YouTube users upload tens of thousands of clips every day, and only a select few go on to become true viral success stories. But what’s really going on when videos that feature things like a dog on a skateboard, otters holding hands or the evolution of dance go viral? Two Switzerland-based scientists have analyzed the popularity of nearly 5 million YouTube videos over a period of eight months to find out.
Riley Crane and Didier Sornette of The ETH Zurich found that 90 percent of all YouTube videos never get any significant bump, but instead just attract a steady flow of viewers. Meanwhile ten percent of them, according to the study that was just published in the Proceedings of the National Academy of Sciences magazine, exhibit “herding behavior,” meaning that the clips either become viral or get a spike in views because of being featured on YouTube or elsewhere. Further, viral videos tend to get more views on average than featured content or one-hit wonders.
Crane and Sornette were able to establish mathematical models to classify YouTube’s success stories into three different groups. The first are true “viral” hits that spread through blogs, emails and social networks. These videos become more popular over time, and their peak can be described as a form of a power-law distribution. Viral videos attracted an average of 33,693 views during the eight months the scientists monitored YouTube.
Videos that get featured on YouTube or that are connected to a current event show a different kind of popularity, peaking very rapidly and then slowly decaying, make up the second group. Crane and Sornette called these clips “quality videos” as they attract massive immediate attention because of their content and then become viral after the fact, meaning they’re still getting forwarded via email days or weeks after they hit the YouTube home page or are featured on a popular blog.
The third class are so-called “junk videos;” they show a sudden spike in activity that then quickly dies off. These clips become popular by chance or through spamming, but never trickle down through social networks. They receive the fewest views — just 16,524, on average, during the research period.
Do those numbers seem low to you? Well, that’s what’s really interesting about these models developed by Crane and Sornette: They actually work for less-popular content as well. Videos can go viral and show sustained interest within a smaller community without ever showing up on top of YouTube’s charts.
Other clips can get really popular really quick and then die off soon after. “This implies that identification of relevance — or lack of relevance — can be made for content that has mass appeal, along with that which appeals to more specialized communities,” the two scientists conclude in their paper.