Categorisation of customer and advisor in contact centers.Abstract: This paper reports on the findings of a research project that had the objective to build a decision support system to categorise customer and service advisor (CSA) within customer contact centre (CCC) environment. We provide the methodology to develop a fuzzy expert system which assigns a new customer or advisor to the pre-defined categories. Categorising of customer and service advisor based on behavioural, demographic and experience variables, was one of the core objectives of the study. In the paper the process of building such a model which categorise customer and advisor through soft computing techniques is described. The behavioural aspects of customer and advisor are not an exhaustive list; but the authors own findings from the case studies at the contact centres. The reported findings and results are very promising, making the proposed model a useful tool in the decision making process, while some of the discussed problems and limitations are of interest to researchers who intend to use soft computing based approaches in other similar real life problems. The framework allows the managers within contact centres to understand the behavioural segments of customers and advisors (CSA) working at the centre and to provide information which would enable them to deliver better customer satisfaction. Keywords: Soft Computing, Fuzzy Expert System, Contact Centres, Customer and Advisor Categorisation, Customised Service Environment. I. Introduction Customer Contact Centres (CCC) and other service industries, often suffer from the problem of high staff turnover and lack of trained staff at the right place for the right kind of customer. The three features of contact centre operations that are crucial to service quality are (a) convenience in fast call handling, (b) cordiality of the advisor, and (c) consistency in advisors [1]. Customer satisfaction may in addition be based on another three dimensions: access, including the advisors communication skills; timeliness, including advisors resolution of queries; and quality, which includes the accuracy, consistency and comprehensiveness of the advisors advice, in addition to the knowledge and politeness of the advisors. Thus the role of the advisors, assumes a heightened importance in customer satisfaction [2]. The advisor/customer relationship can not only be considered critical but sensitive too; for instance there is evidence to suggest that positive employee attitudes appear to be linked to increased customer satisfaction [3]. A model, capable of categorising customer and advisor on behavioural, demographic and experience variables can be a useful decision support tool for a service providing company. Through the proposed methodology for categorisation that incorporates behavioural attributes along with demographic and experience, the authors have demonstrated a way which can help to identify the right amount of information which can enable the advisor to deal with the customer more efficiently and thus providing better customer satisfaction. II. Related Research Service industries have witnessed the widespread use of contact centres in the front of customer service management. In service industries such as hotels, insurance, banking, retail, etc; companies are increasingly paying more attention to customer contact employees to achieve the desired profit and market share goals. Companies are now adopting a people oriented approach as compared to a profit oriented approach [4]. In customer contact businesses, the quality of service delivered cannot be separated from the quality of the service provider. Because service delivery occurs through human interaction, customer service advisor during the service encounter largely determine the level of service quality delivered [5]. The most urgent questions facing most businesses that believe in caring about their customers revolve around (a) what is great service? (b) How can they provide it? (c) How do they get better? [6]. The integration of customer contact centres in daily organizational operations is expected to affect almost all aspects of society from the private sector to government. A Contact Centre can be defined as an internal or outsourced operation largely based on telecommunication and data supports whose primary role is to provide one or many relationship channels for customers, clients, or suppliers (figure 1) [44]. It is critical for companies to identify the need to offer a superior service in order to ensure business survival in a service sector economy [7]. The modern contact centre enables the organization to create a two-way dialogue with their customers [8]. Historically the change within organizations has meant a focus on cost reduction [9]. Cost reduction with stable customer satisfaction has become the priority for organisations [10]. Understanding the customer's capabilities and needs is a necessity when transitioning to a multi-channel environment. [FIGURE 1 OMITTED] A. Customer and Advisor (CSA) Categorisation In contact centres, customer contact employees (i.e. those employees who interact directly with customers over the phone) are called "customer service advisors (CSA's) [4]. CSA's are important for service organisation to provide a link between the external customer and internal operations of the organisation [11]. Although research suggests that CSA's performance is critical to create customer satisfaction, little has been done to analyse which CSA behaviours influence customer encounter satisfaction and which behaviours influence relationship satisfaction [12]. There are five dimensions of CSA's behaviour that influence customer's perceptions mainly; mutual understanding, authenticity, extra attention, competence, and meeting minimum standards [13]. It also showed that advisors attitudinal and behavioural responses can positively and negatively affect customer's perceptions of the service encounter and their judgments of service quality [14]. Many telecoms service operators are subjected to failures in service delivery and better customer satisfaction values because they much depend on CSA to deliver service to their customers. Because of the delivery of the service occurs during the interaction between contact advisors and customers, the attitudes and behaviours of advisors can influence customer's perceptions of the services [14]. As such, this technological change is not simply transforming the methods by which the organization operates, but is impacting the level of skill and education required by both CSA's and management within the contact centre environment. As part of the CRM process, CSA's are increasingly required to identify and act upon cross-selling and up-selling opportunities [15]. The development and widespread use of the Internet for communication and commerce is creating a skills gap within the modern contact centre. There is not enough focus and attention given to the training of CSA's in the area of Internet related support [16]. The loss of an experienced CSA can now be viewed as a major blow to the organization, even more so if they are hired away to a competitor. There is another thought that believes that modern contact centre technologies will reduce the necessity to hire technologically advanced CSA's [17]. Experts we interviewed suggested that as long as CSA's possess good interpersonal skills and are friendly, then screen pops, etc.