I’ve written previously about the difficulties (impossibilities) of predicting whether content will go viral or not. Nevertheless, the prospect of your brand unleashing a Harlem Shake remains tantalizing enough for many to succumb to the charms of marketing agencies that guarantee viral results.
With an hour of video uploaded to YouTube every second though, the success to failure ratio is quite incredibly small, and it’s very likely that your viral masterpiece will languish in the depths, with a relative minority of people tuning in to watch.
Obviously a major part of predicting the virality of content is understanding why some content trends whilst other content tanks. Research led by Jed Hofman from Microsoft Research set out to measure the virulence of online content.
Hofman and his team recorded every tweet containing a link to the 40 biggest websites over an 18 month period. To put that into perspective, that represents over 1 billion pieces of content across sites such as YouTube and the BBC. They then fine tuned that list down to only the content that was linked to by at least 100 different feeds. That narrowed things down to 300,000 bits of content, which between them generated over 1.4 billion tweets.
By then analysing this content in detail it allowed them to build a picture of how content spread from person to person. How closely connected were people when they shared content for instance.
The end result was that each piece of content was rated according to their virality, with a final score awarded out of 100. To give this some context, only a handful of the billion or so pages they originally analysed scored 100 points.
So what interesting trends emerged from the report? Is it possible to predict whether something will start trending?
One interesting finding was that most content only survives for a few ‘hops’ in the network. In other words, content might be shared by some people, and some in their networks might share it too, but after that it tends to peter out. Content that persists through 20 generations is literally one in a million.
They also confirmed something I wrote about last year, ie that a lot of people are all too happy to share something on Twitter without actually clicking through to read/watch what it is they’re sharing. An example they provide is that of a YouTube video that went viral in terms of Twitter mentions, but the video itself only generated tens of thousands of views.
The research forms part of a new application by Microsoft called ViralSearch. The app aims to visualise the virulence of content in a couple of ways. The first is in the form of a family tree of all those who retweeted the content. The second is in circular form and shows the number of people reached on Twitter. Users can then drill down to compare the popularity and virulence of different tweets linking to the same content, say, or focus on various tweets posted by a single user.
It seems logical that such technology will eventually become part of Bing, allowing users to search for what’s hot online. Suffice to say though, whilst the research was able to showcase what content was going viral, they weren’t able to pin down exactly what it was that sent content viral. For the time being, that will remain very much a mystery.