Customer wait time prior to being served is a key performance measure used at the Virginia DMV.
Customer service center wait times are impacted by a wide range of factors including hourly customer volume, staffing and scheduling, transaction type, service time, and geographical location. As such, consideration of all of these factors was essential in creating accurate staffing recommendations.
SIMUL8’s simulation software provided DMV with a powerful tool capable of mapping this customer flow within each customer service center. This enabled analysis of all variables simultaneously and significantly improved the quality of analysis.
A simulation model was developed in SIMUL8 utilizing existing queuing system data from the most recent fiscal year to calculate service type probability, average service time by service type, and average hourly customer arrivals. Resources were defined using aggregated serve time from the current queuing system and then adjusted for shrinkage to determine hourly staffing levels.
DMV conducted an initial trial run with hourly customer arrivals being adjusted until the model’s wait time for each customer service center fell into an acceptable range.
Once normalized, the wait time benchmark was set and the full model was run, using an algorithm designed to identify the lowest number of staff required to accomplish an average wait time at each hour interval.
For each trial, 100 runs were calculated to produce hourly average wait times. The scenario was evaluated chronologically, and at the earliest instance where the average wait time exceeded the benchmark, the staff resource was increased by one.
Further trials were automatically conducted until staffing levels accomplished the wait time benchmark for a full day of operation. Upon completion, recommendations were evaluated against existing limitations such as maximum staff available and service window availability.
For example, if the model returned a staffing recommendation of 25 in the third hour, but only 12 service windows are available at the site, the staffing numbers were manually reduced to reflect that limitation.
Another trial was completed after manual adjustments were made. If the average wait time met the 20 minute goal, recommended staffing levels were compared to the actual staffing levels, and the variance was reported as the staffing adjustment necessary for each hourly interval to achieve the benchmark.