--should provide them with automated responses to effectively handle inquiries. Understanding and adapting to changes of customer behaviour is an important aspect of surviving in a continuously changing environment [18]. It is necessary to understand individual customers from all levels to enable the advisor to help them more efficiently and thus providing better customer satisfaction. Customer choice of a product depends on explicit requirements, implicit requirements, available options and latent requirements implied by the product [19]. Application of the technologies in contact centre plays an important role in accessing more customers and providing better service quality [5]. As suggested in [20] modeling the users can include statements of how the users within a specific user group behave in certain situations or perform certain functions. B. Soft Computing Techniques Soft computing differs from hard (conventional) computing in that it is tolerant of imprecision, uncertainty and partial truth [21]. For all the available research been carried out in fuzzy logic and the development of fuzzy expert system for customer modelling, little has been done to categorise the advisor (CSA) within the contact centre domain. Since the expert knowledge captured in If ... Then statements is often not naturally true or false, fuzzy sets afford representation of the knowledge in a smaller number of rules, and smooth mapping can be obtained between input and output data [22]. Soft computing technologies provide an approximate solution to an ill defined problem and can create user models in an environment such as contact centre to identify: (a) customer willingness to buy (b) companies prediction towards customer purchase intentions (c) advisor reaction towards customer attitude and (d) customer behaviour towards advisors communication [23]. The elements that a user model captures (goals, plans, preferences, common characteristics of users) can exploit the ability of soft computing of mixing different behaviour and capturing human decision processes in order to implement a system that is more flexible in relation to user interests. The goal of fuzzy expert system is to take in subjective, partially true facts that are randomly distributed over a sample space, and build a knowledge based ES that will apply to them certain amount of reasoning and aggregation strategies to produce useful decisions. Mamdani's fuzzy inference method is the most commonly seen fuzzy methodology. Classification is a most important and frequently used technique in data mining. It is a process of finding a set of models that describe and distinguish data classes or concepts. Clustering is a method of combining objects that are some how similar in characteristics and to provide a grouping of similar records. Fuzzy clustering contains two very different areas: (a) the analysis of fuzzy data and (b) the analysis of the crisp data with the help of fuzzy techniques. Fuzzy clustering utilizes fuzzy partitioning to group data such that any given data sample is allowed to belong to several groups with different degrees of similarity bounded within the range of 0 and 1. With two--step cluster analysis, it groups observations into clusters based on a nearness criterion. Two step clustering is more suitable for cases with less number of data points, but it is limited to crisp cluster boundaries. Two step clusters requires only one data pass in the procedure--it passes the data once to find cluster centres (pre cluster stage) and to assign cluster memberships [24]. Soft computing enables solutions to be obtained for problems that have not been able to be solved satisfactorily by hard computing methods [25][26]. Typically FL has been used to implement applications that are based on a recommendation task. In these applications FL provides the ability of mixing different user preferences and profiles that are satisfied to a certain degree. FL has been used to implement recommendation tasks [27], where fuzzy inference is used for recommendation purposes using user profiles obtained with hierarchical unsupervised clustering. Better communication can be attained through fuzzy logic because of its ability to utilise natural languages in the form of linguistic variables [28]. In [29] fuzzy logic was used to model user behavior and give recommendation using this fuzzy behavior model. A system designed to recommend products in an e-commerce site, according to how well this product satisfies user preferences is presented in [3]. Traditionally, NNs have been used for classification and recommendation in order to group together users with the same characteristics and create profiles [30] [31]. NNs have also been used for recommendation, which predicts the next step for a given user trajectory in a virtual environment [32] and in [33] which models student behavior for an intelligent tutoring system. In general GAs have been used for Recommendation in the form of rules, which can capture user goals and preferences, because they perform a global search and cope better with attribute interaction than algorithms used in data mining, where the search is more local [34] [35]. Decision trees have been used to predict insolvencies in such a way that this prediction can be operationally useful for the decision support process of the telecommunications business handling customer insolvency [36]. Fuzzy clustering processes can be appropriate for grouping users in classes by navigational behaviour. Information in user profiles can be used to customise and identify a user with a social group, done by assigning a general profile related to preferences shown by the user [37]. The combination of NN and fuzzy sets offers a powerful method to model human behavior which allows NFS to be used for a variety of tasks. A Neuro-fuzzy system for modelling human operator behaviour in computer generated forces is represented in [38]. The advantages of soft computing techniques over conventional production rule based expert systems may be characterised as follows: (a) fuzzy sets symbolise natural language terms used by experts; (b) since the expert knowledge captured in "If. ... Then" statements is often not naturally true or false, fuzzy sets afford representation of the knowledge in a smaller number of rules; and (c) smooth mapping can be obtained between input and output data [23]. III. Current Practice Within the current contact centre environment there is a problem of high staff turnover and lack of trained staff at the right place for the right kind of customer. Business needs to assign any available advisor to a customer and provide consistent and good quality of service. There is a need to identify the right amount of information to be displayed on the screen considering both the customer and the assigned advisor background. The key observations and findings are shown below. The technical issues encountered during the visits to the centres are as discussed in table (1) [39]. For the purpose