Lies, Damn Lies, and Online Video
And recently I've come to find the off-site -- or embedded autoplay -- as a particularly heinous malodorous video view. Why? Because many embedded autoplay videos often get streamed to pages that have zero content relationship to the video served, and often play below the fold with audio turned off.
But it seems like I'm pissing into the wind here, because according to a new study from video ad-network and metrics company TubeMogul, video aggregators are even more likely now to count an embedded auto-initiated play start as a view.
But first, a bit of history. Tubemogul put together a widely circulated study back in 2008 that showed that almost every online video site counted a view as simply a play-start. The only exception, Dailymotion, waited until at least some of the video had been viewed to call it a "view." And a handful of sites were checking for repeated viewing per unique IP within a set amount of time, which keeps videos views from incrementing when a user simply reloads a page multiple times (the Avril Lavigne problem).
But based on Tubemogul's most recent study, it's gotten even worse. Today, everyone but YouTube considers an off-site autoplay view as a view, and nearly all will also count a page reload as a view too. Only Blip, Metacafe and Dailymotion limited views to one per session, and only YouTube failed to count an embedded autoplay view as a full view.
Tubemogul spokesman David Burch said the results are disappointing. "Now more than ever, a 'view' on most sites just means that a video stream started to play -- nothing more, nothing less." Even worse, said Burch, "the big shock is that all sites but YouTube are counting embedded views set to autoplay."
I share Burch's frustration with counting embedded autoplay views as real video views. It simply encourages, and even legitimizes sleazy practices that my friend Todd Sacerdoti over at Brightroll calls "Stream Fraud." I'm similarly disappointed in the widespread practice of counting multiple viewings of the same video by the same IP address (usually equivalent to the same person) as a view as well. Sure, very infrequently those views are valid, but mostly they are indicative of stream fraud too.
But recently I've begun to soften my stance to play-starts -- whether user initiated or autoplay. It's clearly not a video view in any sense of the word, but it is something. And that something actually hearkens back to a traditional TV metric, called OTS, or "Opportunity to See." I discovered the view to OTS connection from Anne Drake, the head of research and analytics at Kantar Video. According to Anne, OTS simply measures whether a piece of video (typically an ad) was actually delivered to a physical TV set, not whether someone actually saw it. And that makes sense, if you think about pre-DVR TV, where viewers often used ad breaks to grab a snack, use the bathroom, or annoy their little sister.
OTS is applied to many types of media, including billboards, magazines, and certainly ought to be applied to internet video as well. But the play start -- or OTS -- is not enough. With all the tracking capabilities inherent with internet-delivered video, we should be supplementing play starts with more data around duration of view, percent viewed, autoplay on and offsite, embedded plays and more. According to Burch, "measuring web traffic is about much more than counting site 'hits,' and this is even more true for video." And with all that data available, I agree with him that advertisers should "demand much more detail, given tracking technology's evolution (over) the past two years."
Definitely check out the new Tubemogul research, and please leave a comment here and let me know what you think about play-starts, autoplay, session blocking and any other thoughts about the best way to count a web video view.
|ABOUT THE AUTHOR|
Jim Louderback is CEO of Revision3 Internet Television in San Francisco.