The Practicum Pinnacle

Welcome to the Practicum Pinnacle Newsletter! This newsletter aims to highlight the achievements and insights from the Master of Science in Business Analytics program’s practicum projects, showcasing student work, industry collaborations, and the impact of data-driven solutions on real-world business challenges. In today’s edition, we share a Home Depot case study, "Pro Customer Purchase Prediction".

A Message from Scott Radcliffe, Managing Director of the MSBA Program

On behalf of the Emory Goizueta MSBA program, I would like to extend our sincerest gratitude to our esteemed practicum sponsors for the 2025 academic year. This year, students gained invaluable real-world experience applying data science skills to actual business challenges with 1st Franklin Financial, CONA, Cox Automotive, Equifax, GoTo Foods, Liberty Latin America, Macy's, NCR Voyix, Newell Brands, Pennybacker Capital, and The Home Depot. 

The practicum is paramount for student development. Our students honed critical skills, from data cleaning and translating business problems into actionable data solutions to building AI products using diverse tools like Power BI, Python, and SQL. They enhanced project management, teamwork, and communication with technical and non-technical stakeholders. Furthermore, direct client collaboration fostered significant professional growth, marked by their ability to internalize feedback and demonstrate "growth in professional maturity". We sincerely appreciate your partnership in preparing our future data leaders.

We are currently in discussions with potential sponsors for the Spring 2026 projects. Learn more about practicum sponsorship. You may also contact Managing Director Scott Radcliffe, CAP-X (scott.radcliffe@emory.edu), or Academic Director Ramnath (Ram) Chellappa (ramnath.chellappa@emory.edu) via email or LinkedIn messaging.

Pro Customer Purchase Prediction

Business Problem:

The Home Depot aimed to maximize value and margins from its Pro customers (B2B contractors), who drive a major portion of their revenue. The core challenge was to predict when and which products Pro customers would repurchase, and in what quantities, to enable targeted engagement, boost retention through personalized offers, and generate incremental sales by accelerating repurchases. The goal was a predictive model for recommending previously purchased products and their quantities one month ahead.

Technical Approach and Outcome:

The project utilized anonymized transactional, SKU, and customer data. It involved comprehensive feature engineering and explored three modeling approaches: Statistical Models (Poisson-Gamma and Modified Poisson-Gamma), LSTM (Long Short-Term Memory) for sequential purchase behavior, and XGBoost for its performance on structured data. XGBoost achieved strong performance metrics: AUC-ROC of 0.9218 for classification and RMSE of 87.2941 for regression. The models predict repurchase probability and quantity for customer-SKU pairs one month in advance.

Business Outcome:

The solution enables targeted email promotions and personalized offers by identifying likely repurchasers and their product needs. This leads to improved forecasting accuracy, allowing for more relevant promotions and potentially significant sales lift through increased repeat purchase rates. The models support marketing campaigns by triggering promotions before forecasted needs, prioritizing high-margin SKUs like plumbing and hardware, and suggesting product bundles. This enhances customer satisfaction and operational efficiency.

The Master of Science in Business Analytics program was a proud gold sponsor of this year's AI Optimized Conference.

Scott shared how collaboration between universities and industry can lead to impactful, real-world AI and data solutions and how practicum projects are a powerful way to spark innovation while maximizing ROI.

Professor Ramnath (Ram) Chellappa delivered a genuinely thought-provoking and inspiring session, "The Unintended Consequences of AI Deployment," at the Optimized AI Conference. His insights on systemic bias in AI challenged us to think deeper about the impact of the technologies we build and deploy.