of the project, the authors organised visits to three customer contact centres (CCC) in order to understand the current and overall operations of the contact centre environment. The data and information collected from these centres were captured, understood and combined with current information obtained from the literature and, were used to identify the categorisation for customer and advisor. The main objectives were (a) Identify the current understanding of the customer contact centre environment, its overall operation and working (b) Identify the current bottlenecks into the systems process and for using the different customer--advisor characterisation procedures. A total of five centre managers with around 20 different advisors were interviewed through semi-structured questionnaire. The primary key targets for the centres noticed by the authors were mainly the time the advisor spent offline (transferring calls, looking for information). Call times sometime takes up about minimum of 2 minutes and maximum of 12 minutes. Call Handling Time is the effective time; the advisor spends on calls with the customers. The findings showed that there was a high turn over of CSA and it was necessary to allocate the first available advisor to the customer and then provide the advisor with relevant information. IV. Proposed Methodology The proposed research methodology of this work was to categorise customer and advisors within contact centre environment with the use of soft computing techniques. The model was developed to assign any customer or advisor within contact centre environment to a pre-determined category. Once the identification of the type of customer and advisor is known; it would enable to identify the type of information the system should display to the advisor based on the combination of {customer, advisor} which would enable them to service the customer more efficiently thus providing better customer satisfaction. A fuzzy set allows for the degree of membership of an item in a set to be any real number between 0 and 1, this allows human observations, expressions and expertise to be modeled more closely [40] [41]. The steps followed for the categorisation of customer and advisors are shown in figure (2) and explained later in the paper. [FIGURE 2 OMITTED] A. Data Collection Data was collected with the help of semi-structured questionnaires for advisors (CSA) and team leaders/managers with respect to their demographic variables, experience and behavioural variables within five customer contact centre focusing on fault and sales and looking on single to multi profile business customers. A total of 84 advisors were interviewed and assessed, 60 customer calls were monitored, and total of 19 team leaders and managers were interviewed through the questionnaires. Based on the information which was provided from the team leader, the authors then identified the types of advisors which were going to be used for monitoring and observation on the basis of few important attributes such as; Age Group & Gender, Experience within the company, Education background, and Attitudes (positive and negative). The key observations which the authors noticed for the data collection were advisors characteristics and customer observations (voice). The data was collected through observation by the authors and was verified by the advisors at the contact centres. The variables which were used for the questionnaire were from the literature studies, and also verified with expert judgments of the team leaders at the contact centre. The set of criteria used for advisor data collection are as shown in table (2). The author carried out the check list of the following criteria and attributes of the advisor during the monitoring process and was later verified with the advisors and the team leaders respectively. For customer data collection, the collection was done on the basis of the information provided on the screen of the advisor when the call conversation was in progress, and also the author's monitoring to the calls to identify the behavioural aspect of the customer, before the call and once the call was finished. Once the collection was made, it was later verified with the advisors and the team leaders on their knowledge towards the customer attitude within the CC environment. The criteria and attributes used for customer during the monitoring process are as shown in table (2). B. Clustering Analysis--Identification of customer and advisor categorisation The decision to use clustering techniques in this study is based on the observations that contact centre companies collect high volumes of data which are of different aspects of interactions between the company and its customers. A comparative study was carried out which highlighted the usefulness of the use of clustering analysis for this study [42]. Based on the data collection and analysis of data; attributes derived for customer and advisor are as follows (1) Customer--age, education, financial status, time with company, business value and behavioural analysis (2) Advisor--age, education, experience, previous experience, IT speed, and behavioural analysis The final set of attributes used for customer and advisor (CSA) for the clustering analysis to derive the categories are as shown in table (3) [43] [44]. Once the data was collected and analysed it was verified with the team leaders and managers within the contact centres. Based on the verification the data was structured and analysed using data analysis tool. Through the data analysis tool, the customers and advisor were then grouped according to the attributes shared among each other. The next stage for the development was to identify the categories for customers and advisors through the process of clustering analysis. This section shows the method followed for identification of customer and advisor categorisation through clustering analysis by using two--step process within SPSS analysis. The set of advisor categories derived from the clustering analysis is as shown below in table (4). Six advisor categories (A1-A6) were derived out of the 84 data sets for the advisors [43]. Based on the data structuring from the case studies, data set was designed with 60 samples of customer records and 84 samples (cases) of agents (CSA's). Ten different types of experiments were carried out within the two step cluster analysis method ranging from automatic clustering to a maximum of 10 clusters within SPSS. Based on the clustering few results were noted which were (1) Because of the number of clusters increased from 6-10, the total number of cases each cluster is taking is not properly distributed (2) The number of people (customers and agents) in each cluster is too low for making it a significant cluster (3) The rules derived from the cluster results are repeated and are too close to each other. The set of categories derived for customer from the clustering analysis is as shown below in table (5) [43]. The categories derived for customer and service advisor (CSA) are to be used for further development of the system which would determine the type of customer and service advisor for a given set of data. Once the identification of the customer and advisor is