Online Video Viewership Claims Remind Me of Bob Dole

(yah!  This is my first Ad Age column – and is crossposted on their site.  COOL!)

Bob Dole famously cried “where’s the outrage” during his presidential run against Bill Clinton. I’d like to revise that, with a twist, when looking at some recent internet video viewership numbers – but I’ll add in “where’s the analysis” to “where’s the outrage”.

My outrage began when I read a Multichannel News story claiming that Univision’s web series “Vidas Cruzadas” had become one of the “three most-watched online properties ever”, with two million video streams served. The story was served up like fact, without any analysis whatsoever.

My beef: two million views, across 16 episodes works out to roughly 125 thousand views per episode. And if their views followed the typical online video trajectory, the first episode probably accounted for more than half of those total views, with the balance eking out 50,000 or less.

However you slice it, however, that’s hardly enough views to put “Vidas Cruzadas” into the top 100, let alone the top 3. Here at Revision3 we’ve got a lot of shows with more than two million lifetime views – and a couple that consistently do better than 150,000 an episode. Two million’s not bad, but it certainly is nothing to write home about – or to press release over.

But it’s not just Univision and Multichannel News that are off target here. Over the past few weeks there have been numerous other mysterious web video numbers bandied about, with little or no analysis or skepticism in any of the reporting.

Over at CNet, Yahoo called its show “Prime Time in No Time” “the most watched original show in the history of the internet”, claiming 280 million streams in the 20 months or so since it launched last March. I guess Yahoo (and Cnet) have never heard of Fred, Machinima or Smosh, 3 of the 17 YouTube channels – akin to shows in YouTube verbiage – that have each delivered more than 280 million views since they launched.

But wait, there’s more! Internet news site Mashable rolled out a “top 10 most watched” list with partner Visible Measures, claiming to show the most watched web series for September 2009. The numbers are impressive, with Fred coming in at number one with almost 21.5 million views, and Smosh with 13 million. How exactly are those views counted though? The Mashable reporter didn’t know, but was nice enough to introduce me to Matt Cutler, VP for marketing and analytics at Visible Measures. His response:

“For the charts we publish with partners like Mashable, we’re using publicly sourced data that is self-reported by the 150+ video sharing destinations that we cover with our Viral Reach Database.”

But just how realistic are all these “self-reported” views? Visible Measures follows the IAB’s standard reporting guidelines in its own internal metrics, which logs a view after only a paltry 3 seconds of watching. But those “self-reported views” from just about everyone else out there are far more generous.

According to a study from video aggregator TubeMogul, nearly every site – including Yahoo and YouTube – count a “play start” as a view, even if you only watch for a fraction of a second. And what happens if the same person stops, then restarts a video – because of connection or buffering issue? Nearly everyone counts that as yet another view, further spiking the numbers. And what about AutoPlay – where a video starts to play when you load a webpage, even if you didn’t ask for it? Those are routinely considered views as well, both on video sharing sites and on social networking services.

Take the Slide Funspace application for example, one of the top add-ins to Facebook. When a Facebook users loads up the application, a video starts playing automatically. During my last visit, episode five of what was recently touted as “Facebook’s First Hit Series” started playing immediately, without me asking for it.

Is that a view? Unfortunately, according to the currently accepted state of video measurement on the internet, the answer is “yes”. But should it be? I think not. It would be like channel surfing past “ESPN 8” on the way to Monday Night Football, and having Nielsen count you as a viewer of the Dodgeball Championship on “The Ocho”.

I don’t mean to disparage the programs I’ve called out. Most of them are very entertaining, creative and enjoyable to watch. But to call any of them a monster hit based on what appears be at least some inadvertent, or drive-by viewership seems a bit disingenuous. Or at least, as Dan Brekke, one of the best news editors I’ve worked with used to say, it’s “an opportunity for investigation.”

So if the reporters from Multichannel News, CNet, Mashable and MediaWeek won’t do the digging here, you’ll have to. Before signing off on an internet video campaign, make sure you get the answers to the following questions:

• How do you count views? Find out whether they count, or discount, autoplay views. Also ask how long someone has to watch before a view is counted as a view. For many, it’s simply a play start, other follow the IAB’s 3 second rule. These numbers are very important, especially because another recent study from TubeMogul found that more than 50% of online video viewers clicked away within a minute. Engagement matters, and three seconds does not an engaged experience make!
• How about multiple views from the same user? This is also a big deal, particularly for those unscrupulous video producers trying to game the system. According to Tubemogul, “it is common knowledge.. that YouTube caps views at 200 per IP address”. That means someone can hit play on a video 200 times, and log 200 unique views of a video. Multiply that by just a few users, and you have the online video version of “click fraud”.
• How many views do you get for each episode? Make sure you get a per-episode number, rather than an average for all episodes. Most scripted episodic series have a huge open, but subsequent episodes can tail off dramatically.
• Is viewership growing, shrinking, or steady? A mature internet video show with a committed audience should have a consistent per-episode viewership number. If it’s all over the map, or declining, it’s a sign of either artificially enhanced viewership, or a drive-by, non committed audience. Results are better when the audience is committed.
• How do you count unique viewers? It’s very difficult on the internet, especially in a hyper-distributed world, to figure out exactly how many unique viewers watch a group of episodes. Find out where those unique viewer numbers come from: are they unique web-page visitors for show and episode pages, actual unique viewers of all episodes, or something else?
• Where do the views come from? It’s important to know how many of those views happen via a producer’s website, versus distribution partners like YouTube and Metacafe. Sure, you can build engaging, repeat audiences on those platforms, but typically your most enthusiastic audiences will gravitate to the show’s homebase.
• Do you pay for any views, or are they all organic? It’s easy to buy video views, especially if you count autoplay, below the fold, three second experiences as a view. Some marketing is obviously preferred, but if more than 10% or so are purchased, that’s definitely a red flag.

Jim Louderback is CEO of internet video network Revision3. For more details on these issues, you can reach him at jim@revision3.com, or follow him on twitter – @jlouderb