LLM 기반 VOC 데이터 자동 분류 및 Action Mapping 적용 사례 연구

Abstract

This study proposes a novel methodology for analyzing Voice of Customer (VOC) data using a Large Language Model(LLM), and evaluates its efficiency and practical applicability. VOC refers to the structured analysis of customer feedback to improve products and services, serving as a key resource for enhancing customer satisfaction and achieving competitive advantage in modern business. Using real-world VOC survey data from 2022 to 2024, our study focused on analyzing unstructured textual responses. To address the time consumption and complexity of traditional codebook-based manual classification, the study simplified the existing classification system and performed automated categorization using LLM. Furthermore, an Action Mapping Matrix was constructed for key VOC issues, enabling quantitative prioritization and strategic response planning. The LLM-based analysis processed a total of 2,604,282 tokens and was completed in approximately 105 minutes at a cost of 2,157 KRW, demonstrating significant time and cost savings compared to manual methods. The findings indicate that LLM-based VOC analysis can deliver practical efficiency for repetitive and structured tasks, providing a foundation for strategic AI adoption in both public and private sectors.