Online Program

Statistical Practice in North America Telecom Industry
*Andrey Fendyur, University of Calgary 

Keywords: Statistical Practice, Telecom, Methodological Issues, Implementation

The paper communicates a real experience of statistics application for solving performance issues at a major North American telecom multinational company based in Canada. The project was requested to address quality of customer service and increase operations efficiency. A set of statistical techniques was employed: parametric and non-parametric correlation studies, simultaneous and stepwise regressions, data mining and outlier identification. Variable selection and model development issues for North American telecom industry are addressed. Important finding are: most of variance in telecom industry performance comes from outliers while regular performance exhibits low standard deviation, elimination of outliers from the system is a complex and multi-level managerial challenge, variable selection and model development is both a technical and managerial issue, and the implementation of statistics-based improvement recommendations may require changes in the reporting system. The paper benefits academia by identifying methodological needs of telecom industry in statistical techniques. Industry benefits from realizing which performance issues statistical methods may solve.