Causal Process Mining for Temporal Deviations in Business Event Flows

Efficient management of business processes is crucial in dynamic environments. Temporal deviations in execution times can affect workflows. They also increase costs and reduce service quality. Traditional process analysis methods generally rely on correlations. They do not establish causal relationships that explain these deviations. Existing approaches face challenges in identifying the underlying causes of temporal deviations. This difficulty limits the ability to apply precise interventions. Without a causal understanding, corrective actions may prove ineffective. They often address symptoms rather than underlying causes. This study integrates process mining with causal inference. The objective is to explain the factors that generate temporal deviations in enterprise workflows at the variant level. Causal discovery techniques are applied LiNGAM, PC, and GES. Event logs are grouped by variant and stratified into temporal-deviation clusters. This procedure identifies critical activities. Directed acyclic graphs DAG are also recovered to quantify their impact on delays. In addition, alternative scenarios are simulated using the do(⋅) operator. These interventions provide actionable information for optimization. The methodology was validated with a dataset of 562 planning instances from a payment-processing company. The results identified causal relationships consistent with the learned DAGs. Among them, the impact of scheduled hours and complexity level on execution times. This work integrates process mining and causal inference. The result is a framework for data-driven decision-making in operational process management. The proposal improves the interpretability of temporal deviations. It also enables the design of more precise interventions. It contributes to establishing a methodological foundation for continuous improvement through causal analysis.