Engaging the right audiences
As government agencies expand their engagement with citizens through social media platforms such as Twitter, there is an increasing need to better understand their audiences.
Understanding when and why customers interact with these agencies via social media is important. Just as critical is evaluating the effectiveness of the department's messaging. This knowledge equips departments with the tools to improve their business processes and future communication materials. It also reduces the need for customers to contact the department through other mediums, like hotlines, with basic enquiries: saving time and money.
Working with the Department of Human Services, we've developed a platform that allows their communications team to identify their social media followers on Twitter and analyse the audience reach of tweets at a glance.
We designed and developed TweetRipple based on extensive time spent with the department’s social media managers and observing their everyday workflows.
TweetRipple, a web-based software system, allows social media managers to analyse various demographics of their Twitter audience, including any changes over time. The tool can track not just followers, but people who generally interact with the account, for example, by tweeting at the Department's handle. It also allows social media managers to compare the audiences of their managed accounts as well as those of similar organisations.
TweetRipple uses our existing framework of natural language processing and text analysis algorithms, which works by grouping tweets by topic. Presenting the data in this way provides social media managers with clear and immediate insight into the major issues affecting their customers, which could vary on any given day.
Using data to refine tweets and reach the right audience
TweetRipple makes use of machine learning algorithms to group an audience in various ways. Firstly, we segment of the audience into organisations or individuals. For individuals, we further group into social roles - whether a mother, student, journalist, or scientist - and predict that person's gender. We finally geolocate that person, and narrow our focus to Australian-based accounts.
Partitioning the audience into specific groups makes it possible to create more targeted messaging. For example, geographic region may be important after a natural disaster event. Or, there might have been a change to student payments, meaning it's important to spread the word to both students as well as parents.
The platform also analyses all tweets related to the department's accounts looking at both their emotion and sentiment. Looking at tweets grouped by emotion revealed that many of the department's own tweets were labelled as "sad" since they often began with the word "sorry". This has led to the department making a concerted effort to phase their responses in a more positive way.
Audience reach on Twitter is not just about followers. The department's social media team wants to understand the broader 'potential audience' for its Twitter messages, by analysing who follows its followers and who interacts with the department's accounts by retweeting, replying to or mentioning them. This allows the team to identify key influencers and audience trends within their broader Twitter following.
The department's social media team also needs to easily identify and understand followers of its Twitter accounts. This ensures that important health and social services messages reach the right audience and are tailored to suit their needs and preferences. Traditionally, staff in the department's social media team would have to scroll through the 'follower' list for each of the department's official Twitter accounts to identify who follows each account. Tracking who follows multiple departmental accounts was a more onerous task, again requiring manual staff action to scroll through follower lists. Tracking these changes in real time allows the team to respond to emerging issues on Twitter in a more informed way.
We're continuing to work with the department to collaborate on incremental updates to the TweetRipple tool.