|Abstract:|| In the last decades the development of innovative marketing solutions for companies have made telephone call centers an integral part of many businesses and so their economic role has become signicant and growing. Telephone call centers allow groups of agents to serve customer remotely,and they have become a primary contact point between customers and their
service providers. While call centers are technology-intensive operations, often 70% or more of their operating costs are devoted to human resources, and to minimize costs their managers carefully track and seek to maximize agent utilization.
Well-run call centers adhere to a sharply-dened balance between agent efficiency and service quality, and to do so, they use queueing-theoretic models.
Inputs to these mathematical models are statistics concerning system primitives, such as the number of agents working, the rate at which calls arrive, the time required for a customer to be served and the length of time customers are willing to wait on hold before they hang up the phone and abandon the queue. Outputs are performance measures, such as the distribution of time that customers wait on hold and the fraction of customers that abandon the queue before being served. In practice the number of agents working becomes a control parameter which can be increased or decreased to attain the desired efficiency-quality tradeoff. The call center industry in Italy employs more than 300 thousand people and grows with an annual rate of 10%, for a turnover above one billion euros.
The quality and operational efficiency of these services can be extraordinary. In a large, best-practice call center, many hundreds of agents can cater to many thousands of phone callers for hour; agent utilization levels can average between 90% to 95%; no customer encounters a busy signal and, in fact, about half of the customers are answered immediately; the waiting time of those delayed is very short, and the fraction that abandon while waiting varies form the negligible to a mere 1 2%. At the same time, these example of best practice represent the exception, rather than the rule; most call centers do not consistently achieve such simultaneously high levels of service quality and efficiency. In part, this fact may be due to a lack of understanding of the scientic principles underlying best practice.
Therefore accurate forecasting of call arrivals is critical for staffing and scheduling of a telephone call center; this is the main focus of the multivariate statistical models and methods considered in this thesis.
Our methods are illustrated using real data coming from a big call center of one of the most important Italian telephone company.|