بررسی خطرات ناشی از شکست سرویس بر اساس اثر موج دار شدن: روش شبکه های بیزی

This study responds to the need for assessing risks of service failures by focusing on ripple effects. We propose a Bayesian network approach to assessing risks of service failures based on the depdendence relationships among individual service failures. To avoid conceptual misunderstanding and imprecise use in practice, the suggested approach is designed to be executed in three consecutive stages: modeling a Bayesian network; assessing the risks in terms of probabilities of service failures (PSFs) and impacts of service failures (ISFs); and developing a service failure assessment map. A case study of the outpatient consultation service is presented to show the feasibility of our method.

Assessing the risks of service failures has become strategically more important given the large scale and the increased complexity of service systems. It has become the norm for successful companies to consistently monitor the risks of service failures if they are to gain or maintain a competitive edge. Although the damages awarded in service failures vary across industries as well as context of service organizations, many studies have shown that service failures lead to customer dissatisfaction (Parasuraman et al., 1985), customer defection (Keaveney, 1995 and Reichheld, 1996), and negative word of mouth (Richins, 1983). In this situation, service firms are focusing increasing attention on assessing the risks of service failures by defining service standards, establishing a monitoring process, training employees, etc. However, efficient and successful assessment is impeded in many cases by the absence of transparency as well as quality problems due to lack of quantitative and systematic methods (Mefford, 1993 and Harvey, 1998). Consequently, recent years have witnessed a significant increase in attempts to use various models, methods, and tools to utilize existing engineering and technical know-how in the assessment of risks of service failures. In this respect, numerous methods such as quality function deployment (Stauss, 1993), poka-yoke methods (Chase and Stewart, 1994), failure modes and effects analysis (Chuang, 2007), and fault tree analysis (Geum et al., 2009) have been employed. However, while they direct the ways by which we can evaluate the impacts of individual service failures or break a service process into simpler sub-processes, it has been pointed out that these are not useful in large and complex service systems due to lack of dependence relationships among individual service failures. A service process is composed of several components that impact each other in a particular sequence directly and/or indirectly. As a consequence, a service failure at a particular point of a service process can affect the occurrence of service failures at other points (referred to as ripple effects in this study) ( Halstead et al., 1996 and Chuang, 2007). For instance, it is possible that a potential failure point with low occurrence probability can have a significant impact on other failure points. Therefore, analyzing only individual service failures is insufficient; ripple effects among individual service failures should be taken into account at the system level. To counter this problem, we propose a Bayesian network approach to assessing the risks of service failures based on the ripple effects. The primary strength of Bayesian network in service failure analysis lies in modeling and analyzing a large and complex service process that involves several dependence relationships and uncertainties. Firstly, in terms of the large scale of service systems, complex dependence relationships among individual service failures can be easily understood by representing them as a network topology, and can be measured in a quantitative manner. Secondly, with respect to the inherent uncertainties in service failures, various scenarios can be examined based on its ability to infer posterior probabilities. Finally, as far as the data availability is concerned, a Bayesian network is applicable to the service processes in which the objective data do not exist. In addition, a Bayesian network can be easily updated with new information about changes in service systems. Because of these considerations, the suggested approach is designed to be executed in three consecutive stages: service failure modeling, service failure analysis, and service failure assessment. First of all, potential failure points in a service process and their dependence relationships to each other are modeled as a Bayesian network. Second, the risks of service failures are examined in terms of the probabilities of service failures (PSFs) and the impacts of service failures (ISFs). Finally, a service failure assessment map is developed to facilitate effective understanding of the risks of service failures. This study is unique and even exploratory in that the suggested approach is the first attempt that considers ripple effects in service failure analysis. It makes progress in the methodological investigation on service failures by not only introducing the concept of ripple effects but also suggesting substantial methods for reflecting the concept in assessing the risks of service failures. The suggested indicators mirror the ripple effects among individual service failures, thereby measuring the risks of service failures more accurately. Moreover, service failure assessment map can be useful in understanding the characteristics of individual service failures, and can offer implications for reducing the risks of service failures tailored to their characteristics. It is expected that our method can facilitate such decision making process as service redesign and resource allocation, and can serve as a starting point for a more general model. The remainder of this paper is organized as follows. Section 2 presents the general background of service failures and Bayesian network. The proposed approach is explained in Section 3 and illustrated in Section 4 via a case study of an outpatient consultation service. Finally, Section 5 offers our conclusions.

