Not all users are alike. In SocialFlow's thorough study on news consumption and Twitter, there is a clear difference between those who interact with content, and those who read it attentively.
Back in May, when we analyzed the viral spread of news about Osama Bin Laden’s death, we were impressed by the speed and scale at which news could break on Twitter. Yet even with the increased popularity of Twitter as a means of news consumption and dissemination, very few studies have been focused on audiences – who are these people that follow news on Twitter? We wanted to understand how different audiences consumed and rebroadcast messages news organizations were sending out, whether there were substantial overlaps between their audiences, and how their behavior differed.
In the following study, we analyze Twitter audiences of Al-Jazeera English, BBC News, CNN,The Economist, Fox News and New York Times, seeking to understand the makeup, behavior, interests, similarities and differences between networked audiences. We looked at measures of engagement such as the number of clicks shared urls receive, and the number of times content is retweeted by audience members. We visualized the content being clicked on in comparison to the content being shared. Some of our findings:
Wondering how many people follow both BBC News and CNN? Or what the average number of friends and followers are for the audience of the New York Times? Use the interactive Venn diagram below to explore this data.
Some interesting facts from the data:
There’s much more data available about intersection groups. Let us know if this interests you and we can put some more of it online.
Whether you’re a publisher, brand or individual using the web, building up skills to attract and attain attention within social media spaces requires that you engage and get familiar with your audience, followers and fans. Social media, specifically Twitter, has become a critical tool for publishers to connect with their readership, scaling networked information dissemination to levels that were simply not possible beforehand. This shift from a broadcast to networked means of communications completely alters publishing dynamics, requiring entities that have previously controlled their content publishing channels to navigate diverse audience types within a single digital space. Multiple audiences are flattened into one group, a phenomenon described by danah boyd and Alice Marwick as ‘context collapse.’ While this raises the complexity for interaction, social media brings the promise and possibility of learning from information to which previously we’ve never had access. Digital breadcrumbs coming from user interaction online shed light on our digitally networked audience – revealing its makeup, passions and interests.
Fox News, CNN, BBC News, New York Times, Al-Jazeera and The Economist are some of the most read and trusted media outlets. Yet with the ascendant popularity of Twitter for news consumption, very little research has been done on who these audiences are and how they compare to one another. The study is based on bitly and Twitter data from over 20 million tweets posted by some seven million users who follow these accounts on Twitter. We differentiate between types of user engagement such as clicks and retweets, and show how they vary based on the audience. An active audience or a large following does not necessarily promise an engaged one. Additionally, maximizing clicks might not always be the best strategy.
Everyone wants an engaged audience. Yet there are numerous types of engagement, and each comes with a potential consequence. Would you rather your audience always pay attention to your content, but never repost or share it? Or is it better to have followers that rarely click on your links, but constantly share your content with their networks? One drives traffic to your website, but the latter builds trust within your network. Retweets extend published content to new circles, building trust and brand awareness over time.
Our data shows that the larger an audience, the harder it is to maintain high levels of engagement metrics such as clicks and retweets. Finally, we see clear content-based and behavioral differences between audiences: users choose to follow news accounts based on the type of content being posted by the account. Thus news media sites that surface similar type of content see a larger overlap within their audience.
Below is a network graph of topics discussed by the Twitter followers of Al-Jazeera English (@AJEnglish) over a period of an hour. Topics appear as larger nodes the more prominently they are being discussed, and are organized such that related topics appear closer together.
We see a highly sectional network with the top right largely dedicated to European topics, the top left African, the bottom left Mideast and North African, and the bottom right American. There is no one dominant topic but rather there are multiple conversations happening at the same time.
By zooming into the graph, we can see the specific regional topics of discussion. Here’s a segment – from the bottom left – mentioning Tahrir square along with other terms related to the Egyptian revolution of January 2011.
If we move to a different section of the graph, we identify a Japanese contingent, focusing on the Fukushima nuclear meltdown in Japan, using words such as jishin (earthquake) andgenpatsu (nuclear power plant).
And another part of the graph – top left – is also very different, focusing on Africa Day and the separation of Sudan.
Interestingly, the dominant topics that bubble up have to do with newsworthy events, both global and regional. However, if we dig in deeper, we quickly find topics ranging from the latest Oprah update on the Schwarzenegger scandal to people’s obsessions with poetry.
An audience is made up of many people having multiple discussions at any given point in time. The ability to see these discussions ebb and flow in real-time gives an individual the ability to better interact with said audience. We all build mental models in our heads, imagining the people who make up our audience, trying to gauge their interests and how to interact without spamming or boring them. However, as our audience grows in size and diversity, the sheer complexity is just too much to keep in one’s head. Visualizations like the one above show a clearer picture of the conversation space as a whole, and let us zoom into specific sections.
One of the most interesting characteristics of these graphs are the apparent differences between Fox News and Al-Jazeera’s audiences. Al-Jazeera’s network is indicative of a geographically broader conversation of news topics, prominently featuring terms like Jan25, Pakistan, Africa, while Fox News’ graph is much more domestic in scope, featuring terms like tcot, ObamaCare, and fox25. While one may assume that the audiences following these accounts are of a particular political frame of mind, it is important to establish this as a data-backed assertion.
