🎬 Netflix Viewing Analysis

What two and a half years of my Netflix history says about how I actually watch, analysed with SQL and Python.

👉 Live site: https://c20744531.github.io/NetflixViewingAnalysis/

• 152 titles and episodes across 73 shows and films, from March 2024 to September 2026 (about 145 hours, estimated).
• Feast or famine: July 2026 alone accounted for 45 titles, 30% of everything. I watched nothing at all in nine of the 31 months.
• Weekend viewer: 63% of my viewing happens Friday to Sunday, and Saturday is my biggest day.
• Binge watcher: 16 binge sessions (three or more episodes of one show in a day). The biggest was eight episodes of The Night Agent on 3 July 2026.
• Crime & Thriller is my top genre, at about a third of all titles.
• Hollywood first, Nollywood second: after the US, Nigeria is where most of what I watch is made.

1. Cleaning in SQL (sql/01_clean.sql)

I loaded the raw Netflix export into SQLite. Netflix gives dates as text like 9/19/26, so I split them on the slashes and rebuilt them as proper ISO dates, and pulled the show name out of titles like The Night Agent: Season 3: Orion:

SELECT PRINTF('%04d-%02d-%02d', y, m, d) AS watched_on,
       title,
       TRIM(CASE WHEN INSTR(title, ':') > 0
                 THEN SUBSTR(title, 1, INSTR(title, ':') - 1)
                 ELSE title END) AS show
FROM dated;

2. Tagging

I hand-tagged all 73 titles with a genre, film or series, and country of production in data/title_tags.csv, then joined it to the history in SQL.

3. Analysis in SQL (sql/02_analysis.sql)

A recursive CTE fills in the months where I watched nothing, so gaps show up in the chart, and GROUP BY with HAVING finds binge sessions:

-- a binge is 3+ episodes of the same series on the same day
SELECT v.show, v.watched_on, COUNT(*) AS episodes
FROM views v
JOIN title_tags t USING (show)
WHERE t.type = 'Series'
GROUP BY v.show, v.watched_on
HAVING COUNT(*) >= 3
ORDER BY episodes DESC;

4. Pipeline and visuals

analysis.py loads the CSVs, runs both SQL files and exports the results to docs/data.json. The site draws its charts as SVG in plain JavaScript, and the world map is pre-rendered with d3-geo (tools/build_map.mjs).

A note on watch time: Netflix's export has no durations, so hours are estimated at 45 minutes per episode and 110 minutes per film.

python3 analysis.py                 # runs the SQL and writes docs/data.json
npm install d3-geo topojson-client world-atlas
node tools/build_map.mjs            # writes docs/map.svg
cd docs && python3 -m http.server   # open http://localhost:8000

Swap in your own NetflixViewingHistory.csv and add your titles to data/title_tags.csv to analyse your own history.

SQL (SQLite) · Python · JavaScript · SVG · d3-geo · Natural Earth map data · GitHub Pages