In 1990, the Darden Graduate School of Business Administration of the
University of Virginia undertook a major revision of its MBA curriculu
m and instituted a new system for course selection. The administration
determined that one component of education quality was allowing stude
nts to enroll in their preferred courses. Under the new system, no cla
ss times are assigned until all students have made their requests for
classes. A schedule needed to be constructed that simultaneously met c
lass-size and facility constraints, satisfied faculty time preferences
, and maximized the satisfaction of student course requests. The solut
ion technique needed to be quick and flexible, allowing the registrar
to test the effects of various scheduling constraints. Integer program
ming was found to be impractical. We developed a local-search heuristi
c procedure to effectively and efficiently solve this scheduling probl
em.