Development of warranty management visualisation for automotive aftermarket with integration of AI

Warranty management in the automotive aftermarket is increasingly challenged by large volumes of fragmented and heterogeneous data originating from IoT devices, repair logs, and service records. Traditional systems lack the scalability and analytical depth to extract meaningful insights, resulting in delayed claim resolution and higher operational costs. This study proposes a data driven approach to modernize warranty processes by integrating artificial intelligence with interactive visualization. The research utilizes 11,000 historical warranty claim records collected from OEM customers between 2019 and 2023, comprising data attributes such as part numbers, failure codes, service dates, and repair locations. A four-phase methodology based on the Product Design Specification framework was employed: Information Collection, Concept Generation, Product Configuration, and Parametric Analysis. The system architecture follows the Model View Controller design, with SQL and Python forming the backend for data processing and modelling, while Power BI serves as the visualization platform. Advanced analytics techniques including Weibull distribution modelling for failure prediction and Python based anomaly detection algorithms were implemented to identify high risk components and unusual claim behaviours. Integrated dashboards allowed for real time monitoring of key performance indicators such as Warranty Claim Rate, Average Claim Cost, and Claim Resolution Time. The system achieved a warranty cost reduction of RM 23.5K, reflecting a 75% improvement over the five-year period. This study contributes a novel, scalable solution that bridges traditional warranty analysis with AI enhanced predictive analytics. The platform provides manufacturers with improved visibility, accuracy, and strategic foresight. Limitations such as noisy data and model generalizability are acknowledged, with future work aimed at enhancing robustness through natural language processing and adaptive learning models.