Social network analysis of Twitter
Social network analysis (SNA) is a way of graphing the relationships that exist on social networks. It provides a way of visualizing and analyzing how people are connected, how they interact, and who the important influencers are in a network.
In The Tipping Point, Malcolm Gladwell wrote about three types of people: connectors, mavens, and salesmen. Connectors are the ones we are looking for in a network analysis – these people who have a large number of contacts and are central in a network. Knowing who these people are helps you to know who to reach out to if you want to expand your own influence and reach new people.
Let’s look at an example. In Vancouver, a number of Twitter accounts have been set up by the city government to promote civic engagement in various neighbourhoods. The Twitter account, @MarpolePlan, is one of those accounts. For those not familiar with Vancouver, Marpole is on the south side of Vancouver, just north of the airport. While over 20,000 people reside in Marpole, only 103 people are following the @MarpolePlan Twitter account.
In order to figure out how this Twitter account be used to improve reach and engagement, I’ve done a social network analysis. For this diagram, I’ve run a Twitter search for anyone who has mentioned “Marpole” in a tweet and have visualized how those people are interrelated. What I’m interested in are people who do not currently follow @MarpolePlan – we’re looking to expand their follower base to include new users, not to users they already reach.
Here’s the graph of users we get:
In this diagram:
- the black dots are Twitter accounts,
- the lines between the dots are either mentions or replies about Marpole between different Twitter users, and
- the small circles are tweets (representing a self referential loop as a person doesn’t mention another user).
Already you can see that at the centre of this network are two strong connectors who are key in bringing this network together.
To clean this network up, I’ve removed all the Tweets about Marpole and have just left Twitter accounts that either mention or reply to another user about the target keyword. A tweet is just a broadcast message; it’s the replies and mentions that demonstrate a stronger level of engagement within the network. To give this network added meaning, I’ve applied colors to highlight who in the network different groups cluster around. Here’s what the network looks like and the names of the key Twitter users:
As you can see from the network, there are certain users who act as important bridges, ensuring that information about Marpole reaches groups that otherwise wouldn’t be exposed to that information. For instance, if @kathyacouch were to be removed from this network, the other nodes in orange wouldn’t have access to the information she’s broadcasting about Marpole – there are no other paths that exist to these users. Also note the important role of the @theprovince Twitter account in this network. The Province is one of Vancouver’s local daily newspapers, and its influence online helps to inform what will be important for @MarpolePlan’s offline marketing strategy as well.
Let’s try a different network visualization to tackle this in another way. The first set of network graphs look for people who have an interest in Marpole but who do not have a direct or indirect relationship to the city planning Twitter account. But it’s also of value to understand your current social network and who follows @MarpolePlan either directly or indirectly on Twitter. By looking for the connectors in your current social network, you will know who the key people are who help to extend your reach.
In this next diagram, I’ve taken not just everyone who follows @MarpolePlan, but also everyone who follows those people – so we’re not just looking at followers, but followers of followers. This is what the network looks like:
It’s a mess! The black dots represent Twitter accounts and the grey lines represent a follower – followee relationship. But the network is so dense it’s impossible to make out the important details. This is to be expected with a complex network. As Ben Shneiderman would say – for data visualization you should first get a high level overview, then zoom and filter, and get the important details on demand.
One way of simplifying this network and looking for influential members is by trying to find people who have the most connections. This isn’t necessarily going to be the person with the most followers. In this network, I’ve only included Twitter accounts that have at least a second degree connection to @MarpolePlan. So it’s not who has the most followers on Twitter – it’s who has the most connections within this specific network of people.
I’ve changed the network diagram so that the size of the vertices reflects that number of followers that Twitter account has in this network – the bigger the dot, the more followers that account has. I’ve also filtered out accounts without a large number of connections so we can zero in on the very core of the network:
The red dot at the centre of this network is @MarpolePlan. From this diagram, you can see there are actually only 15 key Twitter accounts that are extremely well connected in this network. I’ve limited the number of Twitter accounts that are displayed in this view for simplicity’s sake, however, those large vertices are connected to a large number of other accounts that aren’t represented in this view – many of which do not have a current direct connection to @MarpolePlan. From a marketing perspective, these Twitter accounts are going to be key in spreading messages broadly throughout this network.
Looking at the bio of each @Marpole Plan follower provides an additional data visualization opportunity. Scraping the bios of @MarpolePlan’s followers and visualizing them in a word cloud with size representing the most popular words gives us clues into the common interests of the accounts followers. If we wanted to, for example, run a sponsored campaign on Twitter, this gives us some additional keyword ideas that could be used in conjunction with Twitter’s promoted tweets or promoted accounts advertising product.
Here’s some of the words that stand out as potentially usable keywords for a Twitter promoted tweets or promoted account campaign: city, development, entrepreneur, sustainable, public, business, company, public, planning, community, and urban. Several of the keyword like city and planning are obvious choices for a sponsored campaign, but a few are a bit less obvious, such as sustainability and entrepreneur. With a few geographic limiters set on the campaign to limit exposure to people in Vancouver, this could help the account reach a wider audience.
How can the information we’ve gathered out of this analysis best be used?
- follow key Twitter accounts, ideally so they follow you back
- mention or direct mention key Twitter accounts directly with important information to ensure they reach the full extent of your network
- utilize paid Twitter advertising such as promoted tweets or promoted topics on topics of interest to your users, using geographic targeting for additional relevancy
- check to see if connectors and information brokers in your social network occupy similarly important roles on other social media channels
- also check if key influencers also have well trafficked websites which could be beneficial for reaching a new audience or gaining backlinks for SEO purposes
Knowing the details about how the users in your network relate to each other can be very useful. In this case, we’re trying to understand who the important users are in order to further spread information about community planning in Marpole. But this information can also be used for crisis mitigation in the case of inaccurate information about an organization or product. Essentially, network analysis gives you better insights into who is talking about your organization and how information is passed along – which is what good marketing is all about.