Despite the fact that most people and organizations rely on digital apps for a majority of their banking needs, physical bank branches still face an “analog” question: How much cash should they keep on hand to meet their customers’ daily needs? The situation was made even more complicated by COVID-19, which disrupted usual spending patterns in unanticipated ways.
The inaugural Crossroads Classic Analytics Challenge invited business students from four Indiana universities to solve the problem during its online competition held Jan. 4-18. The challenge’s aim was to provide students with an opportunity to sharpen their analytics skills with a real-world application.
Crossroads Classic was sponsored by the University of Notre Dame's Mendoza College of Business along with Butler University, Indiana University, Purdue University and Teachers Credit Union (TCU), which provided the challenge for students to solve. The competition was hosted virtually on a Kaggle platform, an online community of data scientists and machine learning practitioners.
The Crossroads winners, announced Feb. 5, included:
Graduate Division:
- 1st prize ($7,000): Notre Dame (Bernard Akatu, Claire Kozak, Maureen Malloy, Angela Stitsworth).
- 2nd prize ($3,500): Purdue (Cindy Hsu, Xuan Mai Nguyen, Robyn Campbell, Stefanie Walsh).
Undergraduate Division:
- 1st prize ($3,000): Purdue (Jiacen Liu, Hui Zeng).
- 2nd prize ($1,500): Butler (Victor Aguilar, Tess McTeague, Drew Smith, Sam Williams).
"Competitions like the Crossroads Classic Analytics Challenge are valuable for students for three reasons,” said Scott Nestler, academic director of the residential Notre Dame Master of Science in Business Analytics program, who coordinated the challenge. “The students in analytics programs have the opportunity to apply knowledge and skills developed in multiple classes to a real-world problem; they must work through the analytics process from start to finish; and they get experience presenting their work to business leaders."
The case presented by TCU, which has 57 locations and more than 300,000 members across Indiana and Michigan, asked students to help better predict its branches' weekly cash-on-hand needs, a problem that has become more difficult to forecast in light of how the coronavirus pandemic has shifted member behavior. The calculation had to take into account a branch’s cash limit and also avoid underestimating cash needs, which would result in an expensive “special order” between normal cash deliveries.
“At TCU, we’re focused on serving our members as efficiently as possible, which includes ensuring we have the right amount of cash on hand at each of our branches at any given time,” said TCU Chief Information Officer Dan Rousseve. “The Crossroads Classic Analytics Challenge gave students an opportunity to use an analytics-backed approach to solve a real-world problem. We were very impressed with the quality of their work and the operational efficiencies that were recommended.”
Twenty-nine teams from the four schools submitted 216 entries to the competition. Eleven teams advanced to the semifinals, where they submitted five-minute videos explaining their solution. The top team from each school advanced to the final round on Jan. 29, where they delivered 15-minute presentations to TCU executives.
“The winning teams produced fantastic insights and actionable steps for TCU,” said Rousseve. “They were forward-looking in their recommendations and went above and beyond in their assessments of the results, all while keeping TCU’s members and core values in mind. From developing a self-service analytics dashboard to building a custom web app, the winning teams put their development skills to the test in providing a ready-to-use solution.”
The finalists were judged by Rousseve and TCU executives Todd Brown, chief financial officer; Nicole Alcorn, chief member experience officer; Mitch Speer, manager of business intelligence and business transformation; Andrew Van Goey, business transformations solutions architect; and Jacob Rendall, business intelligence analyst.
“The results will provide a starting point from which our branch staff can interact and adjust their weekly cash orders,” said Rousseve. “This process will marry machine learning with human understanding to produce the best quality forecasts and orders. We will combine this solution with others in the credit union to continue growing into a financial institution for the future.”
Originally posted on Mendoza News.