How to measure social media influence: the UCL challenge
By Lara Carim, on 19 January 2017
The social media landscape offers an overwhelming variety of reach and engagement metrics, but what do they actually tell us about the influence our social media activity has?
The UCL Digital Communications team regularly evaluates the social media channels we run in terms of growth and engagement. We also take a holistic view of performance for cross-network campaigns.
Having read a little about networking mapping and visualising relative influence, I became keen to explore how we could measure and map the influence of individuals and organisations who engage with UCL in the social sphere. I also thought this could help UCL to identify people or institutions with whom engagement would be mutually beneficial.
The underlying aim was to expand our knowledge of who is most interested in UCL and the areas we specialise in, to help us optimise the effectiveness of modest resources. We also wanted to create a bespoke, flexible framework we could use to evaluate strategic social campaigns with different goals, which we could share with colleagues across the institution.
We prioritised research into three types of social media accounts areas that we felt would help us measure UCL influence in a practical way:
1) accounts that engage most frequently with us on our own channels, and their areas of interest
2) accounts that engage most frequently with UCL in the wider social sphere, and their interests
3) the most influential accounts within a particular field.
We brought in the higher education digital strategy company PickleJar to help us tackle this challenge.
How do we define influence?
First, the Communications team took part in a workshop facilitated by PickleJar to investigate the values we set against different types of engagement on different social networks. For example, how does the value of a retweet compare with a reply or a like? Is someone with a large following more influential than someone whose tweets are mostly liked by others?
This was new territory for us with no simple answers, and we were gratified to hear that this was the same for the specialists:
Pickle Jar Communications has been working on communications, research, and strategy for the higher education sector in the UK and internationally since 2007. This project is the first time we have been asked to develop a way for a university to map influence in a strategic way.
After much productive debate, we settled on values that informed an ‘interaction score’ we would apply to different accounts. This was combined with follower figures to arrive at an ‘influence score’ for accounts interacting with UCL in some way.
The values and resulting formula are of course hugely subjective, and might differ from one campaign to the next depending on its goals. They are also subject to the evolution of the networks themselves. But the only constant in social media is change, and for the time being the formula has provided us with an agreed, customised and flexible framework.
The company analysed UCL mentions on our own accounts (focusing on Twitter and Facebook for budgetary reasons) and a range of external social networks over a two-week period. The timeframe was selected due to the characteristics of the social media monitoring tools at our disposal and of the social networks themselves.
Using a combination of free and paid-for software (Falcon Social, SocialRank, Twitonomy and FanPageKarma), the company identified the most prolific engagers on the various social platforms, and calculated their interaction and exposure scores using the values described above.
The handles were also cross-checked across a wider range of social platforms and associated websites were identified. This gave us insight into the scope of the digital presence of these individuals/organisations and other opportunities for engaging with them online.
The top keywords appearing in Twitter users’ interactions were also identified, to give a picture of their areas of interest.
A similar process was employed to identify influencers operating in a specific field. The subject ‘dementia’ was selected due to UCL’s research strength in this field, and the period was limited to three days given the volume of activity related to this topic. The most common keywords used by influencers in this field were also identified, enabling us to learn about different areas of discussion.
Finally, visualisations of the dementia network on Twitter of were created using two additional types of software. NodeXL Pro took the data from Twitter to create lists of edges (interactions) and vertices (Twitter accounts), which it then visualised as relationship and influence maps. Gephi was used to make the resulting maps more attractive.