Completed and submitted a paper made from the experiment named weTest1. We used a variety of analysis. For example network visualization enhanced clusters and highlight patterns:
A few months ago, we had insights thanks to such kind of analysis, after a collaborative observation of pornhub [report page, SFW]. The Adult site tests related content on newly uploaded videos. We reused it on YouTube. If you want a takeaway of this paragraph: pornhub and youtube, acts differently in regarding of testing emerging contents. This at least, with our small sample of observation-contributions, and at the harsher closeness of youtube compared to pornhub.
Apply analysis on recurring videos, and separated what’s in the API, what with unlogged profiles vs logged profiles:
The 14 pages paper it is under peer review, this is the preprint[.pdf]. The data [files on git] CSV and JSON.