**Metric: sleep latency (`sleep_latency_min`), the minutes taken to fall asleep. Category: `primary_bedtime_app`.** I chose these for three reasons: - **Metric:** The dataset's decision context is about reducing sleep latency and sleep debt. Sleep latency also moves much more across bedtime apps than total sleep does. - **Category:** The app is the classifier that separates the outcomes most clearly. - **Sleep duration is flat:** Average `total_sleep_hours` is nearly the same everywhere, from 6.17 h (TikTok / Reels) to 6.38 h (News / Reading). **How sleep latency varies by app** (average minutes, from the gold table by `primary_bedtime_app`): - **TikTok / Reels:** 46.1 min (median 43.0), the slowest, and also the largest group at 2,198 users. - **Instagram / Reddit:** 42.5 min (median 39.5) - **YouTube:** 39.1 min - **Messaging / Chat:** 38.3 min - **Streaming (Netflix/Hulu):** 36.4 min - **News / Reading:** 33.2 min (median 31.8), the fastest, and the smallest group at 596 users. The gap between TikTok / Reels and News / Reading is about 13 minutes, or roughly 39% more time to fall asleep. Next-day fatigue follows the same order: 4.16 for TikTok / Reels against 3.33 for News / Reading. **Caveats** - The data is observational, so this shows association, not that the app causes slower sleep onset. - Age is a likely confounder. TikTok / Reels users are youngest (average age 31.1) and News / Reading users oldest (38.0). - I did not control for bedtime phone minutes, caffeine or other behaviours. ```json {"charts":[{"kind":"bar","x":"primary_bedtime_app","series":["sleep_latency_min__avg"],"title":"Average sleep latency (min) by primary bedtime app"}]} ``` {"action":"final","content":"**Metric: sleep latency (`sleep_latency_min`), the minutes taken to fall asleep. Category: `primary_bedtime_app`.**\n\nI chose these for three reasons:\n- **Metric:** The dataset's decision context is about reducing sleep latency and sleep debt. Sleep latency also moves much more across bedtime apps than total sleep does.\n- **Category:** The app is the classifier that separates the outcomes most clearly.\n- **Sleep duration is flat:** Average `total_sleep_hours` is nearly the same everywhere, from 6.17 h (TikTok / Reels) to 6.38 h (News / Reading).\n\n**How sleep latency varies by app** (average minutes, from the gold table by `primary_bedtime_app`):\n- **TikTok / Reels:** 46.1 min (median 43.0), the slowest, and also the largest group at 2,198 users.\n- **Instagram / Reddit:** 42.5 min (median 39.5)\n- **YouTube:** 39.1 min\n- **Messaging / Chat:** 38.3 min\n- **Streaming (Netflix/Hulu):** 36.4 min\n- **News / Reading:** 33.2 min (median 31.8), the fastest, and the smallest group at 596 users.\n\nThe gap between TikTok / Reels and News / Reading is about 13 minutes, or roughly 39% more time to fall asleep. Next-day fatigue follows the same order: 4.16 for TikTok / Reels against 3.33 for News / Reading.\n\n**Caveats**\n- The data is observational, so this shows association, not that the app causes slower sleep onset.\n- Age is a likely confounder. TikTok / Reels users are youngest (average age 31.1) and News / Reading users oldest (38.0).\n- I did not control for bedtime phone minutes, caffeine or other behaviours.","charts":[{"kind":"bar","x":"primary_bedtime_app","series":["sleep_latency_min__avg"],"title":"Average sleep latency (min) by primary bedtime app"}]}