derived; it would enable the system to identify the type of customer and the advisor to provide with the information based on the following combination. C. Development of Fuzzy Expert System This section discusses the steps followed for the development of the fuzzy expert system for customer and CSA categorisation. Fuzzy expert system was developed by using Matlab Fuzzy Logic Toolbox to assign any customer and advisor to that of pre-defined category which was derived from the clustering analysis. On the basis of the expert system, the authors can determine the type of category the customer and the advisors are given which would enable to identify the categorisation of the customer and the advisor. Once the assignment of the category is done, it would enable the model to identify the type of information which is required to be displayed on the screen of the advisor to enable them to help the customer more efficiently and thus providing better customer satisfaction. Step 1--Identify the critical factors and define membership functions and fuzzy sets The first step of the process involved the combination of a list of critical factors based on the literature review and in-depth interviews with the advisor, team leaders, centre managers and systems expert within the environment. The critical factors were the input variables of the fuzzy ES which were as age, education, and financial background, time with the company, business value and behaviours which would identify the type of category they belong to. The development of the model was done by authors own understanding of the current contact centre environment and from the literature studies. Once the selection was done and the model was developed, it was validated with expert judgment from the team leaders at the centre through nine team leaders and managers at three of the case study contact centres. Each linguistic term is defined by a membership function which helps to take the crisp input values and transform them into degree of membership. Triangular and trapezoidal types of membership function are selected to define the variables as shown in figure (3) [43] [44]. [FIGURE 3 OMITTED] Step 2--Construct the Fuzzy Rules Within the fuzzy expert system model once the membership functions of the input and output variables for customers and advisors were derived, fuzzy if ... then rule were written which identified the type of input for customers and advisors. The rule base specifies qualitatively how the output of the system "Category" for the advisor and the customer is determined for various instances of the input variables. The variables are age, education, financial status, time with company, business value, experience, and behavioral attributes. A total of forty-five rules were derived within the expert system. A sample of the rules derived for the fuzzy logic expert system are shown below and explained as below. The rules for advisors were selected from the understanding of the advisor input attributes and the results from the clustering analysis are explained: IF age is young, education is school, experience is novice, previous exp is low, IT speed is slow, positive behaviour as friendly and negative behaviour as unaware THEN the selected category is A1 as shown in table (6). If ... Then rules for customer were derived similarly to that of the advisors within the fuzzy expert system model. Some of the rules derived for the system are as explained: IF age is young, education is school, financial status is poor, time with company is low, business value is low, positive behaviour is none and negative behaviour is aggressive THEN the category selected is C1 (table 7). D. Validation of Fuzzy Expert System With respect to the model, the authors carried out experiments with the fuzzy expert system model by changing the input variable values and monitoring the change in the output which showed the change in the category for customer and advisors. The results which we analysed are the set of new data points from sampling for customer and advisors (table 8 and 9). The results derived from the experiments carried out within the expert system model were validated within the contact centre environment with the team leaders and managers. 1. Advisors Experimental Examples This section highlights the experimental examples which were carried out within the fuzzy expert system model to assign the customer and advisor to that of the pre-defined category from the clustering analysis. Experiment 5 and 7 shown below are the ones which were different during the validation process. Ex. 5--If Age = 51, Education = 27, Experience = 8.6, IT Speed = 2.8, Previous Exp = 5, Positive Behaviour = 5, Negative Behaviour = 1.2. Then Advisor Category output is 25 which determine that the category for advisor is A5. Ex. 7--If Age = 22.8, Education = 18, Experience = 2, IT Speed = 2.5, Previous Exp = 2.1, Positive Behaviour = 3.2, Negative Behaviour = 1. Then Advisor Category output is 26.1 which determine that the category for advisor is A6. For example, the input values in the first experiment is for age=21.5, and from our membership functions it justifies that the input variable for age is young; education=12=college, experience=5=5-10 yrs, previous exp=1.8=low, IT Speed=1.5=slow, positive behaviour=5.5=friendly. As shown above in table (8) the CC validation shows the results from the validation carried out with the team leader's expert judgment at the contact centre. 2. Customer Experimental Examples The customer experimental results from the expert system, which differ during the validation process, are discussed as below and shown in table (9). Ex. 6--If Age = 40, Education = 25, Financial Status = 5, Time with company = 10, Business Value = 8.5, Positive Behaviour = 9, Negative Behaviour = 0.4. Then Customer Category output is 20 and category is C4 Ex. 7--If Age = 50, Education = 10, Financial Status = 4.3, Time with company = 6.5, Business Value = 0, Positive Behaviour = 7, Negative Behaviour = 3. Then Customer Category output is 30 and category is C6 Based on the model, the authors identified that the results derived from the model, assigned a customer with the predetermined category which were derived from the clustering. These results were also validated with the team leaders at the contact centre to verify that the given selection of the predetermined categories for customer was properly justified. V. Validation The information and the results from the model were verified through team leaders and managers at three of the contact centres where the case studies were carried out. A total of nine team leaders and managers were interviewed with the help of an open set questionnaire, showing the categories derived and the assignment of a particular customer or advisor to these categories through the help of the fuzzy expert system tool developed. From the validation, it was noticed that the expert judgment did correspond to that of the results from the expert system model framework for 80% of the overall experiments that were carried out. The experimental results in table (8) and (9) shows the assignment of a particular customer and advisor to the categories which were derived from clustering. Based on experiment 