Oxford Study Explores Tweeter Reach & Influence
What can social media marketers learn from protests in Spain? Given that Twitter was used to get out the word about these protests, there’s plenty of insight to be gained. To that end, researchers at Oxford analyzed nearly 600,000 tweets over a 30 day period to learn how people were influenced to join these protests, […]
What can social media marketers learn from protests in Spain? Given that Twitter was used to get out the word about these protests, there’s plenty of insight to be gained.
To that end, researchers at Oxford analyzed nearly 600,000 tweets over a 30 day period to learn how people were influenced to join these protests, discovering important information on how tweets spread.
We’ll cover the study’s findings and clarify the technical language thanks to explanations given to Marketing Land by one of the study’s authors, Sandra Gonzalez-Bailon. You can read the study PDF here.
The Study Focused On Measuring Two Things
First is the spread of information, as measured by the number of people who could potentially have seen a tweet in their timeline. Second is the spreading of behavior, as measured by users who hadn’t previously tweeted about the protests sending a first message about the protest, which the study calls “activating.”
The Main Findings
- Spread of information — aka tweet/tweeter reach — depends not just on how many people follow a given person, but on how “central” these people are. What does it mean to be central in the network? You need to be “connected to users who are also well-connected [read: have numerous connections],” says Gonzalez-Bailon. That “gives you an advantage when it comes to reaching more people with your messages.”You are central if you are connected to other central users: to be at the core of the network (according to the definition of core we use) you need to (a) have many connections and (b) have connections that also have many connections. Centrality is less an individual than a ‘social’ attribute.”
At various points in the study, these central users are also referred to as “core” tweeters, “broadcasters” or “global bridges.”Why are they referred to as “global” bridges? Because they connect multiple “local” networks. Others are are mostly influential through their “local” network, which refers to the mutually-following connections tweeters have.
- Local networks are key to spreading behavior — i.e. people’s behavior is most strongly influenced by their mutual connections.The more a person’s mutual connections tweet about something, the greater the likelihood they’ll take action on it. Explains Gonzalez-Bailon, “reciprocal connections are more relevant to the recruitment process — not to the information diffusion process — because they are more likely to reflect offline relationships. Our argument is that spreading behavior — in our case, [recruiting people by getting them to take] the decision to start sending protest messages — is different from spreading information (i.e. reaching a large number of people).”
A case study by Brett Tabke on how WebmasterWorld sold more tickets to Pubcon focused on just this: they added a tweet button to their checkout thank you page so that webmasters could be influenced by seeing lots of friends tweeting about attending the show.
- Bursts of activity are most effective at influencing users. This makes sense since we know from the advertising world that it takes between 3-7 exposures to an ad for the message to be remembered. And obviously, the time-sensitive nature of tweets means that 20 tweets on a topic spread over 20 days are less likely to impact someone than 20 tweets in an hour. (This is what I was getting at in saying that your tweet reach depends on having a maximal share of your followers’ attention.)
Per the Oxford press release about the study, “the time at which different users first got involved and started emitting messages allowed the researchers to distinguish between activists who were leading the protests and those who responded later on. They found that when calls to action came from many different sources within a short time window, their effects were amplified, resulting in ‘recruitment bursts’. The vast majority of users were recruited this way responding to the collective behavior of others, says the study.”
- Protest initiators or “seed” tweeters were not necessarily central users. They were spread randomly through the network, varying from very central to hardly central positions. We sometimes think that memes spread virally because they’re first spread by people who command lots of attention. That’s not necessarily the case — sometimes the seeds are Oprah Winfreys and other times they’re your anonymous next door neighbor. “Mass mobilizations depend not on the influence of central users, who are nonetheless crucial for their growth,” said Gonzalez-Bailon, “but on the actions of many users in local networks that will ultimately reach the influential core.”
- The study finds empirical proof of early-adopters vs late-adopter behavior. Long known anecdotally, the study’s data shows that there exists a variance in how much exposure to a message it takes for people to adopt the desired behavior.
Conclusion on Tweet Reach / Impact:
As confirmed by Gonzalez-Bailon, we can conclude from the paper that the real reach of a Tweeter depends on the relationships with third parties of his own followers, as opposed to just his own immediate impact.