How do videos go viral? How do people share them through social networks? And what are the dynamics of ‘virality’?
Following the success of our Gangnam Style vs. Harlem Shake study (May 2013), Twitter UK invited us to explore four more big viral phenomena. The stories we selected have all been driven by video, and have been chosen to represent various types of video content:
- Commander Hadfield singing Bowie’s “Space Oddity” on the International Space Station (music)
- Dove “Real Beauty Sketches” for advertising (the most-watched advert EVER on YouTube)
- Ryan Gosling Won’t Eat His Cereal series of Vine videos, for serialised narrative content and mobile
- A grass-roots video of June’s protests in Izmir, Turkey, to provide an international and news dimension
Turns out there’s not a single model of virality. Instead, different types of videos spread in different ways. Different types of content appeal to different audiences and the structure of these audiences is what shapes the viral diffusion.
Understanding the dynamics of that spread – quantifying it using metrics, and digging into the influencers and demographics to understand some of the “how”, is what we’re going to talk about in this series of blog posts.
But first, take a look at the diffusion maps below, which show the pattern of tweets and retweets for each video (click to embiggen). It’s immediately clear that there’s something different going on for each. Some, like Commander Hadfield, have one big hub (Hadfield himself) driving half or more of the sharing. Others like Dove Real Beauty and the Turkish protest video show a constellation of many smaller influencers, each being reblogged by smaller groups. Read on and we’ll explain why.
(Blue nodes = tweeters. Yellow nodes = retweeters. Size = author visibility, i.e. estimated reach).
What we did
We used Pulsar’s content-tracking technology to collect and analyse any tweet containing a link to the videos we were tracking and understand how they were shared on Twitter. What we’re studying is content diffusion and content discovery – the way videos are shared, recommended, and retweeted until they become viral phenomena. Of course people share content in other ways too – not least on Facebook – and YouTube search is in fact the second biggest search engine in the world, after Google. But Twitter provides the strongest dataset for analysis, and its role as a “hub” for curating content from across the whole social web makes it an apt case study.
By me and Fran (@abc3d) of FACE, using our social media research platform Pulsar (PulsarPlatform.com). As presented at Strata, the big data conference, today. And there’s Part 2 to follow next week, digging deeper into how this content is spread through social networks.
How Videos Go Viral On Twitter
This network map video shows how people shared the Dove ‘Real Beauty Sketches’ video on Twitter.
We also analysed how the Commander Hadfield ‘Space Oddity’ video spread. Cooler subject, but sadly it made for a slightly less awesome diffusion video because it went viral so quickly. All the secondary nodes (influential sharers like Dara O’Brien and William Gibson) had tweeted it within 2 hours of the video’s release, meaning that the network’s pretty static from then on.
But our video did tweeted by Commander Hadfield himself, which was nice.
I am speaking at a thing:
On: Sunday 12th May, 6pm
At: Auto Italia South East, London N1C 4AE
The explosion of social networks means that the whole richness of human interaction - fights, breakups, love, loss, work, politics, art - are played out across network space by large numbers of the human population. Simultaneous with this is the rise of “Big Data” - huge data sets drawn from these networks that allow researchers (states, the public and private sector) to learn much about the populations under study. Virally and memetics, allegedly spontaneous phenomena can be analysed with the statistical gaze. And everything else.
Alex Andrews is joined by Jay Owens and Ed Manley to ask what does all this mean? What does it mean to be intensively connected like this - selves porous and attached at every waking moment, blurring the boundaries of self-performance work and leisure? How do the micro-banalities of every day life - from the daily commute to the walk in the park - play out on a vast aggregate macro level of Big Data? What is it to have a self on a social network, a data self?
I’ve written a new piece on my work blog:
"Attention marketing" is a current buzzword - and a big problem. In a world of near-infinite media abundance, getting anyone’s attention becomes a profound challenge. However much effort you put into creating content, you can’t guarantee it’s going to get an audience.
But brands and media agencies have come up with a solution. They’ve learnt to listen, to wait and react. Call it real-time marketing, call it culture-jacking – or simply call it “learning from Oreo” and their megawatt SuperBowl tweet in February that went viral. Let the audience grow itself – then tap into that. Any news event will do – maybe.
When former British PM Margaret Thatcher died, some brands immediately tried to jump on this bandwagon. This is what happened…
CHART OF THE DAY: The Difference Between What You Share And What People Want To Read
People are sharing science related content more often than other content category. However, science content is clicked on significantly less than the content that is less popular for sharing, like politics, news, or celebrity gossip.
