Industrial Scientific: Customer Insights
Capstone Team: Ralph Del Negro, Milica Kosic, Rohi Nallamolu, Alexandra Quan, Amy Ran
The sponsor company has noticed an opportunity to improves upsell and cross-sell activity by using historical data to identify which sites are using less product than they need. The Tepper MSBA team met with members of the marketing, sales, ﬁeld services, and IT teams to understand the business and data context. These interviews identiﬁed the key problem of understanding how account characteristics and behavior impact downsells and upsells.
Using product data, Salesforce account information, and contract amendment data, the Tepper team prototyped multiple machine learning models to predict product utilization rates, contract changes, and equipment volume. Ultimately the team reﬁned a customer segmentation model, which used k-means clustering on the principal components of an 80 feature dataset.
The customer segmentation model output 3 clusters: low risk of downsells, moderate risk of downsells, and high growth opportunities. Key insights of these models include strong correlation between customer activity and upsells, the importance of product choice in predicting contract amendments, and the observable predictive value of the company’s own customer health metric.