Using Social Listening Data to Build a Structural Equation Model: An Exploratory Study

Описание: This is an exploratory study based on a pre-print available for view at ResearchGate.

Abstract

This assessment utilizes a new research method. The author, Eric Coyle, has professional experience with social listening for market research in addition to some academic knowledge building SEMs. The author seeks to combine their "real-world" knowledge of social media in combination with academic prowess using SEMs to further social listening capabilities for the greater good. The purpose of this study is to show how to potentially perform SEM (Structural Equation Modeling) using [sample] social listening data on the social media platform X, formerly known as Twitter. This study uses a methodology previously proposed; however, not necessarily tested for functionality to look at causal relationships using social media analysis. Examining causal relationships through SEM can catapult social listening analysis into an echelon of greater functionality in the age of AI (Artificial Intelligence), machine learning and natural language processing. Historically, social media data has been aggregated for review as a leading indicator and to construct simple averages and regression albeit from specific venues (e.g., Facebook, Instagram, blogs, etc…). Social media data can be collected but lacks causality analysis capability that other market research methods (e.g. survey analysis) offer, specifically through SEM. SEMs are typically built out using survey data. This assessment suggests streamlining social listening text-based data processes (and other similar input like voice transcripts converted into text) for eventual adaptation through SEM. This overture is original by way of being the first known scheme a) actually collecting, aggregating and analyzing social listening data, albeit with an AI generated data set based on messages about the brand Disney potentially found on the social media X, to build a test causal model converting social media data for use to build an actual SEM through b) illustrating a way to expand social listening beyond solely text-based analysis or venue-specific constraints while c) exhibiting how to build a SEM using social listening data within the business and marketing realm. The knowledge gap fulfilled advocates using SEM techniques in a way to expand future real-world social media research prospects. Furthermore, this analysis builds upon existing literature around challenges building SEMs with data sets with high percentages of missing high percentages of values while assessing data imputation methods to address missing values for SEM for future study. Past similar research does not provide adequate ways to replicate studies. As a result, a video series detailing methodology is available for public view on YouTube. Both a QR code and link to the video series are featured in Appendix I. This study shows that it is feasible to build a SEM using social listening data. Better indicator assessment, enhanced data imputation methods and improved data collection practices will continue to lend toward more desirable results.