a fatigue score of 9.58. I can pull a stratified sample, or a sample for a specific segment, if you want one.