5, the expert system assigned category A5 to the advisor. However from validation with team leaders it revealed that the category should be A4. On the basis of the validation the changes were made with respect to behavioural attributes from friendly behaviour to customer focus behaviour. Experiment 7 revels that the expert system assigned category A6, which on further validation with team leaders at the contact centres fall into A2 category. The reasons for this swift change in selection of category were due to: a. Education level to be high. b. Positive behaviour to be attentive. c. Less amount of negative behaviour. The rules were fine tuned to predict A2 category and share characteristics of that category. For customer categorisation, the results from the expert system for experiment 6 and 7 did not match that to the validation from the team leaders at CC (table 9). Necessary modifications were carried out within the expert system to assign a category to customer to match with the validation results from the team leaders. As seen in experiment 6, the changes made were education level was changed from graduate to college level to assign customer with C6 category. Experiment 7 revealed that expert system assigned C6 category which on further validation fall into C4 category. The changes made within the expert system were: a. Customer time within company. b. Positive attitude towards the advisor and c. Less amount of negative attitude shown from the customer. VI. Discussions and Future Research This paper reports on a long term research project that was set to investigate the feasibility of soft computing techniques for categorising customer and advisors with specific application in the contact centre sector. However, the findings can have implications far beyond this specific case study. The reported results are considered significant for a number of reasons. Firstly, the study involved a real life problem. This means that the data used, the requirements and objectives set and the scale of the experiment corresponds to a real problem, as defined by a major telecommunications operator. Since modern telecommunication companies have similar characteristics in terms of services provided to their customers, databases and information systems design for monitoring customer characteristics used, the findings of the experiment are applicable to other sectors of service industry. Secondly, throughout this study, it was recognised that domain experts were continuously providing expertise and intuition by directing and pointing to the matters that were important for the company and its customers. It was also shown that the original steps of the process required a mixture of tools and experts intuition, relating to the problem of defining the data set and selection of variables describing the required modelling features. Further research should develop a framework to map customer and advisor behavioural and demographic information directly to the type of information required to be presented on the screen; rather than a fixed template based approach. This paper is focused on the development of customer and advisor (CSA) categorisation within contact centre environment. A fuzzy expert system was developed to assign any customer or advisor to that of the pre-determined category from the clustering analysis. The results showed the assignment from the expert system for the categorisation of the customer and advisor which was validated with the team leaders at the case study contact centres. The proposed methodology can be applied in contact centre environment to minimise the problems of information overload, right amount of information at the right time, retention of advisors, and customer satisfaction through speed of response. Finally an interesting discussion concerns the degree to which the study can be generalised and reused in other problems in service industry. Some of the findings and specific techniques used are of general value such as the clustering approach used for customer and advisors and the discussion on the use of soft computing based approach to assign each customer and advisor to a pre-defined category. More specifically, a large area of problems related to the identification of information and categorisation of customer in the service industry. They bear many similar characteristics to the one described here, and can be based on very similar process, where apart from the initial stages of problem definition and original variables selection, the rest of the process can be repeated. Also for the specific problem of behavioural modelling of customer and advisor within contact centres, the methodology described in the paper can be reused, except of definition of the advisor data. A data mining tool to handle such exception of data can be used as one of the objective for further research. Acknowledgments The authors wish to acknowledge the support of the Engineering and Physical Sciences Research Council (EPSRC), BT Telecommunication Plc, and Decision Engineering Centre at Cranfield University UK. References [1] Haymarket Business Publications. The telebusiness report--the definitive report on telebusiness in the UK to the year 2005. Marketing Direct Magazine 5-17. UK, 1998. [2] Brown, G. and Maxwell, G. "Customer Service in UK call centres: organisational perspectives and employee perceptions", Journal of Retailing and Consumer Services, Vol. 9, No. 6, pp. 309-316, 2002. [3] Schmitt, C., Dengler, D., and Bauer, M. "Multivariate preference models and decision making with the MAUT matching". In: 9th International Conference on User Modelling , Vol. 2702, Johnstown, USA, Lecture Notes in Computer Science, Springer--Verlag GmbH, USA, 2003. [4] Malhotra, N. and Mukherjee, A. "The relative influence of organisational commitment and job satisfaction on service quality of customer contact employees in banking call centres", Journal of Service Marketing, Vol. 18, No. 3, pp. 162-174, 2004. [5] Bennington, L., Cummane, J. and Conn, P. "Customer satisfaction and call centers: An Australian Study", International Journal of Service Industry Management, Vol. 11, No. 2, pp. 162-173, 2000. [6] Feinberg, R., Kim, I., Hokama, L., Ruyter, K. and Keen, C. "Operational determinants of caller satisfaction in the call center", International Journal of Service Industry Management, Vol. 11, No. 2, pp. 131-141, 2000. [7] Prabhaker, R., Sheehan, J. and Coppett, J. "The power of technology in business selling: call centres", Journal of Business and Industrial Marketing, Vol. 12, No. 3/5, pp. 220-232, 1997. [8] Boyd, C., Blood, S., and Wright, T. "UK Contact Centre Market--BT Retail Report Series, No. 220082660, Gartner, UK, 2002. [9] Koole, G., Mandelbaum, A., Gans, N., Ramdas, K. and Fisher, M. "Telephone Call Centers: tutorial, review and research prospects", Journal of Manufacturing and Service Operations Management, Vol. 5, No. 2, pp. 79-141, 2003. [10] Hawkins, L., Meier, T., Nainis, W., and James, H. "The evolution of the call centre to customer contact centre!" In: Information Technology Support Centre, US, 2001. [11] Zeithaml, V. and Bitner, M. Services Marketing, McGraw Hill, New York, 2000. [12] Moshavi, D. and Terborg, J. "The job satisfaction