Operating in dynamic and often unpredictable environments, business processes are exposed to changes that cause their execution to fluctuate over time. Since changes are inseparable parts of process execution, capturing them is crucial for gaining reliable insights into how they behave during their execution. Although prior research has studied continuous fluctuations of processes, most studies have focused primarily on visual representations of such changes. However, drawing more actionable insights requires a structured analysis approach that goes beyond visualization. To this end, this paper explores the potential of leveraging established time series analysis techniques---such as decomposition and correlation analysis---to derive meaningful insights into such changes. We evaluate our approach through backward-looking and forward-looking analysis: the former demonstrates the type of insights it can provide into continuous process dynamics by applying it to multiple event logs, while the latter examines whether the approach is beneficial for a downstream process mining task, the prediction of future process behavior.

Top