Social media customer engagement plays a crucial role in the growth and success of B2B businesses. From a two-way communication perspective, knowing if a social media post will engage customers is of great value. The main purpose of this research is to evaluate the use of linguistic features and machine learning to predict the success of social media posts using user engagement metrics, such as likes, comments, and shares.
To answer this question, we analyzed 51,615 Facebook posts from 121 B2B companies to develop machine learning-based models that can predict the success of a post. We used user reactions (likes, comments, shares) as a success indicator and annotated the data based on that. Then we used the annotated data to train automatic models to predict the success of Facebook posts in user engagement, using the most frequent words in our data as feature words and three well-known machine learning algorithms: Support Vector Machine, Naïve Bayes, and Multi-layer Perceptron.
The performance of our models on the testing dataset showed that our models could predict the success of Facebook posts in user engagement with moderate accuracy. In addition, the statistical analysis of the annotated data demonstrated that the use of certain specific terms can affect user engagement. Several discoveries from our analyses can thus help social media managers create more engaging content and thus optimize their communication strategies:
- Use inspirational and positive words : Words like “inspiring,” “happy,” “moment,” or “welcome” can significantly increase user engagement.
- Incorporate time references : Words related to specific dates or time periods in posts, such as “week,” “year,” “summer,” or the different months, can capture users’ attention and interest.
- Experiment with different types of content : Our research focused on content posted on Facebook. However, we recommend testing and adapting these strategies across different social media platforms to gain additional insights and help refine approaches for different audiences. Indeed, our analyses showed that words related to products, business supply and improvement could be of interest to the audience of B2B companies on Facebook, but we can assume that they would be even more successful on a professional network such as LinkedIn.
- Adapt content based on analytics : While this research offers some room for improvement, we recommend using analytics tools to identify the language characteristics that work best for your specific audience and adjust your posts accordingly.
By implementing data-driven strategies, B2B companies can improve the effectiveness of their social media communication efforts by establishing engaging two-way communication that will foster stronger relationships with their customers and partners.
If you had to remember only one thing from this article: Use machine learning tools to adapt your speech to optimize engagement on your social networks and strengthen your online presence.
Source :
Saravani, S.H.H., Boeck, H., Bourguignon, B. (2024). Using Linguistic Features to Predict Social Media Engagement: Proposing an Approach Based on Machine Learning and Natural Language Processing. In: Reis, J.L., Zelený, J., Gavurová, B., Santos, J.P.M.d. (eds) Marketing and Smart Technologies. ICMarkTech 2023. Smart Innovation, Systems and Technologies, vol 386. Springer, Singapore. https://doi.org/10.1007/978-981-97-1552-7_27