Bank Strategy Client Segmentation
Matheus Guerreiro
Published
Development of a customer segmentation model for a banking credit card portfolio using K-Means clustering, aimed at increasing profitability through targeted CRM strategies. The project grouped 8,950 clients into 8 behavioral clusters based on spending patterns, credit usage, cash advance behavior, and payment profile. As a result, the segmentation enables actionable strategies for revenue growth (upsell/cross-sell), cost reduction (more efficient campaigns), and risk mitigation (loss prevention). The expected annual ROI ranges from ~$30K (conservative) to ~$250K+ (optimistic), with a moderate expected impact around ~$100K/year, subject to A/B testing validation.
Tools used:
Python, Pandas, NumPy, and Scikit-learn.
Jupyter Notebooks.
Machine Learning with K-Means Clustering.
PCA for visualization and cluster interpretability.
Cluster evaluation metrics (Silhouette Score, Davies-Bouldin, Calinski-Harabasz)