Andrés Gonzalez

Data analysis · In-house project · 2026

Why people leave, second by second

A system that crosses YouTube data with an analysis of 21 reference videos to decide with data what to change in each video: pace, figures, length and country.

21Reference videos measured
2Channels analysed
By block and countryRetention

The problem

YouTube gives you hundreds of numbers and no answers. Knowing a video retains poorly doesn't tell you what to change, and changing it blindly costs weeks of work.

What got built

Programs that pull each video's retention by block and by country, measure the anatomy of videos that work in the niche (cuts per minute, words per minute, spoken figures, faces on screen) and compare them with ours. Every conclusion states which data supports it and how strong it is.

What changed

It found what the eye couldn't: the narration was slower than every reference channel, twice the usual number of figures were spoken, and an English channel was being shown first to an audience that didn't understand it. Both channels' script rules come from here.

Built withPythonYouTube Analytics APIffmpegMatplotlib
Contact

Which task is eating your time?

Tell me how it works right now, in your own words and with no preparation. I'll tell you whether it can be automated, how I'd do it, and what isn't worth touching. If the answer is that it doesn't pay off, I'll tell you that too.

I reply within 24 working hours. The first conversation is to understand the problem, not to sell you anything.