Catching a Viral Video

2 minute read

Broxton, Tom, Yannet Interian, Jon Vaver, and Mirjam Wattenhofer. Catching a Viral Video. Journal of Intelligent Information Systems 40, no. 2 (April 1, 2013): 241—259. doi:10.1007/s10844-011-0191-2.

This paper investigates (i) the relationship between the socialness of a video and its popularity, and (ii) provides a ranking of blogs and Web sites based on their ability to spread viral videos.

Method

Social versus Non-Social Videos

The authors segment videos based on their socialness which is determined by classifying their referrer sources either as social or non-social:

  • social referrers: external links, embeds, unknown (the link is directly copied into the browser)
  • non-social referrers: YouTube internal (related links, featured videos, etc.), videos found by search queries

Staying power of viral videos

The popularity ratio (staying power) allows distinguishing between videos that are popular over a short time spawn (e.g. through social effects) and long-term popular videos. \[ \text{popularity ratio} = \frac{\text{Views in the second month}}{\text{Views in the first ten days}} \]

Ranking the potential of Web sites to produce viral videos:

a) remove Web sites that distribute popular and unpopular videos (e.g. Facebook) by computing a ratio $$r(u)$$ of popular $$V_p$$ to unpopular videos $$V_u$$ on the Web site $$u$$. $$W_{100}(u)$$ refers to the set of videos with at least 100 views coming from Web site $$u$$.

\[ r(u) = \frac{|V_p \cap W_{100}(u)|} {|V_p \cup V_u|} \]

b) compute the rank $$R(u)$$ of Web sites specialized on distributing popular videos based on the sum of view they generate.

\[ R(u) = \sum_{v \in V_p} \text{views(u, v)} \]

The authors then compare the obtained top-sites for distributing viral videos with the popular blog ranking from technorati and conclude that 49 of the top 100 video distribution sites where listed by technorati.

Observations

  • People are more likely to watch videos that have been endorsed by means of social networking sites, blogs, emails or instant messaging - if such videos gain popularity they are referred to as viral videos.
  • Viral videos tend to gain and loose popularity very quickly reaching their popularity peak after approximately five days, an effect that is also observed in other social media.
    • Only 5% of the top five stories on twitter remain top stories in the following week.
    • The same is true for 13% of the top blog stories and 9% of the top YouTube videos.
    • In contrast, 50% of the top five news stories remain top stories in the following week.

  • The majority of the popular videos (top 1% of videos) become popular through related videos and search
  • Viral videos have many views over a short period of time but loose too quickly on popularity for being included into the related list, that ensures a sustained number of views over time.