While I've never done A/B testing on YouTube specifically, essentially YouTube would recommended the same video with different thumbnails, measuring an equal sample of users for each. So if it was recommended to 300 users, there would be three groups of 100 all seeing different thumbnails, and then some of those users would watch some amount of the video. Sum up the total watch time within each group, and you can compare which thumbnail led to the most watch time between the three (e.g., if users clicking on thumbnail A watched the video for 20 hours collectively, while B had 16 hours, and C had 14 hours).
Metallicity
2024-07-24 06:26:10 +0000 UTC
What does the percentage represent? It's labeled as "watch time share" but I don't think I understand what that means. I'm not in the biz. Pretty interesting that such minor changes in the thumbnail can have an affect, though.
Bigsicko
2024-07-24 03:31:48 +0000 UTC
In this A/B test, users are randomly split into three groups, each seeing a different thumbnail. All those other factors, such as title and users' general moods, are the same for all three groups. That means the only explanation for differences in the numbers is the different thumbnails. (Plus some amount of randomness which dissipates with more users in the sample.)
Brandon Dilbeck
2024-07-24 01:05:56 +0000 UTC
I would click on the yellow simply because there was a character on it
Sushidashi
2024-07-23 23:59:19 +0000 UTC
I didn’t comment but I was thinking it. Yellow is the ugliest color next to orange (don’t at me) and it doesn’t contrast much with the shade of the background. The red is more striking. I love contrasting, saturated tones that draw my attention.
Veronica Sipe
2024-07-23 21:12:43 +0000 UTC
This is interesting, however I doubt the accuracy of this data as there can be several reasons why someone may not want to watch a video - besides just the thumbnail. And same goes for wanting to watch a video. It could depend on the mood, what they are going through in life and also the title of the video.
Tldr; I'm not sure that any of this matters. But it was fun to engage with you on it and watch the results lol.
DurtyDeeds 69
2024-07-23 20:31:33 +0000 UTC
It takes a big sovl to share the truth even when it didn’t break in your favor. You da real MVP, champ. Team C keep winning. Red Line Go Up!
Dat Guy
2024-07-23 20:29:42 +0000 UTC
But was I right in thinking it was because of the clear background? Further investigation is needed
Jon
2024-07-23 20:17:07 +0000 UTC
I didn't comment on the original post but that is the one I liked the best so I feel validated