Model for Airline Ticket Counter Staffing Using Sabre StaffPlan Staff Forecasting and Planning System
Client: Sabre Holdings: Sabre Airline Solutions
Team: Lauren Duplantis, Hamel Husain, Francisco Villagran
Faculty advisor: Dr. Siems Year: 2003
Documents: Final report (PDF)
Our project involved working with Sabre Inc. on a problem that they were having with their system StaffPlan that is maintained by their Airline Solutions Division. StaffPlan is an Airline Solutions system that minimizes the number of check-in agents at an airport based on the time of day, queue, airline, destinations, and passenger arrival curve categories.
The problem that our client wanted us to handle dealt with the queue, which is a single line of contacts who wait to be serviced. The queue is dependent on the service time, waiting time, quality target, and contact ratio. To simplify the problem they took out the contact ratio, so we just focused on the service time, waiting time, and quality target. The service time is the time it takes for a contact to be serviced from the time he/she gets to the counter to the time he/she leaves the counter. The waiting time is the amount of time that a contact waits starting from the time that they arrive to the time that they reach the ticket counter. The quality target is the percentage of total contacts whose wait times need to be under or equal to a maximum wait time. In this case Sabre wanted the maximum wait time to be 6 minutes and the quality target to be 80%. The problem with StaiThian is that it uses a fixed service time, so the program wasn’t as realistic as it could have been. In the real world, every contact takes a different amount of time at a counter. What our client wanted was for us to put variability in the service time to see how it affected the number of check-in agents needed, which we called ticket counters in our model.
Our method of analysis involved first putting together the C program and then comparing our output to StaffPlan’s output. Then once our output was identical to their output we put variability into the service time by creating a function that returned a random number that had a mean of 5 and was exponentially distributed. Once we got the output from the program we compared the output to that of StaffPlan’s to see what the variance was. Then we analyzed our findings and looked at other ways of making the model more realistic, such as putting in a maximum wait time for those that normally would have to wait longer than six minutes because they meet the quality target.