Dynamic Work Distribution in Workflow Management Systems: How to Balance Quality and Performance

Today's workflow management systems offer work items to workers using rather primitive mechanisms. While most workflow systems support a role-based distribution of work, they have problems dealing with unavailability of workers as a result of vacation or illness, overloading context dependent suitability, deadlines, and delegation. As a result, the work is offered to too few, too many, or even the wrong set of workers. Current practice is to offer a work item to one person, thus causing problems when one person is not present or too busy, or to offer it to a set of people sharing a given role, thus not incorporating the qualifications and preferences of people. Literature on work distribution is typically driven by considerations related to authorizations and permissions. However, workflow processes are operational processes where there is a highly dynamic trade-off between security and performance (e.g., average flow time, service level, etc.). For example, an approaching deadline and an overloaded specialist may be the trigger to offer work items to lesser-qualified workers. This paper addresses this problem by proposing a systematic approach to dynamically create a balance tween quality and performance issues in workflow systems. We illustrate and evaluate the proposed approach with a realistic example and also compare how a workflow system would implement this scenario to highlight the shortcomings of current, state-of-the-art workflow systems. Finally, a detailed simulation model is used to validate our approach.