Arrow DownArrow ForwardChevron DownDownload facebookGroup 2 Copy 4Created with Sketch. linkedinCombined-Shape mailGroup 4Created with Sketch. ShapeCreated with Sketch. twitteryoutube

Case study

Optimising social media campaigns to drive audience engagement

We used our technology to measure and monitor the real-time effectiveness of its social media campaigns. Our analysis resulted with a list of 15 data-driven recommendations to implement, including topics that drive most engagement and the time of day to achieve the greatest impact.


Customer

A large organisation running social media campaigns in the UK.


Problem

The organisation spends several million pounds each year running social media campaigns to promote its message on over 100 owned and affiliated pages. However, the organisation was not able to monitor the impact of these campaigns effectively to see how its messages were being received. Faculty was asked to create a campaign research database to help assess the performance of activity over time and find ways to help the organisation target its campaigns more effectively.


Solution

We built a series of data pipelines using Faculty Platform to capture both historical and ongoing social media data feeds. All unique identifiers were anonymised before being stored. APIs were built to interact with all major social media feeds, including Facebook, Twitter, YouTube, Instagram and Google. We identified that they had created an echo chamber, with interactions evident between owned campaigns (different campaign pages owned by the same company) but almost none with target audiences.

Engagement echo chamber plot
(owned vs. target groups)

 

Topic cluster analysis using HDBSCAN

We then tested the statistical impact of the messaging based on a number of features, including the time of day the posting was made and whether the post was framed as a question. We also used Natural Language Processing (NLP) to analyse how the sentiment (whether positive, negative or neutral) and reading age of posts affected the engagement of different groups.

Finally, we conducted topic analysis on over one million posts and comments using NLP and an unsupervised learning model, HDBSCAN, to cluster posts. This highlighted the key topics driving engagement in different groups across owned and competitors’ campaigns.


Impact

Our data analysis allowed the organisation to measure and monitor the real-time effectiveness of its campaigns. This immediately highlighted the need for a full strategy revamp. Our analysis of effective campaign strategies concluded with a list of 15 data-driven recommendations to implement, including topics that drive most engagement and the time of day to achieve the greatest impact.

To find out more about what Faculty can do
for you and your organisation, get in touch.