The number of call centers for customer service has grown exponentially in Germany in recent years. The quality of a call center is crucial to how well customers feel cared for orders or for service from a company.

The human resource planning for a call center is subject to many uncertainties, as for example, not only the number of the calling customer per day, but also the distribution of the customers over the day varies greatly. A simple approach to modeling of a call center is using so-called Erlang C-formula. However, this only approach takes into account a total of 3 parameters and can thus does not take into account many determining factors such as the impatience of customers, different skill levels of agents, etc.

Every day is a number of customers specified by a probability distribution will enter the system as fresh calls. These calls are collected in a queue. The ACD (automated call distributor) maintains a list of currently available agent, by which is checked which customer can be routed to which agent. The processing of the waiting customers and the assignment of the agents to the customers is controlled by a score values priority system. For the customer this score depends on the customer type and the duration of the elapsed wait time. For the agents the score is determined by their idle time. If there is a appropriate pairing, the customer with the highest score will be forwarded to the agent with the highest idle time score value. After talking it is decided whether the customer leaves the system or whether he needs to be forwarded, to join further processing steps due to the customer's requirement and the conversation history. In this case, the customer is lined up after the conversation again into the queue. The agent goes after talking in a post processing stage. Only after he logs in to the ACD again he is as available for the next call. By each customer in the queue, it is assumed that he is willing to wait a certain amount of time. If this so-called waiting time tolerance is exceeded, he leaves the system without having been served. The simulation model takes into account both the case that the customer finally leaves the system, as well as the case that the customer repeats its call attempt by a randomly distributed time (repeat distance). Retry time distributions and retry probability values are also parameters of the simulation model.

By using the simulation model one can check in a simple manner the impact parameter changes will have on the performance of the system or how many agents are needed to achieve certain objectives concerning accessibility and level of service.

• Schematic diagram of of a complete call center model (pdf)
• Simulation and optimization
• Input parameters for simulation
• Screenshots
• Homepage / download address for Callcenter Simulator
  
• Industry project "Virtual call center"

• A. Herzog: Callcenter – Analyse undManagement Modellierung und Optimierung mitWarteschlangensystemen, Springer, 2017.