Determining staffing requirements of non-stationary queueing systems based on the efficient frontier

  • Tobias Rosinski

Student thesis: Phd


Existing methods for the modelling and staffing of non-stationary queuing systems enable controlling the trade-off between costs and speed. However, the speed of the service is only one aspect of operational performance that affects customer satisfaction. The quality of the service provided by the servers, which may be as or even more important than the waiting time for the customer, cannot be considered yet. This research proposes to determine the time-dependent required number of servers based on a non-parametric estimate of the efficient frontier. The proposed method enables resourcing according to both targeted speed and quality.The accuracy of the efficient frontier based staffing method is tested in the context of passenger and hand luggage screening at airport security checkpoints. Aviation security is a typical example of a service operation where not just the waiting time but also the quality of the service matters, since the purpose of conducting a security check is the detection of prohibited items. The assessed accuracy of the suggested method is compared to the stationary independent period by period approach that would be expected to be the most appropriate of the existing approximate methods for the test case. The resulting waiting times from both sets of recommendations of staffing profiles, based on the same demand forecasts, are estimated by means of simulation.The results suggest that the proposed method does not just enable the consideration of another aspect of operational performance, but is also 18 per cent more accurate than existing approximations in terms of the targeted waiting time in the tested environment. Moreover, the efficient frontier based staffing method is the first direct staffing method that does not require specifying the targeted speed of the service in terms of the impracticable probability of waiting. Instead, the waiting time is the target measure. So far, considering the waiting time was only possible by using methods for the modelling of non-stationary queueing systems, but they entail some trial and error when they are used to determine staffing requirements.
Date of Award1 Aug 2014
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorJian-Bo Yang (Supervisor)

Cite this