We investigate the use of priority mechanisms when assigning service engineers to customers as a tool for service differentiation. To this end, we analyze a non-preemptive M/PH/c priority queue with various customer classes. For this queue, we present various accurate and fast methods to estimate the first two moments of the waiting time per class given that all servers are occupied. These waiting time moments allow us to approximate the overall waiting time distribution per class. We subsequently apply these methods to real-life data in a case study.