Reputation Recovery After Digital Complaint Shocks Using Textual Diagnostics and Causal Machine Learning Evidence from Consumer Platforms

• Hatem Gharbi

  Higher Institute of Business Administration of Gafsa, Department of Management Information Systems, Route de Tozeur Km 4, Gafsa 2100, Tunisia

  Author

• Walid Mansouri

  Faculty of Economic Sciences and Management of Mahdia, Department of Finance and Banking, Avenue Tahar Sfar, Mahdia 5111, Tunisia

  Author

• Karim Bouazizi

  Higher Institute of Industrial Management of Sfax, Department of Business Analytics and Operations Management, Route de l’Aéroport Km 4, Sfax 3018, Tunisia

  Author

Service organizations face a recurring empirical question: why do some reputational injuries fade quickly while others persist across review cycles, search visibility, and local demand conditions? This study examines reputation recovery after digital complaint shocks using a weekly panel of 42,318 local service units belonging to 626 multi-location brands between 2016 and 2024. The dataset combines 19.6 million consumer reviews, 5.1 million managerial responses, platform visibility indicators, local competition measures, and outlet-level operating characteristics. Complaint shocks are identified from abrupt increases in negative diagnostic language, not merely from rating declines. The analysis applies transformer-based complaint classification, dynamic reputation indexing, synthetic-control matching, accelerated failure-time models, double machine learning, and causal forests. The results show that reputation recovery depends less on apology frequency than on operationally specific response content. Managerial responses that name the failure mode, describe a corrective action, and avoid generic assurance language shorten median recovery time by 27.8%. Boilerplate responses are associated with slower recovery and higher complaint-topic persistence. Causal-forest estimates show the largest recovery gains for outlets with moderate pre-shock reputation, high review credibility, and complaint topics involving process breakdowns rather than fairness or cleanliness failures. The findings indicate that digital reputation is repaired through diagnostic credibility and observable correction rather than through response volume alone.

• PDF

Copyright (c) 2026 International Journal of Science, Engineering and Applied Technologies

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.