This paper proposes a trace clustering approach to support process discovery of configurable, evolving process models. The clustering approach allows auditors to distinguish between different process variants within a timeframe, thereby visualizing the process evolution. The main insight to cluster entries is the “distance” between activities, i.e. the number of steps between an activity pair. By observing non-transient modifications on the distance, changes in the original process shape can be inferred and the entries clustered accordingly. The paper presents the corresponding algorithms and exemplifies its usage in a running example.