Examining visual impact: predicting popularity and assessing social media visual strategies for NGOs

Abstract Purpose This research aims to analyze the role of visuals posted on the social media of NGOs and to predict the popularity of a post based on the characteristics of the visual it contains. Design/methodology/approach Two social media platforms, namely Facebook and Instagram, were selected as the empirical study environments. Specifically, all visuals posted on 12 child-related Non-Government Organizations during the period of 2020–2021 (4,144 in total) were collected and subsequently subjected to automatic characterization using visual recognition and artificial intelligence tools. Machine learning algorithms were then employed to predict the popularity of a post solely based on the visuals it contains, as well as to identify the most significant features that serve as predictors for post popularity. Findings The Support Vector Classifier performed best with a prediction accuracy of 0.62 on Facebook and 0.81 on Instagram. For the explanation of the model, we used feature importance metrics and found that features like the presence of people and the emotions of joy and calmness are important for the prediction. Practical implications Companies and organizations serve a large part of their communication strategy through social media. Given that every advertiser would like to use their funds in the most efficient way, the ability to predict the performance of a post would be a very important tool. Social implications The methodology can be used in the non-profit sector, whereby knowing what visual will perform better they could promote their mission more effectively, increase public awareness, raise funds and reduce expenses on their communication strategy. Originality/value The novelty of this work regarding popularity prediction on social media lies in the fact that to make the prediction, it focused exclusively on the visual and its characteristics and achieved high accuracy scores in the case of Instagram. Additionally, it provided important information about visual characteristics and their importance in predicting popularity.