How To Get Your Content Shared in Social Media
Published: 22nd August 2013 Written by:
The idea for this investigation came when I saw Mashable’s Velocity Graph feature which they had launched in December 2012. It attractively displays the speed at which any of their posts is shared across the social media landscape and you can see the changes over time, most of them surging minutes after posting, others finding a second wind when hitting a new timezone’s breakfast crowd.
I thought it would be interesting to run an analysis on this data, to see which aspects of the content (subject, style, timing etc) affected the speed at which it was shared. All I had to do, I realised, was grab enough of the stories from this site to have a data set to analyse.
For the uninitiated, Mashable.com is the social media world’s most popular source of news and content. It covers internet culture relating to business, technology, entertainment, marketing, lifestyle and current affairs. According to its media pack, it has 25 million unique visitors per month and sees 2.5 million social shares of its content each month. As such, it publishes frequently enough and to a big enough social network reaction to allow analysis of a large amount of data in a short period of time.
I first decided on the variables that were likely to be significant and that we could practically record. I chose article category, ‘Article Type’ (defined as News, Feature, Review, List, Guide, Image, Video), time of publication (GMT), headline and social share data. A post could of course belong to several styles; a list of images, video news etc, so we tagged each post with the styles that applied to it.
We chose to snapshot articles at 18 hours after posting; long enough to see the article go round the world and short enough that they were still contained in the RSS feed where we were analysing them from. Articles were grabbed at hourly intervals every day for three weeks.
Once all the data was gathered, we ordered all the variables and began analysis. Due to limits on time and budget, we stopped at 3 weeks and 200 articles. Obviously, this could be taken further by anyone and Mashable themselves could do this with a much bigger data set.
In the end, the Velocity Graph analysis was cut from the study. Upon further investigation, it turned out Mashable uses the data from it to move the story around the site, promoting or relegating its prominence depending on how well it’s performing. Greater prominence for an article would accelerate its performance in social networks, so not knowing its changes in prominence meant we could not interpret anything from the Velocity Graph.
Notes on the Findings
The major social networks have distinct communities.
Facebook responds most to the Watercooler category – the more trivial, fun content. Twitter is all about Social Media which is understandable for the Mashable audience but all categories perform well. Could this be evidence of auto-sharing going on in Twitter where people RT stuff without reading it, simply to have relevant-looking content in their streams? Google+ are clearly most interested in Technology and Reviews do well, possibly as a result of the comments section attached to any update, allowing for public discussions. LinkedIn is overwhelmingly Business.
It’s Facebook that enables content to hit the huge numbers but they tend to respond best to the most trivial content. Visual content is king.
Interestingly, video’s popularity drops only on LinkedIn. Business people don’t have as much time to spend watching video maybe?
Time of day in this case is tied to US consumption.
Mashable publish more content during US hours, which gives those times a greater chance of one of the super-popular posts happening. No doubt, a post needs time to be noticed and it’s unclear how long Mashable allows a post to remain on the homepage before making a call on its popularity.
I’d like to invite Mashable to run this analysis now. It would provide a much fuller data set, weekend analysis and several other variables that weren’t available to us, being limited to manual recording of the data. I’m sure we’ve uncovered many major trends but there are quite likely many more findings lying within the data that will come out with a larger data set.
Embed This Image On Your Site (copy code below):