Our routes and stops aren't perfectly efficient and our bell times are out of balance. We're hosting a challenge to use computer science to improve our routes and equitably balance our bell times - with the goal of freeing up funds to reinvest in schools and to improve the student experience.

In FY16, transportation costs accounted for $110 million or 11% of the district’s budget. On a per pupil basis, BPS’s transportation cost is the second highest and more than five times the average of the largest 200 public school districts. Meanwhile, transportation costs have continued to increase, up $33M from FY11, a 7.5% annual increase.

We are hoping that we can reach out to the Boston community -- and beyond -- for your help addressing the inefficiencies and rising costs in our transportation system. Equitably balancing bell times will impact the lives of nearly 100,000 students in Boston each year. And optimizing bus routes will allow us to increase efficiency while reinvesting in classroom.

Below you will find all the information and resources needed to tackle this problem and receive an award!

Photo "School Buses" by Dean Hochman is licensed under CC BY 2.0

Questions? Ask them here!

The Challenges

Our routing system has evolved since we began busing, when we used string and pins to represent routes and stops. However, we haven’t begun to fully incorporate the tremendous potential of advances in transportation and mapping technology. We believe that if we tap into this technology, we will unlock new algorithms and approaches for automating routing and for doing so efficiently and accurately, with just a few clicks of a mouse. We hope to use this challenge to generate a new system of bus routing that can be used by the spring, to ensure our routes will be ready for the 2017-18 school year.

Improving our routing involves two steps:

Step 1: Optimize Bus Stop Locations

We are working to consolidate bus stops to increase efficiency and decrease student ride time, while also ensuring that students are able to walk safely and quickly to a nearby bus stop. Based on where students live, their destination schools, and the type of service required (i.e. Corner Stop, Door to Door, Wheelchair) how will you assign students to a stop?

Step 2: Improve Routing Efficiencies

We are working to get all students to school as quickly as possible in time for the start of the day. Given the location of bus stops, how would you connect these stops to schools over the course of a day creating a full route for a bus that starts and ends at a given bus yard? How will you account for different types of services required (Corner Stop, Door to Door, Wheelchair)?

Broadly speaking, BPS has three school start times: 7:30, 8:30, and 9:30. These bell times, however, are “imbalanced,” with more buses serving schools at 8:30 than at 7:30 or 9:30 (there are far fewer at 9:30 than at either of the earlier times). In order to maximize resources, we would like for all of our buses to serve three schools each morning at each of the three times. Currently many of our buses perform only one or two trips per morning, causing additional and unnecessary costs.

Complicating matters more, we know anecdotally that many schools do not prefer a 9:30 start time which means we cannot address this imbalance by simply putting more schools at the last of the three bell tiers. So in order to effectively balance our bell times, we need to create an algorithm that also factors in school start time preferences. We need to solve this problem and increase the efficiency of our bell times for the 2018-19 school year.

Step 3: Balance Bell Times

Now that you’ve figured out how to calculate the most efficient routes, what happens when the rules change? That is, how can you assign a start time to a school and minimize the number of buses that we need (as well as the total system wide drive time)? How will you take into account new data that we may give you regarding school preference?

These two challenges -- maximizing both the efficiency of our routing and our bell time structure -- are interrelated but will be addressed on different timelines.

Any participant can submit a response in just one of the challenges or in both. Each contest will have two winning teams -- but the same team could win both contests. And since solving the bell times puzzle will require a routing solution, we imagine that many participants will want to take part in both.

Event	Routing Challenge	Bell Times Challenge
Sample Data and Competition Overview Posted on Website	Week of March 6, 2017
RFI and Background Documents Released	April 1st, 2017
Public Kickoff Event	April 1st, 2017
Requests Can Be Submitted for Actual Student Data	After April 1st, 2017
Question Deadline	April 10, 2017
Consolidated Q&A Posted by BPS	April 14, 2017
Submission Deadline for Round 1	4/30/17	6/18/17
Round 2 Participants Announced	5/7/17	6/21/17
Submission Deadline for Round 2	5/10/17	6/27/17
Interviews with BPS	5/12/17	6/30/17
Public Presentations and Next Steps' Announced	5/16/17	7/10/17

Times and dates subject to change

Note: All deadlines -- unless otherwise stated -- are at 11:59p.m. EST on the date given. All of these dates are subject to change.

