Date of Award
1-1-2025
Document Type
Dissertation
Degree Name
Ph.D. in Health and Kinesiology
First Advisor
Minsoo Kang
Second Advisor
Brennan Berg
Third Advisor
Jong Eun Lee
School
University of Mississippi
Relational Format
dissertation/thesis
Abstract
Social media marketing (SMM) has become an essential strategy for enhancing fan engagement, brand loyalty, and revenue generation in the sport field. The integration of social media platforms allows sport organizations to interact directly with consumers, promoting a shift from passive viewing to active participation. This dynamic landscape has made measuring the effectiveness of SMM increasingly important. However, current research often falls short by focusing solely on financial metrics (e.g., return on investment [ROI]) or non-financial metrics (e.g., return on objectives [ROO]), failing to capture a comprehensive picture of SMM’s overall impact. There is a critical need for an integrated framework that combines financial and non-financial metrics to assess SMM performance and reflect the strategic purpose of social media content.
The purpose of this dissertation is to develop and evaluate a hybrid framework for measuring the performance of SMM in the sport manufacturing industry. Specifically, this study aims to (1) evaluate the classification performance of fine-tuned large language models (LLMs) in categorizing social media content by objectives and (2) apply a slack-based data envelopment analysis (DEA) model to assess the efficiency of firms using both financial and non-financial performance metrics.
Study 1 evaluates fine-tuned LLMs in classifying 10,220 social media posts from 15 sport manufacturing brands across Facebook, Instagram, and Twitter. Posts are categorized into seven strategic content types, and model performance is compared across zero-shot, few-shot, and fine-tuned conditions. Study 2 applies a slack-based DEA model to assess the efficiency of SMM strategies for the same 15 brands. Inputs include sales cost, demand creation, and content frequency; outputs include revenue, net income, and the number of likes and comments. Efficiency is evaluated from both technical and scale efficiency perspectives.
In Study 1, the fine-tuned Gemini-2.0-lite model achieved the highest classification performance, with an accuracy of 98.5% and an F1-score of 0.978, outperforming both zero-shot and few-shot approaches. In Study 2, four brands (F1, F7, F11, F13) were fully efficient in the financial-only model, while firms such as F2 and F12 exhibited notable inefficiencies due to excessive marketing costs and low net income. In the non-financial model, most firms demonstrated high technical efficiency. However, scale inefficiencies were common among those overproducing certain content types. In the combined model, overall efficiency scores improved, but slack analysis revealed ongoing inefficiencies in firms such as F8 and F12, particularly in aligning content strategies with profitability.
The hybrid framework developed in this dissertation demonstrates that combining LLM-based content classification with slack-based efficiency analysis provides a more comprehensive understanding of SMM performance. The findings highlight the need for sport brands to align content strategies with financial efficiency and suggest that integrated evaluation approaches can guide more data-informed marketing decisions.
Recommended Citation
Lee, Seungbak, "Enhancing Social Media Marketing Measurement in Sport: A Hybrid Approach" (2025). Electronic Theses and Dissertations. 3317.
https://egrove.olemiss.edu/etd/3317