Deciphering Sport Marketing: An Ontology-Based Data Mining Core for Adidas Taiwan

Executive Summary

TL;DR: This research bridges the gap between Artificial Intelligence and Sport Management by introducing an Ontology-based Data Mining approach. By structuring knowledge through domain-specific ontologies and applying the Apriori algorithm and K-means clustering, the study provides Taiwan Adidas with a blueprint for matching specific consumer segments with the ideal media channels and celebrity endorsers.

Background: Within the broader field of CRM and market analytics, this work represents a transition from "blind marketing" to "data-driven intelligence," positioning ontology not just as a philosophical concept, but as a practical schema for relational databases.

Problem & Motivation: The "Semantic Gap" in Marketing

Prior sport marketing research often treated consumer data as isolated variables (age, income, gender). However, the "Why" behind a purchase often lies in the complex web of:

• Emotional Connection: Which athlete does the user admire?
• Media Consumption: Where does the user get their news (TV, Internet, or Magazines)?
• Product Perception: Is the product seen as "Performance" gear or "Heritage" fashion?

The authors realized that without a structured Ontology, data mining algorithms might find correlations but miss the contextual logic of the sports industry.

Methodology: Structuring Intelligence

The core of this paper is its three-tier methodology:

1. The Sport Marketing Ontology

Using Protégé, the researchers built a conceptual map of the Taiwan Adidas ecosystem. This ensured that every data point—be it a "Running Shoe" or "Beckham's endorsement"—was categorized with semantic accuracy.

Figure: The ontological association between consumers, products, and endorsers sets the stage for data extraction.

2. Association & Clustering

The researchers utilized a two-step mining process in SPSS Clementine:

• K-means Clustering: Grouped the 800 respondents into distinct clusters based on income and age.
• Apriori Algorithm: Within those clusters, it identified "Rules." For example: If a customer is interested in NBA, they have an 80% confidence of following specific Taiwanese sports channels.

Figure: The integrated research framework from data collection to knowledge discovery.

Experimental Results: The Adidas Case Study

The study identified two primary "Personas" that represent the bulk of the Taiwan market:

Cluster 1: The Traditional Professional

• Media: Prefers United Daily and Apple Daily (Newspapers) and TV channels like ESPN.
• Endorsers: Highly influenced by professional athletes like Chien-Ming Wang (Baseball).
• Strategy: Focus on performance-driven narratives in traditional media.

Cluster 2: The Lifestyle Enthusiast

• Media: Devours BANG and GQ Magazines; high internet engagement (Yahoo Sports).
• Endorsers: Follows local stars like Tien Lei and global icons like Kobe Bryant.
• Strategy: Market "Heritage" and "Style" lines through digital and fashion-forward outlets.

Figure: The resulting Knowledge Map serves as a strategic compass for Adidas marketers.

Critical Analysis & Conclusion

The Takeaway

The genius of this paper lies in its Semantic Schema. By forcing the data into a pre-defined ontology, the authors ensure that the resulting Association Rules are not just "statistically significant" but "commercially relevant." For Adidas, this means the difference between a wasted multi-million dollar endorsement and a pinpoint-accurate campaign.

Limitations & Future Work

While robust, the study relies on static questionnaire data. In the era of Web 3.0, the next frontier for this research is Real-Time Ontology Updates—streaming social media data (Twitter/Instagram) directly into the mining engine to capture 24-hour shifts in fan sentiment. Additionally, the shift toward "Micro-Influencers" over global celebrities like Beckham might require a more granular update to the "Endorser" branch of their ontology.

Conclusion

Liao et al. prove that data mining is not a "magic box." Its power is unlocked only when guided by a structured understanding (Ontology) of the human behavior it seeks to predict.