Thousands of music tracks are uploaded to the Internet every day through websites and social networks that focus on music. While some content has been popular for decades, some tracks that have just been released have been ignored. What makes a music track popular? Can the duration of a music track’s popularity be explained and predicted? By analysing data on the performance of a music track on the ranking charts, coupled with the creation of machine-generated music semantics constructs and a variety of other track, artist and market descriptors, this research tests a model to assess how track popularity and duration on the charts are determined. The dataset has 78,000+ track ranking observations from a streaming music service. The importance of music semantics constructs (genre, mood, instrumental, theme) for a track, and other non-musical factors, such as artist reputation and social information, are assessed. These may influence the staying power of music tracks in online social networks. The results show it is possible to explain chart popularity duration and the weekly ranking of music tracks. This research emphasizes the power of data analytics for knowledge discovery and explanation that can be achieved with a combination of machine-based and econometrics-based approaches.