and performance of contingent and regular customer service representatives", International Journal of Service Industry Management, Vol. 13, No. 4, pp. 333-347, 2002 [13] Dolen, W., Ruyter, K. and Lemmink, J. "An empirical assessment of the influence of customer emotions and contact employee performance on encounter and relationship satisfaction", Journal of Business Research, Vol. 57, No. 4, pp. 437-444, 2004. [14] Hartline, M. and Ferrell, O. "The management of customer contact service employees: an empirical investigation", Journal of Marketing, Vol. 60, pp. 52-70, 1996 [15] Storey, A. and Cohen, M. "Exploiting response models --Optimizing cross-sell and up-sell opportunities in banking". In Proceedings of the Eight ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Jul 23-26 2002, Association for Computing Machinery, Edmonton, Alta, Canada, pp. 325-331, 2002. [16] Rose, E. and Wright, G. "Satisfaction and dimensions of control among call centre customer service representatives", International Journal of Human Resource Management, Vol. 16, No. 1, pp. 136-160, 2005 [17] Mohr, L. and Bitner, M. "The role of employee effort in satisfaction with service transactions", Journal of Business Research, Vol. 32, No. 3, pp. 239-252, 1995. [18] Chiu, C. "A case-based customer classification approach for direct marketing", Expert Systems with Applications, Vol. 22, No. 2, pp. 163-8, 2002. [19] Zeelenberg, M. and Pieters, R. "Beyond valence in customer dissatisfaction: A review and new findings on behavioral responses to regret and disappointment in failed services", Journal of Business Research, Vol. 57, No. 4, pp. 445-455, 2004. [20] Bushey, R., Mauney, J., and Deelman, T. "The development of behavior-based user models for a computer system". In: Proceedings of UM99: 7th International Conference on User Modeling, 20-24 June 1999, Springer, Banff, Alta., Canada, pp. 109-18, 1999. [21] Zadeh, L. "The birth and evolution of fuzzy logic (FL), soft computing (SC) and computing with words (CW): a personal perspective", World Series in Advances in Fuzzy Systems, pp. 811-819, 1996. [22] Ngai, E. and Wat, F. "Design and development of a fuzzy expert system for hotel selection", Omega, Vol. 31, No. 4, pp. 275-286, 2003. [23] Frias-Martinez, E., Magoulas, G., Chen, S. and MacRedie, R. "Modeling human behavior in user-adaptive systems: Recent advances using soft computing techniques", Expert Systems with Applications, Vol. 29, No. 2, pp. 320-329, 2005. [24] Johann, B., Knut, W., and Melanie, V. "SPSS Two Step Cluster--A first evaluation, Company Report from SPSS available: http://www.statisticalinnovations.com/products/twostep.pdf (accessed 2004), 2001. [25] Dote, Y. and Ovaska, S. "Industrial applications of soft computing: a review", Proceedings of the IEEE, Vol. 89, No. 9, pp. 1243-65, 2001. [26] Zha, X. "Soft computing framework for intelligent human-machine system design, simulation and optimization", Soft Computing, Vol. 7, No. 3, pp. 184-98, 2003. [27] Nasraoui, O. and Petenes, C. "Combining web usage mining and fuzzy inference for website personalization". In Proceedings of the Web Mining as Premise to Effective Web Applications, Washington DC, USA, 2003. [28] Kuanchin, C. and Gorla, N. "Information system project selection using fuzzy logic", IEEE Transactions on Systems, Man & Cybernetics, Part A (Systems & Humans), Vol. 28, No. 6, pp. 849-55, 1998. [29] Ardissono, L. and Goy, A. "Tailoring the interaction with users in electronic shops". In Proceedings of tth International Conference on User Modeling, 20-24 June 1999, Springer, Banff, Alta., Canada, pp. 35-44, 1999. [30] Yasdi, R. "A literature survey on applications of neural networks for human-computer interaction", Neural Computing & Applications, Vol. 9, No. 4, pp. 245-58, 2000. [31] Fix, E. and Amstrong, H. "Modeling human performance with neural networks". In Proceedings of the International Joint Conference on Neural Networks (IJCNN), Vol. 1, SanDiego, CA. USA, 247-52, 1990. [32] Sas, C., Reilly, R., and O'Hare, G. "A connectionist model of spatial knowledge acquisition in a virtual environment". In Proceedings of the 2nd Workshop on Machine Learning, Information Retrieval and User Modeling, USA, Springer Verlag, USA, 2003. [33] Beck, J., Jia, P., Sison, J., and Mostow, J. "Predicting student help request behavior in an intelligent tutor for reading", In Proceedings of the 9th International Conference on User Modeling, Vol. 2702, Johnstown, USA, Springer Verlag--Lecture Notes on Computer Science, USA, 2003. [34] Ishibuchi, H., Nozaki, K., Yamamoto, N. and Tanaka, H. "Selecting fuzzy if-then rules for classification problems using genetic algorithms" IEEE Transactions on Fuzzy Systems, Vol. 3, No. 3, pp. 260-70, 1995. [35] Rees, J. and Koehler, G. "Evolution in groups: a genetic algorithm approach to group decision support systems", Information Technology & Management, Vol. 3, No. 3, pp. 213-27, 2002. [36] Daskalaki, S., Kopanas, I., Goudara, M. and Avouris, N. "Data mining for decision support on customer insolvency in telecommunications business", European Journal of Operational Research, Vol. 145, No. 2, pp. 239-55, 2002. [37] Martin-Bautista, M., Kraft, D., Vila, M., Chen, J. and Cruz, J. "User profiles and fuzzy logic for Web retrieval issues", Soft Computing, Vol. 6, No. 5, pp. 365-72, 2002. [38] George, G. and Cardullo, F. "Application of Neuro-fuzzy systems to behavioural representation in computer generated forces", Technical Report available at: http://www.link.com/pdfs/neuro-fuzzy.pdf (accessed 2004), 1999. [39] Shah, S., Roy, R., Tiwari, A., and Majeed, B. "Customised customer support using a soft computing approach", In Proceedings of the International Conference on Computational Intelligence for Modelling, Control and Automations, Vienna, Austria, IEEE Computer Society, Vol. 1, pg 939-945, 2005. [40] Wong, K. "Data mining using fuzzy theory for customer relationship management". In Proceedings of the 4th Western Australian Workshop on Information Systems Research (WAWISR 2001), Perth, Australia, Western Australian University, Australia, 2001. [41] Zadeh, L. "Soft computing and fuzzy logic", IEEE Software, Vol. 11, No. 6, pp. 48-56, 1994. [42] Shah, S., Roy, R., and Tiwari, A. "Technology selection for human behaviour modelling in contact centres", In Decision Engineering Report Series, Cranfield University Press, UK, 2005. Online Link--http://hdl.handle.net/1826/1212 [43] Shah, S., Roy, R., and Tiwari, A. "Development of fuzzy expert system for categorising customer and advisors in contact centres", In Proceedings of the 10th Online World Conference on Soft Computing in Industrial Applications, Cyberspace, Applications of Soft Computing--Recent Trends, Springer Series on Advances in Soft Computing, Vol. 36., ISBN 10:3-540-29123-7, 2005. [44] Shah, S., Roy, R., and Tiwari, A. "Optimising customer support in contact centres using a soft computing approach", In Decision Engineering Report Series, Cranfield University, UK, 2006. Online Link--http://hdl.handle.net/1826/1211 Author Biographies Satya Shah received his B.Eng. (Hons) in Electrical and Electronics Engineering in 2001 and M.Sc in Telecommunications Engineering 2002 (with Distinction level), all from London South