Of particular interest is the lower-middle section of the Fox News graph – the change in edge colors and shading from orange to brown is indicative of a conversation that shifts away from topics and toward specific actors on the network. The accounts in this area, such as JoeNBC, VotingAmerican, slkbrooke, and PL1776 are all established conservative accounts and command a fair amount of attention from Fox News’ followers. By analyzing the relational hashtags surrounding each term, we can get a sense for the topics that influence and are influenced by the accounts. In this case, #p2, or Progressives 2.0, seems to be an ideal hashtag to use for targeting highly connected conservative accounts on Twitter. Adding the right hashtag at a given time increases the visibility of a tweet to additional audiences who are searching on that tag. #p2 is relatively underused, yet closely connected to major hashtags such as #tcot along with a number of influential Twitter accounts.
People are enticed to click on content through many means, invited to explore a recent development relating to one’s interest, a recommendation from a friend, or even a manipulation of language and words. By mapping out the top articles that users from each audience clicked on we see clear differences between the user groups. Below is a ‘tag cloud’ representation for four of the accounts featuring top keywords that drove the highest levels of traffic from Twitter to their website.
While there are overlaps between some of the accounts, the identity of each media outlet is evident through the data. The New York Times’ audience enjoyed stories on sports and education this particular day, while followers of Fox News consumed content surrounding the Casey Anthony Trial coverage. Some topical overlap includes the Joplin tornado, the IMF chief scandal and President Obama’s UK speech. The Economist’s audience clicks on content that is quite different from the others, including topics that tend towards the niche that the account serves.
What do the numbers look like?
For every URL shared by The Economist on Twitter, we see more clicks on average, and a substantially higher median compared to the rest. Even though the size of The Economist’s audience is less than a third of the New York Times’, it is generating hundreds more clicks per shared link. This high proportion of clicks per tweet is suggestive of a high level of attention paid to the content shared by The Economist from its audience. The lack of a major disparity between the median and average indicates that this high level of attention is relatively consistent throughout the data, rather than exposing a few highly-clicked, outlying tweets driving up an average for a majority of underperforming tweets.
Next we look at both clicks per tweet and total retweets normalized by audience size.
The Economist still has the most engaged audience in terms of content clicks and, relative to its size, Fox News is a close second. While Fox News does not have the highest nominal click rates, it is generating a substantial volume of clicks given its number of followers. One would expect an active Twitter audience such at Al-Jazeera’s to click on content at a higher volume, yet in this case, Fox news performs much better.
This raises some interesting questions. Why do we see less clicks coming from an audience that is supposedly more advanced and passionate on Twitter? Perhaps Fox News has a better recipe for generating provocative titles and attracting the attention of their followers, while Al-Jazeera has a trusting audience that generally retweets its content, but doesn’t necessarily click links to the shared articles.
One is not necessarily better than the other. Trust and engagement are comparative to future investment in a brand; people who think highly of you will actively recommend you to their friends and colleagues. This is obviously productive over time. Yet, click-through metrics are extremely important in order to maintain earnings. There’s no easy answer to this question, as noted by a recent discussion started by Clay Shirky describing the need for new types of funding models.
The elements of successful audience development on Twitter are clear. Having realtime information about topics being discussed by a networked audience at any given time yields high potential for engagement. Context matters immensely on Twitter, especially as audiences are transient and quick to jump from one conversation to another. One cannot expect to get attention at any wanted time. Knowing when an addressable audience is available and what topics they’d like to engage in is key to earning their attention.
Further, businesses need to take advantage of the benefits that real-time data unlocks. Unlike TV, radio, or even most spaces on the web, the actions audiences take on Twitter have a measurable impact that can be assessed in real time. This data is extremely powerful and should be used to inform content development strategies and marketing planning.
Historically we’ve used intuition to interact with our imagined audience, keeping mental models in our heads of what we think will be most entertaining and interesting. With the advent of social network spaces, we can leverage available data to power thoughtful decisions on ways to interact with an ever-growing networked audience. Rather than using simplistic measures such as audience size or randomly choosing topics, there’s great potential in parsing real-time signals, identifying prominent discussion topics and getting to know an audience based on past actions – what they click on, retweet and post.
Using real-time signals, publishers can make wiser decisions on what to post and when. We showed how topical network graphs can be used to identify different conversations taking place and how certain hashtags can be used to reach a certain audience. We also showed how click maps provide a reflection of oneself: what one’s audience finds compelling out of the total content posted. Using these two approaches, one can identify existing opportunities to match content with an audience.
Finally, it is important to think deeply about the significance of engagement metrics. While clicks bring immediate returns, retweets and other forms of audience participation raise trust and brand awareness, both imperatives for maintaining sustained growth. A high number of followers is not indicative of an engaged audience; a high click-through rate doesn’t necessarily yield other engagement metrics such as retweets and new followers. By paying attention to long established demographics, collective audience behavior and the mercurial and fickle moment-to-moment signals, we step away from conjectures, generalizations, and assumptions, and leverage the audience itself in determining how best to interact.
This study was originally published on SocialFlow.
Full topical network graphs are available for download: AJEnglish, FoxNews. Feel free to explore and use as visuals. But please remember to give SocialFlow attribution. Network graphs lovingly made with Gephi.
Sorry, there was no activity found. Please try a different filter.