As a former math major, it was hard to skip over the article by this title in Forbes (Sept. 12,2011) which describes the multimillion dollar simulation software created by Schneider National to manage its massive trucking business. With 13,000 
drivers, 10,050 trucks on the road at any one time, and over 33,000 trailers waiting to be hauled, this 76-year old, $3.1 billion company faces many daily decisions regarding scheduling of drivers and equipment. For example, drivers are on the road between 4 days and 3 weeks at a time, and then must be back home by a certain date. The government regulates driver breaks and hours per day at the wheel (11 max). And customers are only open during certain hours.
Prior to hiring Princeton prof Warren Powell to build its fleet-wide “tactical planning simulator” (which is actually based on dynamic programming algorithms), Schneider relied on pilot projects to answer key logistical questions. A group of, say 20, drivers was carved out, experimented with, and results were drawn–at a cost of $100,000’s each time. But too often a system that worked well on the small sample did not scale well for the whole company. Sometimes the pilot even cost more than the money saved with the policy change.
The simulation model works like this: it runs forward in time 3 weeks to set the value of having a truck and driver at a certain location at a certain time. Then it runs backward in time to the “present”, reconciling the results with those that happened in the simulated future. It runs forward 3 weeks again and then backward, continuing to improve its estimate. When the schedule “converges” after hundreds of thousands of decisions, the process is complete. Schneider has saved tens of millions of dollars with the new system, which not only schedules, but helps determine price hikes, hiring patterns, and fleet size.
Discussion questions:
1. What other aspects of OM can Schneider use the simulator to analyze?
2. Why is simulation such a valuable tool in OM?