Updated on Feb 22: I just realized that we gave an incomplete projection at the very end of this post... see below for the rest (clarifications updates in bold italics, completed calculcation offset). -Matt C.
----
Late last week, our friends at comScore reported that, according to their respected Video Metrix service, US Internet users consumed a record 10 billion videos in December of 2007. comScore's Erin Hunter attributed this surge in Internet video consumption to the much-publicized Hollywood writer's strike:
“With the writer’s strike keeping new TV episodes from reaching the airwaves, viewers have been seeking alternatives for fresh content. It appears that online video is stepping in to help fill that void.”
Subsequent coverage of these figures by the widely-read IT publication InformationWeek went on to quantify things at a more granular level:
Video lovers watched an average of 3.4 hours of video during the month, representing a 34% gain since the beginning of 2007. The average online video lasted 2.8 minutes, and the average viewer watched 72 videos.
So let's do some speculative math here:
72 videos per viewer (on average)
times
2.8 minutes per video (on average)
equals
3.4 hours of viewing time per user (on average).
Is anyone really suggesting that the US Internet video audience is watching every second of every video they view? We don't think so, as we all know that this is not how viewers behave. Whether video viewers are on YouTube, FunnyOrDie, MTV, or your corporate training intranet, people watch pieces and parts of each video.
They stick around if it looks interesting, they disappear if it isn’t. They rewind to re-watch the compelling parts and they skip ahead to see what else worth watching might show up. In other words, Internet video viewers have a well-established tendency to interact with video while it's playing... so measuring actual viewing time isn't as straightforward as multiplying views by average length, but must also take into account observed audience engagement.
This is where Visible Measures, and our ability to measure audience behavior while a video is playing back, comes into play. In working with the early adopters of VisibleSuite, we have seen -- in no uncertain terms -- that the vast majority of viewers who start any given video (i.e. they press the 'play' button) don't watch all the way through to the end. If anything, an end-to-end linear play-through is the exception, not the norm.
And we get asked all the time to generalize about the viewing habits of the online audience... which is about as useful as generalizing about how people drive in the US. The answer, of course, is that 'it depends': it depends on the particular audience, it depends on the content, it depends on the player, it depends on the stream performance, it depends on marketing, and more. Our client experience shows that even a popular video can show large audience drop offs if the video is mis-named, aggressively promoted, or located on a frequently viewed destination.
So, what does the latest figure of 3.4 hours of viewing time per month per US Internet video viewer really represent? We would say that this represents the total potential viewing time per user. That is, the theoretical maximum. But reality is always less than the theoretical maximum, and is usually much less. How much less? Who knows. Until quantifying viewer engagement by measuring in-stream audience behavior is a generally accepted practice, the theoretical maximum is the best we'll be able to do. Hint hint.
In conclusion, and in the interest of stirring the pot, here's our back-of-the-envelope, only kinda SWAG estimate of the actual average viewing time per US Internet video user per video:
A little more than a minute.
71.4 seconds -- 1.19 minutes -- to be semi-exact.
Or 42.5% of the theoretical maximum per video.
----
Feb. 22 update: So, to compare this to the original viewing time per view calculation:
72 videos per viewer per month on average (no change)
times
1.19 minutes per video (42.5% of the theoretical maximum)
equals
1.43 hours of viewing time per viewer on average per month.
----
Interested to know how we came up with that number? Well, our estimates are based on the ~75 years of cumulative viewing time we've collected so far. Ask us a question and we'll email you with our thinking. :)
-Matt Cutler & Matt Reider