Bank University, UK. From 2003 onwards he is working as doctoral researcher in the Decision Engineering Centre at Cranfield University on Application of Soft Computing in Telecommunications. The research is jointly sponsored by Education and Physical Research Council, UK (EPSRC) and British Telecommunications, UK. He is currently towards completion of his PhD thesis at the university. His main research interest is intelligent systems development, expert systems, soft computing, call/contact centre optimizations and design environment. Rajkumar Roy is the Head of the Decision Engineering Centre. Professor Rajkumar Roy has a background in manufacturing engineering and artificial intelligence. He started his professional career in manufacturing industry back in 1987, and worked in the area of knowledge engineering, decision support and shop floor implementation of expert systems. His research projects have a strong focus on industrial applications. His research focus includes soft computing applications to solve real life problems in industry. He is the Editor in Chief of the Applied Soft Computing Journal from Elsevier and Decision Engineering Book Series from Springer. He is a Chartered Engineer and members of international organisations such as IEEE, IED, and ACostE and has co-authored four books in his area of research. Ashutosh Tiwari is a Lecturer in Decision Engineering at Cranfield University. He graduated in Mechanical Engineering from the Indian Institute of Technology (IIT) Kanpur. He started his research career at Cranfield University in 1998 through the MSc in Engineering and Management of Manufacturing Systems. From November 1999 to January 2002, Dr Tiwari worked as a Research Officer in Cranfield University on an EPSRC project. He is actively involved in research, Virtual Organisation teaching, I-DEAS teaching in postgraduate modules and short courses, CAD/CAM software evaluation and support. Basim Majeed is a Principal Research Professional at the Intelligent Systems Research Centre within BT Research and Venturing at Adastral Park. He holds a Masters degree (1987) and a PhD degree (1992) in Intelligent Control Systems from the University of Manchester. He is part of a team working in the area of real-time business intelligence and business process management. He is a Member of the IET and IEEE, and a Chartered Engineer. Satya Shah (1), Rajkumar Roy (1), Ashutosh Tiwari (1) and Basim Majeed (2) (1) Decision Engineering Centre, School of Applied Sciences, Cranfield University Cranfield, Bedfordshire, MK43 0AL, United Kingdom {s.shah,r.royanda.tiwari}@cranfield.ac.uk (2) Computational Intelligence Group, BT, PP 12, Orion Building, Adastral Park. Ipswich, IP5 3RE. UK. Tel: +44(0)1473605320 basim.majeed@bt.com
Table 1. List of major concerns observed in the contact centre [39].
Contact Centres
Categories of Contact Centre Contact Centre Contact Centre
Concerns 1 2 3
System Complex Long time Wrong calls
Functionality systems. Not needed on transferred
easy to training to through IVR.
understand and enable the
use. advisor
understand
properly.
Advisor Time Time sped on Time spend on Advisor spent
Management arrangement of collection of time other than
work once the prior answering the
call is information customer query
finished. about the in searching
customer, for
knowing later information.
the call is not
for that
specific
department.
Customer Understanding Unsure of the Lack of
Interactions the delay which type of sufficient
occurs when the problems information
advisor is encountered. which can
processing the Lack of enable the
customer query. information advisor to help
from advisor. the customer.
Advisor Lack of Up to date High staff
Training training given information on turnover
to understand the changes results it
the system more made to the increase
efficiently. system is not training
always given to budget.
the advisors.
Call Problems Calls to the Long queues and Call quality
wrong long wait (e.g. overseas
department times. contact
which needs to centres).
be transferred.
Future Need to be a Customer focus User friendly
Challenges user friendly and system
and less time satisfaction. appropriate
consuming Real Time training.
system. Visualisation Systems that
of information only shows the
display relevant
information
required to
deal customer
query
Table 2. Advisor and Customer categorisation criteria's.
Situation / Condition Criteria Categories
* CS 's Demographic Age & sex Age 18-30, 3050,
Values Above 50
Sex Male, Female
* CSA's Knowledge of Knowledge Level Knowledge--School,
service College, Graduate
* Level of experience Service Experience Service Experience--
Novist (<1 yr),
Experienced (> 1yr)
* Computer experience IT experience IT experience--Little,
moderate, extensive
* Characteristics Char. Behaviour Char. Behaviour--
Behaviour with the Competence,
customer Altitude,
Communication
* Speed with the Speed Speed with Service--
service Slow, Medium, Fast
* Relationship with Relationship Relationship--Helpful,
the customer Very helpful
* CSA's positive and Positive & Negative Positive--Attentive,
negative emotions Concentrated,
while dealing with Joyful, Happy
customers Negative--Sad,
Discouraged, Angry,
Mad
* Mutual understanding Understanding Understanding--Open,
of the customers Close
situation
* CSAs Competence Competence Competence--Capable,
Efficient,
Organized, Thorough
* Performance Performance Performance--
Understanding,
Attention, Meeting
standards
Situation / Condition Criteria Categories
* Customer Demographic Age & sex Age 18-30, 30-50,
Values Above 50
Sex Male, Female
* Different types of Customer Types Customer Types--
customers Prospectus,
Responders, Active,
Former
* Customer's education Education Level Education--School,
College, Graduate,
Professional
* Financial Level of Income Group Income--Poor, Average,
the customer Good
* How long has the Relationship Relationship--Old--
customer been with Less than 2 yrs,
the company More than 2 yrs, New
* How often does the Lifecycle Buying Patterns--
customer buy from Frequently, Rarely
the company
* What is the Purchasing Power Purchasing Power--Low,
customers purchasing High
power in the family
* Customers payment Payment Problems Payment--Regular,
difficulty in Irregular
previous
transactions
* Customers method & Complaint Frequency Comp. Freq.--Rarely,
frequency of Regular, Often
complaints of
service
* Customer Emotions Positive & Negative Positive--Attentive,
Concentrated,
Joyful, Happy
Negative--Sad,
Discouraged, Angry,
Mad
Table 3. Advisor and customer variables within clustering analysis.
Advisors (CSA)
1. Age--young, middle age, old
2. Education--school, college, graduate, professional
3. Experience--novice, medium, senior
4. IT Speed--slow, medium, fast
5. Previous Exp--low, moderate, extensive
6. Positive Behavior--attentive, friendly, customer focus
7. Negative Behavior--unaware, annoyed, angry
Customer
1. Age--young, middle age, old
2. Education--school, college, graduate professionals
3. Financial Status--poor, average, good
4. Time with Company--low, moderate, high
5. Business Value--low, medium, high
6. Positive Behavior--joyful, co-operative, understanding
7. Negative Behavior--angry, annoyed,
Table 4. Advisor Categorisation.
