This study presents a research approach using data mining for predicting the performance metrics of posts published in brands’ Facebook pages. Twelve posts’ performance metrics extracted from a cosmetic company’s page including 790 publications were modeled, with the two best results achieving a mean absolute percentage error of around 27%. One of them, the “Lifetime Post Consumers” model, was assessed using sensitivity analysis to understand how each of the seven input features influenced it (category, page total likes, type, month, hour, weekday, paid). The type of content was considered the most relevant feature for the model, with a relevance of 36%. A status post captures around twice the attention of the remaining three types (link, photo, video). We have drawn a decision process flow from the “Lifetime Post Consumers” model, which by complementing the sensitivity analysis information may be used to support the manager’s decisions on whether to publish a post.
Keywords: Social networks, Social media, Data mining, Knowledge extraction, Sensitivity analysis, Brand building
The worldwide dissemination of social media was triggered by the exponential growth of Internet users, leading to a completely new environment for customers to exchange ideas and feedback about products and services (Kaplan and Haenlein, 2010). According to Statista Dossier (2014), the number of social network users will increase from 0.97 billion to 2.44 billion users in 2018, predicting an increase of around300% in 8 years. Considering its rapid development, social media may become the most important media channel for brands to reach their clients shortly (Mangold & Faulds, 2009; Korschun and Du,2013).
Companies soon realized the potential of using Internet-based social networks to influence customers, incorporating social media marketing communication in their strategies for leveraging their businesses.
Measuring the impact of advertisement is an important issue to be included in a global social media strategy (Lariscy et al., 2009). Several studies focused on finding the relationships between online publications on social networks and the impact of such publications measured by users’ interactions (e.g., Cvijikj et al., 2011). However, fewer studies devoted attention to research for implementing predictive systems that can effectively be used to predict the evolution of a post before its publication. A system able to predict the impact of individual published posts can provide a valuable advantage when deciding to communicate through social media, tailoring the promotion of products and services. Advertising managers could make judged decisions on the receptiveness of the posts published, thus aligning strategies toward optimizing the impact of posts, benefiting from the predictions made.
Also, it has been shown that social media publications are highly related to brand building (Edosomwan et al., 2011). Therefore, the predictive tool outlined in this paper could leverage managerial decisions to improve brand recognition.
This research focused on modeling performance metrics extracted from posts published on a company’s Facebook page through the usage of data mining (predicting social media performance metrics. Moreover, the support vector machine technique was employed by feeding it with seven input features, all provided by Facebook’s page, except a content specific categorization provided by the page’s manager. Twelve performance metrics were modeled with these input features, from which the two models achieving the best performance modeled the “Lifetime Post Consumers” and the “Lifetime People who have liked a Page and engaged with a post” output features, with a mean absolute percentage error of 27.2% and of 26.9%, respectively.
Based on the “Lifetime Post Consumers” model, this study showed how it could benefit through its predictions of brand building by providing insights on social media engagement. The advantages of using the model were also linked to all the stages of branding (cognitive, affective, and cognitive stages). A data-based sensitivity analysis was then applied for extracting valuable knowledge from the model of “Lifetime Post Consumers.” The “Type” of the content published was considered by far the most relevant input feature for the model. Posts from the “Status type” are likely to result in twice the impact of the remaining“Types.” Also, seasonality was found regarding the “Month” the post was published. Publications related to special offers and contests are likely to produce posts with greater impact than “Product” and other non-explicit brand-related content. We also produced a decision flow process based on rules extraction from the model. Facebook page managers can use this knowledge to make informed decisions on the posts they publish, enhancing their impact, thus contributing to brand building.
Several ideas arise from this study for future research. First, the model may be enriched with other context features (e.g., if the product I being advertised elsewhere) for tuning its performance. Also, text mining methods could be employed to the content for extracting additional knowledge. Finally, text mining the comments of each post for user sentiment analysis could reveal the feelings each post is generating.
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:Predicting social media performance metrics and evaluation of the impact on brand building: A data mining approach
Received 29 September 2015
Received in revised form 11 February 2016
Accepted 15 February 2016
Available online xxxx
Content lists available at ScienceDirect
Journal of Business Research
Sérgio Moro a,b,Paulo Rita a, Bernardo Vala c,
a Business Research Unit, ISCTE–University Institute of Lisbon, Portugal b ALGORITMI Research Centre, University of Minho, Portugal
c ISCTE Business School, ISCTE–University Institute of Lisbon, Portugal
Please cite this article as:
Moro, S., et al., Predicting social media performance metrics and evaluation of the impact on brand building: A data mining approach (predicting social media performance metrics), Journal of Business Research (2016), http://dx.doi.org/10.1016/j.jbusres.2016.02.010
We would like to thank the two anonymous reviewers for their valuable recommendations, which highly enhanced the value of the final manuscript.
Considering the novelty of the proposed approach, the knowledge hidden in the “Lifetime People who have liked a Page and engaged with a post” was also evaluated, which achieved a mean absolute percentage error of 26.9%. Nevertheless, such a performance metric was influenced by considering only users who have liked the page, as argued in Section 4.1. Therefore, this Appendix shows in Fig. 15 the relative relevance of each input feature to the model (similar to the exercise displayed in Fig. 6). The results are aligned with those for the model of “Lifetime Post Consumers,” even though the “Type” is now more relevant (40%) than for the latter model (36%). Also, “Page total likes” are more relevant, while “Month” is less. Further studies would be required for a deeper analysis of the differences. Moreover, these studies would require additional data for differentiating the engagements of users who have liked from those that didn’t but also engaged.
PDF reference and original file: Click here
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.