טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentTali Zeitler
SubjectHow to Utilize Callbacks: Time-Varying Queues with a Call
Back Option
DepartmentDepartment of Industrial Engineering and Management
Supervisor Dr. Yom-Tov Galit


Abstract

Many service centers nowadays offer their customer a callback option, in which customers are waiting offline for the agent to call them back. The customers, who arrive to the queue, are informed regarding the anticipated waiting time and the maximal delay till callback. They can choose between waiting online in the regular queue or wait offline for a callback. In many cases, the maximal offline delay guarantee is constant, offered to all customers and independent of the systems load. For example, big companies in Israel, are required by law to offer such callback option if the online wait is longer than 3 minutes, and they are required to call back to the customers within 3 business hours. Such a policy is in the interest of the company, as it has the potential of both improving service performance, and server utilization, by balancing load between overloaded and underloaded periods. The problem with current practice is twofold: 1. It ignores the fact that the time till returning influences customer choice as well. The deterministic rule ignores customer sensitivity to the online and offline delay values. 2. It also ignores the daily pattern of load, and may move customers from overloaded periods to even more overloaded ones, causing worse delays instead of improving them. Therefore, planning delay guarantees that vary over time and choosing when to offer them at all is important. In this work, we provide a simulation based optimization algorithm, called the Iterative Delay Guarantee (IDG) algorithm, to determine what offline waiting guarantees a company should offer in a time-varying environment. Those guarantees change over time and depend on service level targets that the company wishes to provide. Moreover, they depend on the load of the current state of the system as well as the anticipated load of future time-periods.