Comment by cimi_ | Hacker News Reader

Comment by cimi_ | Hacker News Reader

From the the linked post in the article (https://www.wsj.com/articles/facebook-overestimated-key-vide...):

> Ad buying agency Publicis Media was told by Facebook that the earlier counting method likely overestimated average time spent watching videos by between 60% and 80%, according to a late August letter Publicis Media sent to clients that was reviewed by The Wall Street Journal.

From the FB post (https://www.facebook.com/business/news/facebook-video-metric...):

> The metric should have reflected the total time spent watching a video divided by the total number of people who played the video. But it didn’t – it reflected the total time spent watching a video divided by only the number of “views” of a video [...].

How does this work?

They explain that a "view" is when the user watches the video for more than 3 seconds.

View time is probably highly heavily skewed towards plays that last less than 3 seconds. (also because facebook autoplays videos without any pause time)

If 60% of viewers play the video for 0-3 seconds, and only 40% view the rest of the video, you're grossly overstation the average view time.

Removing users with less than 3 seconds seems 100% intentional (to avoid exposing the fact that most viewers are not really viewers, they immediately dismiss the video), not an error.

Thanks, this makes sense. If you think of this in terms of distraction, it's huge.

Right on.

In other words, fb measures were correct for "people watching videos for 3 seconds or more". But they didn't: 1. Qualify their numbers with that. 2. Disclose how many people were NOT watching videos.

Without these two critical details, the numbers represented "all Facebook users".

No, they were still incorrect for "people watching videos for 3 seconds or more." For that to be correct, the total view time (numerator in the figure in the article) would have to be view time only for those users who watched >3 seconds.

Reply on news.ycombinator.com