Process mining techniques can be used to discover process models from event data and project performance and conformance related diagnostics on such models. For example, it is possible to automatically discover Petri nets showing the bottlenecks in production, administration, transport, and financial processes. Also basic statistics (frequencies, average delays, standard deviations, etc.) can be projected on the places and transitions of such nets to reveal performance and compliance problems. However, real-life phenomena such as overtaking, batching, queueing, concept drift, and partial blocking of multiple cases remain invisible when considering basic statistics. This paper presents an approach combining Petri-net-based discovery techniques and so-called performance spectra based on token flows. Token production and consumption are visualized such that the true dynamics of the process are revealed. Our ProM implementation supports a range of visual-analytics features allowing the user to interact with the underlying event data and Petri net. Event data related to the handling of orders are used to demonstrate the functionality of our tool.