A simulation-based staff scheduling analysis for call centersin cargo industry

Abstract

Call center staff scheduling problems have been studied frequently on a regular basis for many years. Call centers generally conduct their business on customers by delivering service. Problems occur especially when the density of queues is high, causing long waiting times. In terms of customer satisfaction, this outcome is undesirable. One of the reasons that causes this outcome is the inefficient personnel/job scheduling. In call centers, most of the time, the demand of service is inhomogeneous. Therefore, the incoming number of calls varies depending on the time interval of the day, in which the firm operates. The aim of this thesis is to understand appropriately how a call center operates and generate effective and efficient solutions by employing simulation. The raw data regarding the call center operations were first acquired which contained several important aspects that are beneficial to capture the behavior of the system and propose a better alternative to it. Gathered raw data had incoming call dates and several other statistical records that were kept. Firstly, this data was analyzed using MATLAB. Subsequent calculations on Matlab made it possible to put the resulting data that illustrate different aspects of the queue to the ARENA simulation program as inputs. To achieve effective and efficient scheduling, the incoming calls were categorized based on different time intervals by taking the density of calls into account across a day for 27 days. Afterward, input analysis was conducted on the ARENA Input Analyzer tool to get the best-fit distributions that were necessary in order to create a successful simulation analysis. After processes were clearly defined and mapped on ARENA, several simulation parameters were calculated and considered before conducting the analysis, one of which stands for huge importance in terms of reliability and accuracy of the outcomes, which was determining the appropriate number of replications. The results showed that there was insufficient number of employees working at that time. Additionally, employee scheduling was not conducted in an efficient way. These were the main reasons why bottlenecks occurred in the first place. In order to reinforce the solutions, a scenario analysis was conducted. Additional three scenarios were included in the simulation program and all of the scenarios were compared against each other in terms of KPIs. Finally, results showed that the simulation conducted appeared to be appropriate, and represented the overall system successfully. The scenario analysis illustrated thatrelaxing the bottleneck could be achieved when effective and efficient job scheduling is conducted. Moreover, the outcomes also indicated that customer satisfaction could be achieved by improving the existing system according to the simulation results presented. In the first section, an entry to the call center staff scheduling problem has been made. In section 2, a literature study was conducted in order to see the span and spread of techniques employed to solve the problem. A comprehensive study was conducted to see the blindspots in literature. These studies were heavily analyzed, and the outcomes were clearly illustrated. It was deducted from these studies that the use of simulation would be appropriate, specifically, discrete-event simulation. In Section 3, the methodology is proposed. After determining which technique to employ in order to solve the problem, related academic books were analyzed to come up with a fine approach to the problem. In Section 4, after creating the framework to follow, the simulation application was conducted. Firstly, the raw data was pre-processed and analyzed in Matlab, the descriptive statistics were presented, and the results were given to the ARENA simulation program to conduct input analysis. Input analysis included conducting data analysis, and determining best-fit distributions. Data analysis and finding best-fit distributions are crucially important because the ultimate aim when conducting a simulation analysis is to create an artificial system that perfectly reflects the original one. Because the raw data acquired showed seasonality and, therefore needed to be more carefully analyzed. Best-fit distributions were found for the five main data categories which were call durations, interarrival times, waiting time for missed calls, assignment time to the queue, and connection waiting time from the queue to the operator. The application chapter also included creating a process flow chart. After the system is conceptualized and input analysis is conducted, the process flow chart, which also reflects the actual simulation model that needs to be built, was drawn. Subsequently, the ARENA model was created. In the ARENA model, along with several other aspects, one main parameter was to define the minimum necessary number of replications. In order to do that, some statistical calculations and comparisons were made, and the minimum necessary number of replications was found. Finally, output and scenario analysis were conducted to better capture the the effects of changes made in order to improve the overall system. Important KPIs were collected and run for four scenarios. It seemed that waiting time in queues was caused by the insufficient number of employees working without an appropriate schedule. Additionally, this caused the problem of long waiting times in queues, in other words, bottlenecks, which also decreased the customer satisfaction level. Scenario analysis pointed out the existing system and proposed better alternatives in the subsequent scenarios in terms of KPIs. According to the results of the scenario analysis, the firm needs to hire more personnel and schedule those effectively and efficiently, as illustrated in this thesis to improve the overall call center operations. All the necessary figures and tables regarding the simulation application were given in Section 4. Additionally, in Section 5, the results of this thesis were presented and suggestions for future studies were given.