|Ph.D Student||Khudyakov Polyna|
|Subject||Statistical Analyses of Call Center Data|
|Department||Department of Industrial Engineering and Management||Supervisors||Professor Emeritus Paul Feigin|
|Professor Malka Gorfine-Orgad|
|Full Thesis text|
This study looks into management problems of call centers and the opportunity to analyze a large quantity of data collected over a long time period. The aim is to develop and apply methods of statistical analysis to call center data in order to identify basic problems, to find the sources of such problems, to develop ways for their solution and to estimate their possible impact.
We consider Markovian models for a call center with and without an Interactive Voice Response (IVR) system and approximate performance in the Quality and Efficiency Driven (QED) asymptotic regime, which is suitable for moderate to large call centers. In contrast to exact calculations, the approximations are both insightful and easy to implement (for up to thousands of agents). We validate our models against data from a US Bank Call Center, and our results demonstrate that simple models still provide very useful descriptions of much more complex realities.
We also present a statistical analysis of customer patience. This work is the first attempt to apply frailty models to an analysis of customers' patience while taking into account the possible dependency between calls of the same customer, and estimating this dependency. We extended the estimation technique of Gorfine et al. (2007) to address the case of different unspecified baseline hazard functions for each call, to address the case in which the customer's behavior changes as s/he becomes more experienced with the call center services. Then, we provided a new class of test statistics for hypothesis testing of the equality of the baseline hazard functions. The asymptotic distributions of the test statistics were investigated theoretically under the null hypothesis and certain local alternatives. We also provided a variance estimator. The properties of the test statistics, under finite sample size, were studied by an extensive simulation study and verified the control of Type I error and our proposed sample size formula. The utility of our proposed estimating technique is illustrated by the analysis of the call center data of an Israeli commercial company that processes up to 100,000 calls per day.