Videos watched
Were ads
Plays were replays
Mostly music on repeat
Different channels
Timeline
Two homes, one watch history
Detected in Chicago
Detected in Vietnam
Language
Coming back to Vietnamese
Vietnamese
English
Korean / Japanese
Chicago vs. Vietnam
Homesick, in numbers
Channel loyalty
A long tail with a loyal core
On repeat
The replay habit
Rhythm
A night watcher
Sessions
Rare binges, big share
How it works
Method
1 · Cleaning
Takeout lists ads in the same file as real videos. The pipeline drops every entry tagged “From Google Ads” and anything that isn’t a watched video.
2 · Time zones
Takeout stores every time in UTC. For each month, the script tests Vietnam (UTC+7) and Chicago (UTC−5/−6) and picks the zone where I was least active from 2am to 8am. The result matches my real calendar.
3 · Language
Each title is classified with regex: Vietnamese diacritics, Korean Hangul or Japanese kana. Everything else counts as English.
4 · Sessions
SQL window functions (LAG plus a running SUM) start a new session after a 30-minute gap. The raw history never leaves my laptop.