Where Data Meets Optimization

In the year after graduating with her PhD, in what was financially a lean time, Margrét Bjarnadóttir and her husband decided to buy a new Macbook Pro to replace their aging one.

It was a big purchase, and Bjarnadóttir, an expert in data-driven decision-making who would later come to work at the University of Maryland’s Robert H. Smith School of Business, was determined to make the smartest decision possible.

“We collected data on eBay auctions on Macbook Pros, and we analyzed everything,” she says. “What were the key indicators for success? What are the characteristics of postings that were the most successful? For example, good pictures are very important. When was the best time to list a Macbook Pro – time of day, day of the week?”

They came up with a regression equation. "We had a lot of time," she says, starting to laugh. "I’m not saying the return of investment per hour was very good.” 

They eventually bid on a low-performing auction for a new laptop and used the best techniques to auction the old machine. And they broke even. 

Whatever the decision you are facing, Bjarnadóttir says, there are likely analytics that can help you find the best answer. “That’s what I do – data analytics," she says. "I live on the intersection where data meets optimization.”

Bjarnadóttir is an assistant professor of management science and statistics in the Smith School’s Decisions, Operations and Information Technology department. She applies innovative data modeling methods to specific domains – lately healthcare, complex financial networks and sports. 

A recent paper explores decision-making for cancer patients. “What the literature tells you is that doctors are biased,” she says. “When they are speaking to their patients, they tend to overestimate their survival. And sometimes they don’t even have that conversation.”

She and her co-authors built good predictive models to estimate survival and wrapped that into a tool to support those physician-to-patient conversations. With the app, patients could see how patients like them appear in the data and how successful various treatments were. “Basically, we try to unbias the conversation so that patients can make the decisions that are best for them, whatever that might be,” she says.

A newer study focuses on the opioid crisis, building an early risk model that aims to help physicians determine, based on data, how likely a patient is to become dependent on opioids if they receive a prescription. “It’s important work,” she says. “Once a person becomes a chronic opioid user, the risk of all of the bad outcomes is magnified.”

She is meanwhile also studying the gender pay gap, and potential algorithms that companies might use to close it optimally, cheaply and equitably. It’s been fascinating research, she says, highly relevant in the #MeToo era and more complicated than it might seem at first glance.

“If you are tasked with eliminating the pay gap at your large organization, how would you go about it?” she asks. “If you had money to eliminate the pay gap, how would you hand out raises? No one had really addressed that question before we took on this research project and developed algorithms that eliminate the gender paygap balancing equity and efficiency.”

In sports, she is studying how Major League Baseball teams might identify good buys in the free agent market. And she is studying the progression of young competitive swimmers, developing a model that university swim recruiters might use to identify strong high school prospects. 

“There are always externalities not captured by our models that you have to take into account, in any decision scenario” she says. “But we can build models that can help.”

Bjarnadóttir earned her undergraduate degree in her native Iceland at the University of Iceland, where she initially pursued a major in civil engineering. It didn’t stick.

Early on in her studies, in a course on the chemistry of civil engineering, she encountered a 20-page chapter on ingredients that could be stirred into concrete. “And then I realized I couldn’t do civil engineering, because those were the most boring hours of my life, the hours spent reading a chapter on how to make concrete better,” she says.

She pivoted to industrial and mechanical engineering and took a course on operations research, and that’s when she had her eureka moment. “I fell in love with operations research,” she says. She later received a PhD in operations research from the Massachusetts Institute of Technology. “The math is beautiful, but it’s also very practical. Operations research is really problem driven. It’s all about: How can we use math and models and data to improve decision-making?” 

Math always came easily to Bjarnadóttir. When she was just 8 or 9 years old, she recalls, she went on a business trip with her dad, an electrical engineer for Iceland’s national power company. To keep her learning and amused, she says, he written out the multiplication table for her to study and practice.

Her brothers pursued STEM careers as well. One is the head of software development for MAREL, an Icelandic food production line company; the other works in cybersecurity in Finland.

Her husband is also from Iceland. They met while studying at MIT and were surprised, as Icelanders naturally would be, to learn find that of their many Facebook friends and family members back home, there were no common contacts.

“Iceland is very small," she says. "It has just 340,000 people, in total. Everybody knows pretty much everybody.” 

It’s the reason why author Michael Lewis admonishes readers of his 2011 book “Boomerang” that they’re just being annoying when they ask Icelanders whether they know the singer Bjork. “Of course they've met Bjork,” he writes. “Who hasn't met Bjork? Who, for that matter, didn't know Bjork when she was two?"

Bjarnadóttir doesn’t know Bjork. “But I’ve been on a treadmill next to her,” she says.

View More Profiles