More Information

Superintendent Tommy Chang

Superintendent of Boston Public Schools

John Hanlon

Chief of Operations for Boston Public Schools

Regina Robinson

Dean of Student Affairs at Cambridge College and Boston School Committee Member

Andy Rotherham

Co-founder and partner at Bellwether Education Partners

Dimitris Bertsimas

Professor of Management, a Professor of Operations Research, the Co-Director of the Operations Research Center and the Director of the Master of Business Analytics at MIT Sloan

Jascha Franklin-Hodge

City of Boston, Chief Information Officer

Mike Hughes

Assistant Director of Transportation

Steve Poftak

Executive Director of the Rappaport Institute for Greater Boston

Velecia Saunders

Headmaster of the McKinley Schools in Boston, Massachusetts

As part of this process, we are posting a number of documents to help you solve the problem.

Background Information

The Challenge Overview: This is your one-stop shop for everything you need to know about the challenge - including an overview, evaluation criteria, timelines, and more!

Kickoff Presentation: This is a copy of the presentation shared at our kick-off

Non Disclosure Agreement (NDA): This NDA must be submitted in order to advance through the Challenge. This NDA, accompanied by a one-page letter outlining your qualifications (read more in the Challenge Overview!) will enable access to a more realistic routing dataset

Principal Transportation Handbook: A helpful resource created by our transportation team providing more context around our transportation policies

Rules of the Road: A punnily named resource that summarizes the constraints and guidelines that need to be considered as part of your solution

Sample Output: An example of a file to submit with your solution that will help you, and us, check your solution for compliance

List of Answers to Submitted Questions: A constantly updated list of FAQs surrounding the challenge

Bell Time Preference Data Overview: A constantly updated list of FAQs surrounding the Round 2 Challenge on changing school start times

Raw Data

Simulated "Fake" Student Address Dataset*: This file was created with the support of students and faculty from BU's Hariri. This file contains data similar to the data-set we will share with those who choose to submit an NDA - a neccessary step to move into Round 2 of the Challenge - but contains simulated data very similar to our overall student assignment patterns

One Page Overview of Simulated "Fake" Dataset*: This file contains instructions on the information included in this dataset

*A note on this dataset: this information is based on fake students - but real addresses in Boston, randomly pulled from the property tax records - assigned to fake schools. We did this to protect student confidentiality.

In order to win either competition, you must have requested and received the real dataset. This dataset will be given out after the Kickoff Event (on April 1st, 2017) on a rolling basis to participants who submit the following three documents:

A brief (less than a page) note summarizing why you are interested in solving this problem, what qualifications you have, and how you would think about solving this problem based on the sample data

A signed Non-Disclosure Agreement (see Appendix 2 in the RFI) signed by everyone on your team who will be using the confidential BPS data

An agreement outlining the terms and conditions of the competition (see Appendix 3) signed by everyone on your team

In the Challenge Overview (above) we outline the specifics of what we are looking for in your response to each challenge. Broadly speaking, we are looking for a few things:

A high level description of your optimal routing or bell time solution

This should include key performance metrics (i.e., the number of required buses, routes, stops, and driver hours)

An excel file pairing each student to a stop, route, and bus - this allows us to check your solution at the student level

An interface -- of your choosing -- that allows us to visualize your solution. This can be as high tech (secure web portal) or low tech (printed directions) as you'd like

In order to win either competition, you must have requested and received the real dataset. This dataset will be given out after the Kickoff Event (on April 1st, 2017) on a rolling basis to participants who submit the following three documents:

A brief (less than a page) note summarizing why you are interested in solving this problem, what qualifications you have, and how you would think about solving this problem based on the sample data

A signed Non-Disclosure Agreement (see the NDA file in the "Resources" section above, or see "Appendix 2" in the "Challenge Overview"). Everyone on your team who will be using the confidential BPS data must sign this NDA

In the Challenge Overview we outline the specifics of what we are looking for in your response to each challenge. Broadly speaking, we are looking for a few things:

A high level description of your optimal routing or bell time solution

This should include key performance metrics (i.e., the number of required buses, routes, stops, and driver hours)

An excel file pairing each student to a stop, route, and bus - this allows us to check your solution at the student level

An interface -- of your choosing -- that allows us to visualize your solution. This can be as high tech (secure web portal) or low tech (printed directions) as you'd like

The Challenges

Step 1: Optimize Bus Stop Locations

Step 2: Improve Routing Efficiencies

Step 3: Balance Bell Times

More Information

Superintendent Tommy Chang

John Hanlon

Regina Robinson

Andy Rotherham

Dimitris Bertsimas

Jascha Franklin-Hodge

Mike Hughes

Steve Poftak

Velecia Saunders

Background Information

Raw Data