Boston cut 50 school buses, saving $5M a year, by solving routes and start times together

Maintaining a bus fleet is a major cost for US school districts, and many spread start times to reuse buses between schools. But that reassignment is done by hand, and the resulting early high-school starts harm student health.

MIT researchers, with Boston Public Schools, built the first algorithm to optimize bus routing and bell time assignment jointly. It decomposes the routing problem, solves subproblems with mixed-integer optimization, and uses that to price out what each start time would cost.

The method cut a process that took 10 people more than 3,000 hours to one that runs in about 30 minutes, while making the start-time tradeoffs — cost, equity, health — explicit choices for the district.

Choose each school's bell time from what routing will cost

It solved one joint problem instead of guessing at two separately, so buses and times reinforced each other.

A transport-cost proxy made the start-time decision a solvable optimization rather than a hunch.

It made student-health considerations an explicit objective, not an afterthought.

Running in 30 minutes let the district explore options instead of committing to one.

Two decisions people take in isolation — school start time and bus routing — are one cost problem; solve them together and the fleet shrinks, the budget frees up, and students get later mornings.

The algorithm led to $5 million in yearly savings and maintained service quality despite a 50-bus fleet reduction. In December 2017 the Boston School Committee unanimously approved the first school start-time reform in thirty years, moving to later middle- and high-school starts, and the work became a 2019 INFORMS Edelman finalist.