Online video watching—is it massive or is it overblown? These days, it's almost a political question. Traditional television distributors assure us that 99% of video is watched on a television screen, and a massive observational study concluded that people say they watch more online video than they actually do in order to sound "new and cool." Consumers "cord-cutting" their cable subscriptions en masse? That's a "myth."
Meanwhile, YouTube's audienceand content library just keep blowing up the charts. And TV streamer Hulu, just a gleam in NBC and Fox's eye a couple years ago, is now a Super Bowl ad-driven household name and one of the biggest web success stories of the last year.
So what's really going on in Web video, and how do we measure it? Counting a video play is more complicated than a simple load of a Web page, and page view stats are still not standardized.
YouTube, for one, offers public view counts on a per-video basis, but not overall. Meanwhile, the best Hulu does is tell us which shows are most popular on a daily, weekly, and monthly basis—no view counts attached. Various sites also have different ideas of what qualifies as a video view; many say a view happens anytime a video starts, something in-stream advertisers must just love.
TubeMogul, an independent video syndication service, just made a play at aggregating impartial data by launching a free analytics service that measures videos directly as they stream on 15 different video sites—but not the biggest ones.
The two big-name online measurement firms that offer dedicated video research, Nielsen and comScore, both use a combination of panels (where thousands of willing participants install software that monitors their Web habits) and active measurement, where they integrate a "tag" or "beacon" into a video host's players.
Even though the two firms have similar methods for gathering data, their numbers can be very different. Some of that can be attributed to the fact that Nielsen does not count plays of video ads (though that must be hard to break out, given the growth of branded content and the fact that Web series are often distributed through ad units. But some disparities are harder to grapple with. Hulu, for instance, recently complained to Nielsen in a letter that was picked up by The New York Times. Nielsen showed Hulu losing audience in April, with a little more than a fifth of the audience that comScore counted.
Active measurement is the better method, because it helps accurately count views of embedded and protected content. But Nielsen and comScore only make the inclusion of embedded beacons available to paying clients of their market research. comScore says it's willing to include sites that are "really interested," while a Nielsen representative explained, "This granularity in reporting is only made available to clients due to costs incurred with server infrastructure, client support, QA and daily (monitoring)."
If you look at the latest video stats from April, comScore says Americans streamed 16.8 billion videos, up 16% from the month before, to a new high. Nielsen counts 9.5 billion videos viewed, down 2.3% from the month before. Directionally (and that may be the best way to look at this data), both sites agree that video viewing is going up, but the video audience as a percentage of the overall Internet audience (at least in the U.S.) is pretty stable at around 70% to 80%.
I recently ran into the inexact science of it all when I asked Nielsen analyst Jon Gibs why the number of unique viewers his firm measured for April was down 2.5% from the year before, to 117 million. He replied the difference was excitement about last year's election (which at that time was in the primary stage). "The election was driving up a lot of that consumption cycle," he said. I thought that was really interesting, so I asked for some follow-up numbers on the size of the video audience tuning in for election video. After looking into it further, Nielsen's team found that political video wasn't a huge factor, and a representative alternatively stated, "The 2.5% dip is simply due to minor fluctuations."