Assortment optimization: meet new ISE faculty James Davis
February 2, 2016
The front page of an online retail website or news website displays only a few items or news articles. Yet these aren’t the only items or articles on the entire website. How do companies decide what few items to display?
This specific problem has to do with assortment optimization, which requires implementing algorithms, or sets of operations, to make optimal decisions. It’s this type of problem that Assistant Professor James Davis enjoys solving by applying techniques from revenue management and algorithm design.
“I like to actually solve problems with algorithms that are not just efficient, but can also be implemented by someone at a company and can actually make an impact. That’s my goal,” Davis said.
Before joining the ISE faculty in the fall of 2015, Davis earned his bachelor’s in mathematics from Rutgers University-Camden and his Ph.D. in operations research from Cornell University. His initial interest was in approximation algorithms — those that attain approximate solutions to problems — which eventually shifted to an interest in revenue management.
“I can apply those techniques for approximation algorithms to revenue management problems, and get good, interesting results that people can actually implement in revenue management systems,” Davis said.
Previously, Davis worked for Amazon, using his thesis research to apply assortment optimization techniques in order to make inventory decisions for the company.
In the theoretical computer science community, algorithms are made frequently but are implemented infrequently, Davis explained. What he aims to do, then, is make algorithms that can be used in the real-world industry. He strives to make algorithms that are both applicable and fundamental in their purpose.
“When I design an algorithm, it’s maybe one step away from actually being implemented in a system,” he said. “Then a year later, someone will take over this algorithm and put it into practice.”
Davis said it felt natural to join the Department of ISE, as he is “an industrial engineer by heart.” Additionally, he enjoys the interdisciplinary atmosphere at the University and the opportunities he has to talk to researchers in other disciplines.
Currently, Davis is working on a project related to search engines. Many search engines today return the most relevant results for users. But sometimes, he said, the most relevant results might not actually be what the user wants.
“If I type in ‘couch,’ I don’t want to see 50 identical couches that are just relevant,” Davis said. “Maybe I want to explore the space of furniture . . . maybe I can also see some loveseats or something else.”
To get to this point, Davis said companies should instead have an explicit objective search function to help a company find the goal of a search result. Finding this goal would require a company to explore what a user is actually asking for when they enter a specific search term.
But there’s also a human component to designing algorithms, and Davis said this component should always be taken into account — whether it’s using people’s decisions to inform prices, showing items on a website that will be most attractive to a customer, or planning bus routes that people will actually use.
Along with this is the notion of the “customer-centric” focus. Davis predicts that retail and other industries will continue to be increasingly customer-centric.
“It’s going to get to the point that individual people won’t be able to make the best decisions and will need to turn to automated systems,” he said. “Those automated systems are the things I’d like to bring into existence.”
Other problems he’d like to address include figuring out what to show on a website in a data-driven way and making advances in pricing techniques.
When he’s not working on research or teaching his integer programming course, Davis has a variety of side projects. One of these projects is a computer numeric controlled milling machine — a robot that can perform a variety of milling operations — that he built himself and now keeps in his garage.
“I just read a bunch of stuff, got super interested in electronics for a year or so, and made it,” Davis said. “I’m continually working on it. Most of it is duct taped together, but I’m getting rid of the duct tape and replacing it with real parts.”
Davis said his biggest goal for his career is making sure his research is hand-in-hand with the problems that are developing in industry, and that his algorithms can be implemented to solve these problems. Seeing this implementation, he said, is the most exciting part of his work.
“That’s something that’s really important to me,” he said. “Consistently staying abreast of changes in the world is important.”