Quantifying risk of service failure in customer complaints: A textual analysis-based approach

This study introduces a novel data-driven Failure Mode and Effects Analysis (FMEA) approach for service risk management, leveraging textual analysis of customer complaints to identify and assess service failure risks. Unlike traditional FMEA methods that rely on expert opinions, our approach utilizes guided Latent Dirichlet Allocation (LDA) with prior knowledge to mine service failures, reflecting actual customer experiences. Additionally, a unique text analysis framework, utilizing sentiment polarity, mention frequency, and information entropy, objectively quantifies Severity (S), Occurrence (O), and Detection (D) values of service failure modes. Enhancing the model further, we introduce the Criteria Importance Through Inter Criteria Correlation (CRITIC) and Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) framework. This combination, replacing the traditional Risk Priority Number (RPN) method, utilizes objective weights and evaluates risks against ideal and nadir scenarios, offering a more accurate and comprehensive risk assessment. Applied to the hotel service industry in New York City, our method demonstrates enhanced accuracy and reliability over other methods. This research contributes significantly to service quality improvement, providing a more objective, efficient, and customer-oriented risk assessment approach.