Manual queue management at Bank Mandiri Taspen Medan Branch Office gives no automatic priority to the elderly, persons with disabilities, and pregnant women and may lengthen waiting times when visits increase. This study designs and implements a web-based customer queue system using a four-level Multilevel Queue (MLQ) that combines fixed priority between levels, First Come First Served (FCFS) within a level, and an anti-starvation rule with parameter K. The contribution is an integration of application and queue policy, not a new MLQ algorithm. The Waterfall model was used for development. The evaluation is entirely simulation-based rather than field measurement: a constructed deterministic scenario of 14 customers, replicated stochastic simulation (2,000 replications), and sensitivity analysis of arrival intensity, service-time variability, counter configuration, and threshold K. In the 14-customer scenario, the average waiting time of the priority group fell from 24.50 to 11.00 minutes, while the regular group rose from 37.13 to 50.50 minutes. In the stochastic two-counter simulation (λ = 0.20 customers/minute), the 95th percentile of priority waiting time fell from 25.71 to 17.47 minutes with K = 3, while the regular group’s 95th percentile rose from 22.99 to 32.83 minutes. Thresholds K = 2–4 met the proposed fairness criterion at this load, but no K met it at λ = 0.25. These results describe the behavior of the simulation model and require validation with field data
Copyrights © 2026