Quality is a minimum qualification for successful services in the service-oriented economy. Service quality has become a primary concern in firms’ strategy for achieving a competitive advantage. Because of its significant effect on the improved organizational productivity, increased market share, and thus enhanced financial performance, a superior quality has been becoming an increasingly important means of differentiation not only for service firms but also for manufacturing firms (Gummesson, 1994, Rapert and Wren, 1998, Hays and Hill, 1999 and Tay, 2003). On the contrary, a poor service can lead to bankruptcy at worst (Fitzsimmons and Fitzsimmons, 2008). Despite its central importance, it has been reported that service organizations have spent less effort on quality improvement relative to manufacturing firms (Berry and Parasuraman, 1991 and Stuart and Tax, 1996). In order to effectively support quality management activities, more practical and substantial methods have been emphasized (Gummesson, 1994 and Stuart and Tax, 1996). In this vein, this study focused on the service failure among the various issues related to service quality. Reducing the risks of service failures is an undisputed critical concern and a prerequisite for establishing quality services. In particular, the ripple effects among service failures have become a major concern of firms as services are becoming more complex and large-scale. The suggested approach that accounts for ripple effects among service failures, which is an inherent nature of service failures, in assessing the risks of service failures is expected to provide more accurate and richer information on the risks of service failures. The contribution and potential utility of this research are threefold. First of all, this study is the first attempt to consider the ripple effects in service failure analysis. It makes progress in the systematic investigation on service failures by not only introducing the concept of ripple effects but also presenting substantial methods for reflecting the concept in assessing the risks of service failures. Second, the suggested method can be applicable to modeling and analyzing various service situations based on the probabilistic methods, regardless of the availability of objective data. Third, the proposed indicators and the service failure assessment map can enhance the understanding on the risks of service failures. We believe that our method can facilitate such decision-making in such managerial issues as service redesign and resource allocation, and can serve as a starting point for developing more general models. Despite all the possibilities offered by the proposed approach, this study still has some limitations that stand in the way of our future research plans. First of all, the suggested approach could be more elaborated by combining other methods of identifying the possible problems systematically for more sophisticated and concrete modeling of Bayesian network. For instance, while characteristics of service activities like the degree of difficulty and customer participation may have an effect on service failures, our approach does not yet model and analyze the detailed information on service activities explicitly. Integrating other types of methods specific to a service process, such as service blueprint and cause and effect diagram, could improve its accuracy and applicability by making unexpected errors more explicit. Secondly, this study has focused only on service failures and the relationships among them. Our method can be expanded to incorporate more various factors. For instance, the impact of service failures on customer satisfaction can be included by employing other methods for measuring customer satisfaction such as Kano analysis, which is capable of categorizing customer needs into several different types according to the relationship between the degree of fulfillment of customer needs and customer satisfaction. Thirdly, evaluation of managerial priorities among service failures can be an interesting topic of importance. Incorporating the factors that influence a decision making of the managerial priorities among service failures is expected to make the suggested approach become a comprehensive and powerful tool for assessing risks of service failures. For instance, the expected cost and time to rectify the problems could be combined with the value of PSFs and ISFs. Investigating the causal relationships between problems would also be a crucial factor for prioritization. Finally, the proposed approach was illustrated by only one industry. More testing on service processes from a wider range of industries could help establish the external validity of our proposed method. These topics can be fruitful in areas for future research.