Background and Related Work

Batch processing is when a resource accumulates cases to process them together as a group. Batch processing is often used to solve scheduling problems, reduce processing times and costs by executing multiple cases together. However, introducing batch processing requires accumulating cases, which increases waiting time. As such, batch processing implies a trade-off between processing and waiting time. As batch processing impacts processing and waiting times, we can use the CTE metric to measure batch processing efficiency. CTE calculates the ratio of the processing time relative to the cycle time of batch processing activities. CTE ratio close to 1 indicates that the process has comparatively low waiting to processing time and, thus, little room for improvement. However, low CTE indicates a comparatively high waiting time to processing time. Therefore, there is an improvement opportunity. CTE is used to identify time-related process inefficiencies. For instance, it has been applied for measuring time-related performance in factories to detect inefficiencies. Ignizio reports that CTE is the most efficient metric in identifying workstation instability in production processes. Furthermore, CTE has been used to assess the efficacy of process redesigns. We, therefore, use the CTE metric to analyze batch processing (in)efficiencies.

Process mining techniques enable the discovery of batch processing behavior from event logs. For instance, Nakatumba et al. address the problem of accurately reproducing batch processing behavior in simulation models. Wen et al. propose a process mining technique to discover batch processing from events logs. Pufahl et al. use event logs to enhance process models with batch processing. Andrews et al.  propose an approach to identify and quantify shelf time, i.e., idle time that exceeds acceptable duration, in business processes. Pika et al. propose an approach for discovering batch processing from event logs and discusses, in particular, how to identify batch processing from multiple perspectives of a business process such as activity, resource, and data perspectives. Martin et al. focus on identifying rules that trigger a batch processing activity. Similarly, Martin et al. use batch processing metrics, such as frequency of batch processing, batch size, duration of activity instances, and waiting time in a batch, to describe batching behavior. This paper focuses on discovering and analyzing the waiting times associated with batch processing. Thereby, we build on existing research to identify batch processing inefficiencies.

Process mining techniques have also been applied to assess batch processing performance and explore their impact on process performance. Thus, in addition to batch discovery, Batch processing behavior is visualized and quantitatively analyzed to identify specific patterns such as detecting outliers. Similarly, Klijn propose quantifying batch processing performance using measures such as intra-batch case inter-arrival time, case inter-arrival time, batch interval, batch size, and batch frequency. Pufal et al. examine how overall process performance can be improved in terms of time and cost by simulating batch processing. While these studies use event logs to analyze various performance dimensions of batch processing, they do not consider the different types of waiting times associated with batch processing. Furthermore, their focal point is on batch-processing performance measures. We extend existing work by identifying the different types of waiting times related to batch processing and their impact on process performance to identify potential improvement opportunities.