The solution
During COVID-19, Columbia Business School wanted some students physically present while classmates joined online. A naive course-by-course allocation could strand a student online for one class in the middle of an otherwise in-person day.
The school modeled hybrid attendance as a mixed-integer scheduling problem. One formulation minimized those excess online cases for individualized enrollments; another spread team attendance across the term while maximizing cross-team interaction.
The system scheduled more than 2,500 students in student-level scheduling and about 790 students in team-level scheduling from fall 2020 through summer 2021.
What it achieved
Schedule attendance by day, not by course
Why it worked
Student experience depended on the whole day, not isolated classroom capacities.
A MIP could encode distancing, enrollments, team structures and balance objectives together.
Separate formulations matched the two real operating modes instead of overgeneralizing one model.
The solution was deployed across multiple terms under changing pandemic constraints.
What can be applied
When capacity changes by room and day, fairness improves if the unit of planning is the person's whole day, not a single class section.
Aftermath
The mixed-integer scheduling system ran through the fall 2020, spring 2021 and summer 2021 terms. It gave the school a repeatable way to allocate scarce in-person capacity while keeping online exceptions explicit.
FOLLOW THE EVIDENCE