In March 2013, at Microsoft’s annual research and development event TechFest, a new project was introduced that aimed to let “users interactively explore the full chain of events whereby individual news stories, videos, images, and petitions spread from one user to the next over a social network.” The program, in effect, aims to understand how content spreads through a social network such as Twitter. By aggregating large amounts of data and tracking how users share things on their Twitter accounts, ViralSearch turns the transmission of content into a visually friendly genealogy of media, which Microsoft terms its “virality.” The more descendants a video has, for example, meaning those who have shared it (which is broken up into generations, or subsets of users that represent one wave of shares) the more viral it is according to ViralSearch’s virality percentage. More than this, it actively differentiates between virality and popularity, by looking precisely at how the information is shared. As researcher Jake Hofman says,
This is what people sort of typically have in their mind when they think about one of these viral videos, but nobody’s really been able to actually look at the structure of these things to date. And so what we’re able to do is going through these billions of events we reconstruct these trees by looking at all the followers of everyone who adopts the content and using a large cluster to reconstruct these things and then a novel scoring method to actually distinguish this tree as being viral from just being popular.