Predictive maintenance is the promise of the future in infrastructure asset management. Predictive maintenance more and more uses sensor data. Sensors are relatively cheap, and their data mostly comes in huge quantities. Trends or flags may be observed in the data, sometimes with traditional statistical analyses and more often with advanced analyses such as machine learning techniques. These trends and flags may indicate a developing problem, allowing maintenance professionals to act before a failure occurs. However, as of today predictive maintenance is far from being common practice in infrastructure asset management. Failure mechanisms are often extremely complex. Besides knowing how to search for flags and trends, one should first know what to search for. Also, on the organisational side barriers are found. Often, big data is available in infrastructure organisations but underutilised for various reasons such as inaccessibility of data, client unfriendly user interfaces and a lack of tools to analyse data effectively by maintenance engineers. The current research investigates the potential for more predictive maintenance in current professional practices, not by adding new sensors but through exploitation of existing data (data mining) and removal of barriers which professionals experience in using such data.