The data collected from the Wheel Impact Load Detector (WILD) system must be reliable to ensure operational efficiency and improve maintenance strategies. Wheelsets, one of the important rolling stock components, has a very important role of maintaining safety standard, maintain stability and preventing damage to the infrastructure. However, the variations present in the data collected from WILD system, influenced by various environmental or operational factors, can make it challenging at times to rely on these data. These data inconsistencies create difficulties to effectively identify and assess the degradation patterns. This study focuses on analysing the quality and consistency of WILD data by implementing statistical and visualization techniques, and to find out if it is suitable for predictive maintenance strategies. By analysing the distributions, trends, and anomalies in wheel impact forces across different conditions, the study observes patterns such as bimodality, which may represent the presence of operational or measurement inconsistencies. The results indicate that combining visual analysis and statistical tests may enhance the interpretability and reliability of WILD system data. This study provides a foundation for refining data and improve its quality, with the aim of building reliable data driven models for maintenance of wheelsets.