Hacker News new | past | comments | ask | show | jobs | submit login
Median YouTube views/subscribers ratio – 30% (among channels I'm subscribed to) (github.com/efojs)
3 points by efojs on Feb 11, 2020 | hide | past | favorite | 2 comments



Recently I found a cooking channel with 139K subscribers, but less than 500 views per new video.

I was confused and still don't know what's the deal, but to dig deeper firstly wanted to check if this channel really stands out and find:

— if it's unusual to have <500 views per video for 150K channel,

— if there is any visible relation between subscribers and views count.

So I used YouTube API to fetch data about all 346 channels I'm subscribed to and found that:

— some channels provide no stats, so after filtering them out only 333 left,

— having less than 500 views per video for 100K+ channel is unusual, and

— that in general 30% of subscribers watch new posts.

If you want to repeat the exercise with your channels, you'll need some basic Git and Python experience and get YouTube Data API key.

Post with more details on GitHub: https://github.com/efojs/yt_views_subs


Back on Google+ a "views per follower" (VPF) metric was in vogue for a while, see: https://globerunner.com/introducing-new-google-plus-quality-...

A few things stood out:

- VPF was typically low. Ghost town...

- High-follower counts typically had low VPFs, lower-but-modest follower counts: higher. This follows from an attention basis: low follower counts tend to be more directly engaged.

- Actual values were all over the map, and informal sampling tends to be highly biased.

- Useful sampling probably starts with at least both of most-heavily-viewed channels and most-heavily-subscribed channels. Those are the ones which are succeeding by at least some metric. You'd likely also want to include most-viewed videos, though one-hit-wonders can also distort results. (Tom Scott's observation that you don't want instant success, but more of a slow burn, is very apt.)

Again, from G+, here looking at Communities, what I found was that it was not the total number of members that mattered most, but the rate of new content -- and no, more isn't always better, but none is fatal.

Looking at visible metrics (community name, subscribers, visible recent comments, plus-ones ("likes"), and re-shares, correlation was highest with more posting activity, at least to a point (based on the visible rate of new posts).

Large groups with no posts => no activity.

Small groups with relatively frequent posts, often => high activity.

After a point, size seemed to count against communities (~10k members or so).

This is based on a comprehensive dataset of summary data from all 8 million G+ communities, provided me by Friends+Me during the G+ shutdown. See:

https://social.antefriguserat.de/index.php/Migrating_Google%...

Also:

Activity crossplots: https://joindiaspora.com/posts/13680950

Membership distribution (log-log): https://joindiaspora.com/posts/13738258




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: