An Interactive Tool Helps School Districts Redesign Their Bus Schedules—and Get Kids a Little More Sleep
Parents and teachers know that teenagers struggle to learn first thing in the morning, but the complexities of bus routes mean that most school districts still start class earlier than recommended. Yale SOM operations scholar Zhen Lian and her co-authors created an interactive tool that helped San Francisco reach consensus on school schedules, move start times later, and save millions of dollars in transportation costs.
The research is clear: the circadian rhythms of adolescents make it hard for them to wake up early. The American Academy of Pediatrics and the Centers for Disease Control recommend that middle and high schools start no earlier than 8:30 a.m. so students can get enough sleep, making it more likely they attend class, excel in their studies, and graduate. But most school districts ignore the science, building schedules around bus routes and transportation budgets and asking millions of teenagers to start learning before their bodies are ready.
The few school districts that have tried to make a change have found it challenging. When a school board proposes new start times, parent and teacher emotions often run high. In 2017, for example, Boston tried to make school start times later for high schoolers and earlier for many elementary students. But parents pushed back, and the district backed down.
In 2019, California became the first state to mandate a later school day, setting middle school start times at 8:00 a.m. or later and high school start times at 8:30 a.m. or later. Faced with the mandate, the San Francisco Unified School District (SFUSD) looked for help designing school schedules that complied with the law while helping to reduce its budget deficit—and that would be palatable to students, parents, and teachers. The district contacted Zhen Lian, then a PhD student and now assistant professor of operations management at Yale SOM, and two colleagues who had helped Boston revise its schedule: Arthur Delarue, now at the Georgia Institute of Technology, and Sébastien Martin of Northwestern University.
“Deciding on school schedules is a pretty complicated logistical problem,” says Lian. “The question is, how can you design school schedules in a way in which everyone feels that they’re engaged in the process?” Their solution, which is described in a newly published study, helped SFUSD implement the first ever interactive optimization-driven school start time change in a major U.S. school district.
Hey, I’m AInsights
Ask me questions about the article and its underlying research. Let’s chat!
When SFUSD, a district of over 50,000 students and more than 120 schools, contacted the researchers in fall 2020, its school schedules, which had been mostly standardized in the 1980s, had splintered, with 18 different start times requiring 159 buses at an annual cost of $200,000 each. The software the district had been using to schedule buses was slow and offered a limited number of possible solutions.
The researchers began the project with the experience of the Boston school district in mind. “Right from the beginning, we were very careful to include everybody in the loop,” says Lian, whose research focuses on how to use new technologies to bring stakeholders closer to large-scale, complicated problems.
Lian and her colleagues didn’t want an opaque, black-box algorithm driving the process, which could lead to suspicion and distrust. Instead, they set out to create a more dynamic tool that administrators could use on their own, hoping it would pave the way for a smoother design and implementation.
The researchers built an algorithm that calculated 1,000 near-optimal school schedules based on later middle school and high school start times. Then they created a user-friendly interface in a spreadsheet and fed in the algorithm’s solutions. Transportation staff, principals, and teacher groups were able to test-drive options using the interactive spreadsheet: they could pick different start times for individual schools and see how choices impacted times across the district and bus budgeting. A dynamic map illustrated start times too.
“We had levers people could play with,” Lian says. “It shows them the ripple effect that changing times for one school can create to the entire system. We were really trying to put them in the driver’s seat, and because the process was so transparent, and they had so much control, they were able to do so much by themselves.”
The tools really helped them with the underlying structure of this very complicated problem. It makes it easier to reach consensus. Everything just becomes more concrete.
The interactive tool helped the district make tough decisions, such as whether or not to sacrifice bus savings for an individual school’s preferred start time.
Built into the algorithm was the district’s willingness to give up about 7% of potential transportation savings in exchange for more scheduling options to evaluate. Late in the process, they helped the district factor in the additional variables, including early dismissal days and the length of the school day.
“We got feedback from the district that the tools really helped them with the underlying structure of this very complicated problem,” Lian says. “It makes it easier to reach consensus. Everything just becomes more concrete.”
The tools also revealed a pleasant surprise: coordinating elementary early-release days so everyone could take part in professional development—a change teachers had been requesting for years—would cost far less than the district assumed.
In August 2021, SFUSD implemented new start times, with elementary schools starting at 7:50 a.m. and middle schools and high schools beginning at either 8:40 a.m. or 9:30 a.m. The change saved the district $5 million annually in transportation costs, which it put toward classroom development. In 2022, the researchers surveyed over 27,000 district families and staff members from 70 schools to see how they felt about the change.
“They were more positive than expected,” Lian says. The survey showed that a slight majority of parents were at least somewhat satisfied, and 37% very satisfied. Among staff members, a plurality of 32% reported being “very satisfied.” Still, about 23% of respondents were “very unsatisfied”—some of them likely parents with more than one child in different schools who received their preferred start time for one child but not the others.
Lian and her co-authors have since presented their work for school districts at conferences across the country. Multiple districts are currently considering a change, including Fairfax County Public Schools in Virginia, where officials are developing a pilot plan to bump up start times at select middle schools, and Florida’s Duval County Public Schools, which by the start of the 2026-27 school year must comply with a state bill requiring 8:00 a.m. or later middle school start times and start times of 8:30 a.m. or later for high school.
The researchers are making their tools available to the public, but districts would need to create interactive spreadsheets tailored to their own needs. If more districts express interest, the team may formalize the process.
The research is a starting point for future projects tackling collaboration between policymakers having to make difficult decisions and researchers, Lian adds. “We’re thinking about other collaboration angles,” she says. “Tools like these can help policymakers be clearer about their priorities and at the same time learn about the process of reaching consensus and getting a good outcome to actually be implemented.”