Please Note: This project is not currently active. The content on this site is provided for reference and is not actively maintained.

YouTube’s Hot Spot Analyses: Changing Behavior Raises Questions

by May 18, 2011

A few days ago, I wrote a post describing how three of our videos on the topic of ocean acidification measured up using the hot spots tool within YouTube’s Insights dashboard. Returning to this today, I planned to look at another batch of videos. When doing some comparisons, I realized that the hot spots graph had changed rather dramatically for one of the videos, “New Neighbor.” Here’s the before taken on 5.16.11:


Today, I found a considerably different hot spots trace:


It would appear that the video performed better early on in the second trace. There is still an upward trend toward the end of the 90 second video, although it is somewhat muted in the second, and more recent trace.

What could have caused this behavior? To the best of my knowledge, YouTube is determining hot regions within a video based on how far people make it through a video and how frequently they replay an area (here an answer in their help topic). I mentioned in my earlier post that the soft voices in the New Neighbor video could have led many users to replay a section. Regardless, the majority of the New Neighbor video remains below average in YouTube’s comparison to other videos of similar length.

In contrast to the hot spots, YouTube determines cold spots if viewers click away from the video mid-stream. As discussed on this YouTube forum page, there is a good bit of speculation as to what might cause a depressed hot spots trace early on. One reason is if they are set to auto-play when somebody follows a link to them. That behavior could yield a number of people turning away due to lack of interest once they realize what the video is about.

The best clue to this shift in behavior would appear to be that the first trace was taken at the beginning of brief campaign to increase the number of views of the video through promotions within YouTube. In fact, we increased the views from about 580 on 5.5.11 to just over 1,100 on 5.17.11. Incidentally, we started this promotion on about 5.3.11 specifically so that we would have a sufficient number of views to do the hot spots analysis. In the process of preparing this post, we realize now that videos auto-play via YouTube promotions, as well as via Google AdWords.

The fact that we were getting views organically before beginning the promotion campaign may have influenced the hot spots traces. One hypothesis would be that people clicking through YouTube promotions were, on average, more likely to watch through more of the video.

A final note is that it is counter-intuitive to have the traces for the hot spots analyses begin at any value other than “average.” This was pointed out in the forum post mentioned above. It would seem that all videos should be considered equal when the clock is at 0:00. I would expect to see the trace rapidly diverge from average as the counter increased, second-by-second, assuming a video’s performance deviated from the average.


Leave a Reply