This thesis describes a throughput-based technique for identifying the bottleneck of a production system using discrete-event simulation. Most literature defines a bottleneck as being a workstation that has the largest impact on reducing the throughput of a system. However, when identifying a bottleneck using discrete-event simulation, throughput is rarely considered as the parameter of interest. Instead, parameters like percentage utilization and waiting time in an upstream buffer are considered. The technique suggested in this thesis identifies the bottleneck as being the workstation that causes the largest drop in throughput if added to a system. The technique is explained and tested on four different shoploor arrangements of workstations (serial, job split, conditional branching, and a feedback/rework production line). This demonstrates that the throughput-based technique can be used in most any shoploor arrangement of workstations and eliminates some of the drawbacks of the other more commonly used bottleneck identification methods, such as percentage utilization and waiting time in queue. A major failure of the percentage utilization technique in identifying system bottlenecks is seen in systems that have static and dynamic resources. However, the throughput-based technique correctly identifies the bottleneck of such systems.