There’s some rationale behind the split in what’s read and what’s shared. People will want to share science content because it makes them seem smart. People will click on celebrity content because no one will know.
Full Story: Business Insider
Slightly surprised politics & news are so heavily clicked-on - surely politics suffers from the same “Share because it makes you look smart, but secretly find a bit boring” issue that science & tech do?
This is awesome;
“Rubbish, reckon, blimey”
Read the full paper here: Word usage mirrors community structure in the online social network Twitter, by John Bryden, Sebastian Funkand Vincent AA Jansen. EPJ Data Science 2013, 2:3
The significant part of the article:
Royal Holloway has now filed a patent application to secure the intellectual rights to many of the techniques used in the research, which staff hope could have commercial applications. “There are numerous applications of our method, including social group identification, customising online experience, targeted marketing, and crowd-sourced characterisation,” the paper adds.
Here’s the patent application: WO2012/080707
So they publish in an open-access journal, and then try and lock down the method? That feels like a nasty move - I know nothing at all about patent law, but it’s fairly clearly a patent on the analysis method, not a proprietary commercial application. So what, any other researchers wanting to use these methods would not be allowed, or have to pay a fee? Fuck that - imagine if the chemists patented titration…
I’d also like to know about just how original their method really is - again, I’m certainly not an academic network analyst, but I get the gist of what they’re doing and it seems perhaps a unique combination of surely fairly standard network analysis moves? Perhaps it hasn’t been patented already because the compsci community know better than to lock down methods - but then along come a bunch of biologists…
Anyway, if you know more about this (Alex Leavitt, looking your way) I’d be very interested to hear your thoughts.
Final thought: it’s pretty old data - tweets collected in December 2009 from as far back as 2007. These communities do not necessarily exist now - pretty naff to use the word “tweetup” at SXSW this year, right? And the foodies are into kale now, not chard!
Seven seminal dilemmas about moral behaviour on the internet, from 1993 to today.
Featuring trolls, addicts, anonymity, suicide, community, and big data.
On my longform blog - Tumblr, pray forgive me! ;)
As Susan Sontag wrote in On Photography,
there is something predatory in the act of taking a picture. To photograph people is to violate them, by seeing them as they never see themselves, by having knowledge of them they can never have; it turns people into objects that can be symbolically possessed.
Sontag notes that this makes for a nostalgic gaze, an understanding of the world as primarily documentable. For those who live with status updates, check-ins, likes, retweets, and ubiquitous photography, such an understanding is near inescapable. Social media have invited users to adopt a sort of documentary vision, through which the present is always apprehended as a potential past. This is most triumphantly exemplified by Instagram’s faux-vintage filters.
There’s always tension between experience-for-itself and experience-for-documentation, but social media have brought that strain to its breaking point. Temporary photography is in part a response to social-media users’ feeling saddled with the distraction of documentary vision. It rejects the burden of creating durable proof that you are here and you did that. And because temporary photographs are not made to be collected or archived, they are elusive, resisting other museal gestures of systemization and taxonomization, the modern impulse to classify life according to rubrics. By leaving the present where you found it, temporary photographs feel more like life and less like its collection.
I replied to a question on Quora a few months ago - What are the *unofficial* rules of Tumblr (and what ought they to be)?
"Conventions, courtesies, credits, comments — what are the spoken and unspoken rules of the Tumblr community? Where is the community deficient and where does it excel?"
Here’s my take:
1. Be interesting
Have a point of view and add something to your posts - pull out what’s interesting about the argument, connect it to another idea, or just explain why you like it. Give people a hook to engage with to inspire questions or reblogged responses
2. Be honest
Text-based Tumblr likes personal accounts, confessionals and the exploration of personal vulnerability. (It’s the closest thing to the Heir To LiveJournal out there.) That doesn’t have to be your thing, but it’s definitely not a bragging, self-promoting kind of place.
3. Promote your friends
If a Tumblr or Twitter friend has written something good, give them a boost. Pro-social behaviour is good for community-building.
4. Credit your sources
Not widely done, but should be. Use both the linked URL and the body text to credit who made the image, what it’s called, and their website. (Don’t know? Use Google Image Search or TinEye.com.
5. Don’t be an asshole
Tumblr’s probably more sensitive towards sexism, racism, transphobia etc than other communities. Being a bigot isn’t cool.
