Understanding the complexity of urban dynamics requires the combination of information from multiple city data sources. Besides traditional urban data, geo-localized social media provide human-generated content, which may reflect in (near) real time the activities people undertake in cities. The challenge is to devise methods and tools that enable the integration and analysis of such heterogeneous sources of information. Motivated by this, we developed SocialGlass, a novel web-based application framework to explore, monitor, and visualize urban dynamics. By deploying our platform in three real-world use cases, the paper elaborates on the benefits and limitations of integrating social media with related city datasets. It further shows how the inherent spatiotemporal, demographic, and contextual diversities of social data influence the interpretations of (dynamic) urban phenomena.