A1 A2 A3
Categories (Novice (Customer (Annoyed
Attributes Advisor) Focus Advisor) Advisors)
AGE 18-25 18-25 25-40
EDUCATION SCHOOL GRADUATE GRADUATE
EXPERIENCE <1 YRS <1 YRS 5-10 YRS
IT SPEED LOW MEDIUM HIGH
PREVIOUS NONE NONE EXTENSIVE
EXPERIENCE
POSITIVE BEHAV. -- CUSTOMER ATTENTIVE
FOCUS
NEGATIVE BEHAV. ANGRY & ANNOYED ANNOYED
UNAWARE
Total Cases 16 18 20
(out of 64)
A4 A5 A6
Categories (Experience-- (Experienced-- (Attentive
Attributes Cust. Focus) Friendly) Advisor)
AGE 40-50 50+ 18-25
EDUCATION PROF. PROF. COLLEGE
EXPERIENCE 10-15 YRS 15+YRS 1-5 YRS
IT SPEED HIGH MEDIUM MEDIUM
PREVIOUS EXTENSIVE MODERATE NONE
EXPERIENCE
POSITIVE BEHAV. CUSTOMER FRIENDLY ATTENTIVE
FOCUS
NEGATIVE BEHAV. -- -- UNAWARE
Total Cases 4 7 19
(out of 64)
Table 5. Customer Categorisation.
A1 A2 A3
Categories (Novice (Customer (Annoyed
Attributes Advisor) Focus Advisor) Advisors)
AGE 18-25 25-40 18-25
EDUCATION SCHOOL GRADUATE COLLEGE
FINANCIAL STATUS POOR GOOD POOR
TIME WITH COMPANY 1-5 YRS 5-10 YRS >1 YRS
BUSINESS VALUE LOW MEDIUM MEDIUM
POSITIVE BEHAV. -- UNDERSTANDING JOYFUL
NEGATIVE BEHAV. ANGRY & ANGRY ANNOYED
AGGRESSIVE
Total Cases 12 9 13
(out of 60)
A4 A5 A6
Categories (Experience-- (Experienced-- (Attentive
Attributes Cust. Focus) Friendly) Advisor)
AGE 40-50 25-40 40-50
EDUCATION PROF. PROF. COLLEGE
FINANCIAL STATUS AVERAGE GOOD AVERAGE
TIME WITH COMPANY 10+ YRS 5-10 YRS 5-10 YRS
BUSINESS VALUE HIGH HIGH LOW
POSITIVE BEHAV. JOYFUL UNDERSTANDING CO-OPERATIVE
NEGATIVE BEHAV. -- AGGRESSIVE ANNOYED
Total Cases 6 11 9
(out of 60)
Table 6. Sample of Advisor Fuzzy If..... Then Rules.
Previous IT
Age Education Experience Experience Speed
Young School Novice Low Slow
Middle Graduate Medium Moderate Medium
Old Profess. Senior Extensive Medium
Young College Novice Moderate Fast
Young Graduate Novice Low Fast
Old Graduate Senior Extensive Fast
Young Graduate Medium Moderate Fast
Positive Negative
Age Behaviour Behaviour Category
Young Friendly Unaware A1
Middle Attentive Annoyed A3
Old Focus None A5
Young Focus Unaware A6
Young Attentive Annoyed A2
Old Friendly None AA
Young Attentive None A2
Table 7. Sample of Customer Fuzzy If.... Then Rules.
Financial Time with Business
Age Education Status Company Value
Young School Poor Low Low
Middle Graduate Good Moderate Low
Old Graduate Average Moderate Medium
Young College Poor Low Medium
Middle Professional Good Moderate High
Old Professional Average High High
Middle School Poor H Medium
Positive Negative
Age Behaviour Behaviour Category
Young None Aggressive C1
Middle None Annoyed C2
Old Understanding Angry C6
Young Co-operative None C3
Middle Joyful None CS
Old Joyful Annoyed C4
Middle None Aggressive C1
Table 8. Sample of Customer Fuzzy If.... Then Rules.
Previous IT
No Age Education Experience Experience Speed
1 21.5 12 2 1.8 1.5
2 30 21 4.2 5 4
3 20 5 1 0.5 1.3
4 28 24.6 0 1.5 3
5 51 27 8.6 5 2.8
6 43 16.5 7 5.1 4.2
7 22.8 18 2 2.1 2.5
8 15 2 1 1 0.8
Positive Negative CC
No Behavior Behavior Output Category Validation
1 5.5 3.8 25 A6 A6
2 1.8 5 10 A3 A3
3 1.2 1.8 5 A1 A1
4 8 4 5 A2 A2
5 5 1.2 25 A5 A4
6 6 0 20 A4 A4
7 3.2 1 26.1 A6 A2
8 7 0 2.33 A1 A1
Table 9. Sample of Customer Fuzzy If.... Then Rules.
Financial Time with Business
No Age Education Status Company Value
1 20 10.2 2 0.8 4
2 25 5 3 5 2.5
3 30 7 8.9 9 6.8
4 36 16.5 6.5 4.5 5
5 28 10.7 0 0 5
6 40 25 5 10 8.5
7 50 10 4.3 6.5 0
8 18 1.2 1.5 3 1.2
Positive Negative Output CC
No Behaviour Behaviour Value Category Validation
1 10 1 15 C3 C3
2 1.2 5 5 C1 Cl
3 5 0 25 C5 C5
4 6.2 10 10 C2 C2
5 10 2.1 15 C3 C3
6 9 0.4 20 C4 C6
7 7 3 30 C6 C4
8 1.2 8 5 C1 Cl
|
|
||||||||||||||||||||
Printer friendly
Cite/link
Email
Feedback
Reader Opinion