Chapter Preview
TopIntroduction
Service sector forms a growing portion of world economy; Archarya (2006) estimated that service economy accounts around 50% and 70% of the total value adds in the developing and developed countries, respectively. Wolfl (2006) reported that service industries, including social and personal services as well as business services, are performing poorly in terms of productivity growth. In overall sectorial comparison, the service sector still seems to lag behind manufacturing. There were various thoughts dedicated to improving service productivity. Sherman and Zhu (2006) proposed a data envelopment analysis (DEA) method to benchmark service performance. Li (2008a) compared between a manufacturing process and a service process, and suggested opportunities of improving service productivity across service cycle via software technology.
In the service business, a recurring issue is how to optimize demand and supply to achieve business benefits for a service enterprise. Traditionally, academic researchers and industrial practitioners resorted to manufacturing production planning techniques such as System Dyanmics (SD) and Discrete Event Simulation (DES) to help solve service production planning problems. Antoniol, Cimitile, Di Lucca and Di Penta (2004) used queuing simulation to access staffing needs for a software maintenance project. Liu and Wang (2007) applied constraint programming as the searching algorithm to optimize resource assignment problems of linear construction projects. Li and He (2008b) devised new local search heuristics to optimize lead time and resource utilization for utility service enterprise; previously the focus of resource allocation was more on cost saving than on customer experience.
Manufacturing production planning methods use statistic models to approximate demand and supply, and match demand with supply to identify and fix gaps. In a geographically-distributed service operation, resource planners tend to group demand and supply into regional buckets based on geographical and skill distributions of the business. They then apply production planning methods to optimizing each service operation bucket. During the matching, individuals in the demand and resource are often treated indifferently such that a uniform mapping between demand and supply can be arranged. Examples of matching are the required number of programmers for software work packages, civil engineers for construction activities and field technicians for telecom service tasks.
In industrial practice, however, traditional manufacturing production planning methods have constantly met fierce challenges from service planning and operation community. The reason is that statistical approaches often try to abstract real-world problems so that standard mathematical models can be used to represent the problems. In doing so, original problems are often over-simplified perhaps even distorted and a large amount of key information gets lost. Take field service for example, every field technician has unique skills, geography and productivity; the matching between a task and a technician is not a simple linear mapping but determined by scheduling rules, location and type of next task, current location and skill of technician, travel time from current location to the next, onsite time spent by the technician for the task, etc. Taking a simple average and variance from the bucket won’t reveal a technician’s individual circumstance, performance and opportunity for optimization.
Agent-Based Service Analytics (ABSA) is a new analytical framework, coined by the author, to address the deficiency of manufacturing production planning methods when reused in the service domain. It is composed of a suite of techniques for the end-to-end analytical tasks such as service modeling, visualization, analysis, simulation and optimization. The key drive of this framework is to understand the behavior and optimization opportunity of individual agents in a service process, be it customer, employee or supplier, which are often the ultimate questions asked but cannot be answered by the service planners and operators. This is in contrast with optimization in a manufacturing process where the interest is in process activities and the aggregated number of identical entities flowing through each activity.
Key Terms in this Chapter
System Dynamics (SD): Is an approach to understanding the behavior of complex systems over time. Originally created during 1950s by Professor Jay Forrester, it deals with internal feedback loops and time delays that affect the behavior of the entire system. In comparison with other simulation techniques, SD uses feedback loops, stocks and flows as its constructs, which can often be represented as a set of differential equations in math. Initially used to help corporate managers improve understanding of industrial processes, SD is currently being used in social and economics domain for policy analysis.
Service Optimization: Aims to improve a service process that involves close interaction between customer and provider that bear human factors and greater uncertainties in comparison to a manufacturing process. The key focus of service optimization is to make customers happy by satisfying their individual requirement as well as improving individual employee productivity whose collective impact on service performance will emerge at top-level. This is often in contrast with manufacturing process optimization where improvement is managed at activity level to produce more identical components as needed.
Discrete Event Simulation (DES): Emerged during 1960s in the manufacturing boom; it models the operation of a system as a discrete sequence of events in time, typical of a manufacturing process. Each event occurs at a particular instant in time and marks a change of state in the system. This contrasts with continuous simulation in which the simulation continuously tracks the system dynamics over time. Conventional DES constructs are entities, activities and queues; these constructs are linked to form a complex process in which entities flow. Probability functions are also used to model stochastic processes.
Agent-Based Simulation (ABS): Is a relatively new paradigm that simulates the simultaneous operations and interactions of multiple agents, in an attempt to re-create and predict the appearance of complex phenomena. In OR context, allowing agents to behave differently and communicate among themselves distinguishes ABS from other simulation techniques such as SD and DES. Existing ABS applications are primarily found in the science, social, economy and market domains; the rules used in these agents are either theory-based or empirical-based and the results of simulation are often lack of convincing power.
Agent-Based Service Analytics (ABSA): Is a newly coined term for capturing an end-to-end analytical framework to optimize a service process. The framework consists of tasks such as agent modeling, agent visualization, agent analysis, agent simulation and agent optimization. In comparison to DES framework where focus is often at activity level and from a central-control perspective, in ABSA the focus is to understand the behavior of individual agents in the service process and to identify opportunities of optimizing agents in order to improve the service process, as is often required by service business owners.