Data-driven solutions are becoming more available to infrastructure asset managers. Roads are among infrastructure sectors that have good data availability in most jurisdictions. The collected data are not always usable as it has different types of errors and mistakes. This is especially an issue in pavement performance data. The current study investigates common data quality issues and their root cause in the domain of pavement performance. For this purpose, 40,000 international roughness index (IRI) and 4000 pavement condition index (PCI) records were queried from the Long-Term Pavement Performance (LTPP) database to study the anomalies in performance data. Machine learning predictive models and statistical testing were used to identify the factors associated with error. It was observed that the following factors are among the key contributors to error and mistake in performance data: initial condition, age and climatic region. Furthermore, a random forest algorithm was developed for predicting the likelihood of error and mistake in data as well their magnitude. The model was able to predict errors with an accuracy of 86%. The findings of this study can be useful to municipalities, departments of transportation and consultants who are interested in the notion of data quality management in asset management.