ETA: Imperfectly phrased. Obviously don’t be a bigot because it’s wrong and hurtful, not because you’re concerned with “cool”. But, yes, community norms mean you ain’t going to be cool on Tumblr if you’re a hater.
What do you reckon?
A long tail of bots, read-onlies and the inactive - you can see as well as I do that the majority of likes & reblogs I get are from the same core of 20 people (thanks, guys ;). Hell, I got 50 follows this afternoon but only 5 of them were “real” (?) enough to show up in my notifications page… Weird.
95% a function of Tumblr tech listing, too - Stowe Boyd posted the exact same stats just a couple of days ago. Content is barely a factor at all - he’s produced 10x as many posts as my 480-odd, yet the follow-rate is bang the same.
ETA for 100,000 probably late March
Coca-Cola’s new ‘Coming Together’ advert
"There’s an important conversation going on about obesity out there. We want to be a part of the conversation."
I’m interested in this sentence - the positioning of the advert moreso than the actual content.
Now there’s a brand that’s bought into social media listening and the narrative that social has turned advertising on its head. No longer are brands able to broadcast what they mean at passive audiences - no! instead consumers are having conversations, and constructing webs of signification with each Like and Reblog, and all a brand can hope for is to humbly generate content in the hope that it might be used as currency in this attention economy.
"Consumers’ most valuable relationships are not with brands but with other consumers," says Mark Earls - hell, I’ve even used that line myself…
It’s “The Real Thing”, right? Well. I’m less interested in this as a truth-statement or Big Idea that Coca-Cola are reorienting their entire marketing strategy around; instead it’s something we need to examine as ideology. This is the story that brands are choosing to tell us. Us as marketing professionals, us as consumers.
Is it a defensive strategy - as in, this IS the new reality and their only hope of survival is to acknowledge the new paradigm?
Or does capital’s capture of signification continue unabated, and we are now only being told that we are free and powerful in the hope that we might slip the bonds of regulatory guidance and indulge in that refreshing icy-cold beverage we have always wanted, but, since 1989, have been replacing with mineral water and sports drinks?
"The crisis consists precisely in the fact that the old is dying and the new cannot be born; in this interregnum a great variety of morbid symptoms appear."
Antonio Gramsci, Selections from the Prison Notebooks, 1971.
In a nice turn of events, my post Reblogs and content sharing on Tumblr: a personal network analysis was followed up by Union Metrics CEO Hayes Davis. Union Metrics are the company who’re sitting on exclusive rights to Tumblr analytics, btw.
It’s heartening to see we are much on the same page. Davis actually identifies something very interesting: in some ways, Tumblr is actually better than Twitter for analysis of network effects:
I think one of the most fascinating things about Tumblr is that you can actually pinpoint who caused amplifications downstream. This is unlike Twitter where all retweets are “flattened” into a single list against the original tweet. We’ve seen this go absolutely insane; including things like a post by Barack Obama where reblogs occurred out to 109 degrees removed from the original post.
(With Twitter, about the best you can do is order the retweets in chronological order, plot that on a timeseries and by number of followers, and then hypothesise that “big name tweeps” who reblog the content generated any subsequent spike in retweets, or perhaps also clickthroughs. It’s a best-guess rather than definitive model - people could have seen the tweet via any intermediary’s retweet, but the retweet is only attached to the parent tweet without the ‘family trees’ or genealogies of influence Tumblr can provide).
I’m also very happy to be corrected: Union Metrics DOES give you the granular detail of individual Tumblr likes & reblogs. This is what’ll make it a useful tool for research purposes - many social media analytics platforms aren’t in fact designed this way. Adland and PR buyers seem to mostly want nice summaries of the impact their campaigns have had, with the analysis often heavily automated and pretty top-line. For research, however, you need to be able to cut the data in whichever way your hypothesis suggests - aka export everything & play around. Glad to know Union Metrics are meeting our needs too!
We go so far as to show you the top 100,000 posts, tags or users in the web interface (if there’s that much activity) and give you full-on exports of *all* of the data. And, as I said above, you can get that level of detail for individual posts as well.
*starts formulating New Year’s projects…*
Impressive Tumblrlytics from Alex Leavitt, a PhD student at the Annenberg School at the University of Southern California. And it’s a gif too :)
A network graph of 109,703 Tumblr users and a sample of the connections they make through reblogging (colored by algorithmic clustering which represents hypothesized topical communities). Large version here.