Engineers Optimize Marathon Volunteers

As runners tackle the Bank of America Chicago Marathon on Sunday, they will have an unseen support system to help keep them – and the race – running smoothly.

Northwestern Engineering’s Karen Smilowitz, her student team, and the Feinberg School of Medicine’s George Chiampas are using data analytics to help marathon organizers ensure that medical tents are well-staffed throughout the 26.2-mile course.

Four years ago, Smilowitz and her students began developing a custom-designed data visualization system that provides a computer simulation of the race as it unfolds. Using data from the past nine Bank of America Chicago Marathons, the system can forecast where large concentrations of participants will be along the course and help race officials plan accordingly. Staff members stay connected to the system via mobile dashboards to ensure that the race is seamless and safe.

To update the system to help optimize medical personnel, Smilowitz and her team first examined historical data and calculated the demand at each tent location over time. More recently, to determine whether medical tents are well-staffed, they turned to another data source: “stress levels”. One volunteer per medical station gauges the general stress level of the tent’s atmosphere on a scale of 1 to 5 and reports it to a radio team. (One being the most relaxed, and five being the most overwhelmed.) Every time the level of stress changes, the volunteer updates the report. Smilowitz’s students have access to the report data and input it into the dashboard system. By monitoring this input, Chiampas, the marathon’s medical director, can determine areas with staff shortages and make informed decisions about redeploying teams to keep up with demand.

Smilowitz, Charles Deering McCormick Professor of Teaching Excellence and professor of industrial engineering and management sciences, is an expert in the study and application of data analytics for mass-scale events. For this project, she teamed up with Chiampas, an emergency medicine professor at Northwestern’s Feinberg School of Medicine. The pair recently received a grant from the National Science Foundation to fund the commercialization of their system.

“The marathon is always a big challenge,” Chiampas said. “But Chicago has set the bar globally with regards to large event planning, and Karen’s work is a part of that.”

As medical director, Chiampas has witnessed first-hand how difficult it can be to effectively distribute 1,200 medical personnel. Volunteers often do not make it to their pre-assigned tent, sometimes getting confused about locations or drifting instead to tents where their friends are working.

“With any event that relies on volunteers, there is a natural attrition rate,” Chiampas said. “Now we can be more confident in knowing if they make it to the right place.”

Accounting for adjustments

Marking the Chicago Marathon’s 40th anniversary, this year’s race is expected to have one of its highest attendance levels. A total of 50,000 runners have registered for the event — 10,000 more than ran last year. But just because 50,000 runners registered does not mean that number will finish — or even start — the race.

“There is the number of how many people sign up, how many people pick up their packets, how many people show up on race day, and how many people finish,” Smilowitz said. “There is a known drop-off between registration and packet pick up and another known drop-off between packet pick up and the start line.”

In the past, Smilowitz’s team has experienced difficulties in accounting for drop-outs. As runners exit the race, Smilowitz and her students must quickly adjust the numbers and alter the simulation.

“This is particularly relevant this year because of predictions about the heat,” she said. “That could make the drop-off between registration and packet pick up even bigger.”

This year, Smilowitz and her team updated their system to better model drop-outs. Kate Larsen, a student in the Master of Science in Analytics program, has focused on analyzing the drop-outs that occur between the starting line and the finish line.

“Our goal to better predict the frequency and volume of drops is two-fold,” Larsen said. “Knowing where the drop-outs occur can help allocate resources to prevent drops, and it can help us know how many runners are actually on the course at any given time.”

Racing improvements

Smilowitz credits her diverse student team members, who represent computer science, industrial engineering, and analytics, with improving the system each year. They have worked with Chicago Marathon researchers to give the system more capabilities, including accounting for drop-outs, adding more seamless automation, and better tuning the simulation to reflect weather conditions.

“This summer, we completely overhauled the behind-the-scenes work of how the data are stored, accessed and manipulated in order to increase security, streamline the movement of data, and prepare for large expansions,” said Charlie Collar, a junior double-major in computer science and music theory. “Since we redesigned the system, we were able to learn from past mistakes and create a very stable environment.”

To prepare for race day, Smilowitz and her students ran mock simulations to pinpoint issues and work out the kinks. This Sunday morning, they will be stationed in the marathon’s forward command center to monitor the incoming health, situational, and performance data on two dashboards. The data they collect this year will go into perfecting their system.

“Every year, our product is just better,” Smilowitz said. “In the past, we mostly focused on developing and prototyping the system. This year, the students really cleaned it up and fine-tuned it. They have done a phenomenal job.”