Unlocking Business Insights with Customer Segmentation
To move beyond a “one-size-fits-all” marketing strategy, this project segmented a customer base into distinct groups based on their purchasing behavior. The goal was to provide actionable insights for targeted ad campaigns, personalized promotions, and customer retention strategies.
Five new features were engineered: purchase Frequency and Recency, CLV, Average Purchase Cost and Customer Age.
The analysis used unsupervised machine learning with K-Means clustering. The Elbow and Silhouette methods guided the choice of the optimal number of segments. To handle high-dimensional data and enable clear visualization, I applied dimensionality reduction with PCA and t-SNE.
The final model identified five distinct customer personas:
- Core loyal customers, who bought cheaper items more frequently and recently
- High-value spenders, who bought more expensive items less frequently
- Older and potential churn risk customers, whose purchases were less recent and so might be leaving the customer base
- Two overlapping clusters, both consisting of customers who recently purchased lower-cost items, differentiated primarily by age.
Tech Stack
- Algorithms: K-Means Clustering, Elbow and Silhouette Methods
- Dimensionality Reduction: PCA, t-SNE
- Libraries: Python, Pandas, Scikit-learn, Matplotlib, Seaborn