For a two-week period, TubeMogul measured viewed-seconds for a sample of 188,055 videos, totaling 22,724,606 streams, on six top video sites (due to partnership limitations, we cannot disclose which sites). Limitations of this approach include the fact that we are only studying short-form content (full television episodes being streamed on sites like Hulu were not included), and that, because our code is not yet in their player, YouTube was not included (although many of the videos in the sample mirror identical videos on YouTube).However, another analyst found a different way to interpret the same data, to determine the ideal length for Web video:
The results are dramatic: most online video viewers watch mere seconds, rather than minutes, of a video.
Online video viewers' short attention span seems especially relevant to advertisers looking to strategically trim ad budgets as the economy contracts. For starters, it is clear that post-roll ads are of limited effectiveness. A three-minute video that has a post-roll ad in the final seconds, for example, will only be viewed by 16.62% of the initial audience, on average.
Another takeaway is that overlay ads should be displayed as early as possible in a video, preferably within the first few seconds. On YouTube, where most overlay ads appear at about 10 seconds in, 10.39% of a video's initial viewers are not likely seeing the ad.
Nothing too surprising about the numbers. They seem to support the notion that “shorter videos work better on the web”. But that conclusion depends on us using a metric that is based upon “complete views”. If we count only viewers who watch the entire video then a shorter video will always have “more viewers”. To retain even half the viewers the video has to be under a minute - 30 seconds would be even better.But because video is linear, does that mean producers will try to guarantee ad views by leaning toward the "inverted pyramid" print model (i.e. putting the important stuff at top, and then filling in the details)? If so, that puts the kibosh on good old-fashioned storytelling -- with a beginning, middle and end; and rising action, climax, denouement, conclusion. Wouldn't it be better to put a "hook" at the top, inject some foreshadowing, and let the suspense build throughout, to keep viewers riveted to the end?
But supposing we add another metric - say “minutes watched.”
Going from the chart:
1000 people watch a one-minute video, 464 are likely to finish - total 464 minutes watched
1000 people watch a two-minute video, 237 are likely to finish - total 474 minutes watched
1000 people watch a 3-minute video - total 484 minutes watched
1000 people watch a 5-minute video - total 471 minutes watched
1000 people watch a 30-second video - total 330 minutes watched
CONCLUSION: Ideal run-time for web video - 2.5 to 4 minutes.
Also: As with "browsing" a text article, are we taking into account folks who "fast-forward" through a video, sampling a little bit here and there, and skipping other parts? That's how plenty of people watch DVDs or shows/movies on Tivo. (Note: Some platforms forbid viewers from "sliding" through the story, usually at their own peril -- if you're impatient and can't fast-forward, you're more likely to abandon the piece entirely.)
And the one seemingly immeasurable factor that no study has yet taken into account is quality. Presumably folks will stick around a couple extra minutes for excellent videojournalism that holds their interest. That's what we at KobreGuide are banking on!
What are your online video viewing habits? We know you're busy, but under what circumstances would you watch a video that's longer than 3-4 minutes?