Enhancing value in time-sensitive service delivery systems using intelligent scheduling

• Main Author: Seakhoa-King, Shireen
• Format: Dissertation
• Language: English
• Published: Imperial College London 2020
• DOI: 10.25560/86003

In a typical service delivery system, a provider offers a service to a customer, based on an agreed commitment. Frequently, customer service requests are time-sensitive requiring prompt processing. Indeed, service providers are aware that the timely completion of processing of such requests can enhance the value which is accrued from the service encounter by both customer and service provider. However, the challenge facing many service providers is that time-sensitive requests are increasing superlinearly while resources available to carry them out remain limited, highly utilised and subject to much weaker growth rates. The problem is exacerbated when service durations are highly stochastic. In this context, the aim of this research is to enhance value to the service provider in time-sensitive service delivery systems through intelligent scheduling. The contributions are fourfold. Firstly, through an analysis of previous literature, key dimensions of a time-sensitive service delivery system are revealed and subsequently, a new taxonomy of service delivery systems is developed. Emerging from this, a quantitative concept which captures the varying value of service delivery as a function of time is proposed, namely the Time Value of Service function. Secondly, a Service Delivery Framework that supports modelling and decision-making for time-sensitive application domains is designed and implemented with the objective of enhancing value to the service provider. Thirdly, a new scheduling algorithm is proposed, the Least Lost Value (LLV) algorithm, that takes set of service requests, each comprising a Time Value of Service function and a duration distribution function, and outputs an ordered list of prioritised service requests for processing. Finally, the application of the framework and the LLV algorithm is demonstrated in the context of three case studies: (1) Big Data Processing, (2) Drone Delivery Service and (3) Emergency Services, where LLV outperforms traditional scheduling algorithms in enhancing value to the service provider. In the big data processing case study LLV extracted 300% more value than First Come First Served (FCFS), 200% more value than Earliest Deadline First (EDF), 160% more value than Shortest Remaining Time First (SRTF) and 110% more value than Highest Utility Density (HUD) algorithms. In the drone service delivery case study LLV extracted 48% and 31% more value than FCFS and Shortest Job First (SJF) in terms of realised revenue. Finally, in the emergency services case study LLV extracted 500% more value than the FCFS strategy.