🧑‍🤝‍🧑 UK Customer Segmentation

🧠 Objective

Segment customers of a UK-based online gift retailer using transaction data to uncover behavior-driven groups for targeted marketing.

📦 Dataset

• Transactions from Dec 2010 to Dec 2011
• 541,909 records, 8 columns
• Key columns: InvoiceDate, Quantity, UnitPrice, CustomerID, Country

🔍 Methodology

1. Data Wrangling:

   • Removed canceled orders and missing customer IDs
   • Filtered out negative or zero Quantity and UnitPrice
   • Parsed dates into datetime format

2. Feature Engineering:

   • Built RFM features:
     • Recency: Days since last purchase
     • Frequency: Total purchases
     • Monetary: Total amount spent

3. Preprocessing:

   • Log-transformed and scaled RFM values
   • Applied PCA for 2D visualization

4. Clustering (K-Means):

   • Determined optimal k=3 using Elbow + Silhouette methods
   • Assigned customers to 3 clusters

5. Visualization:

   • PCA scatter plot of clusters
   • Boxplots and heatmap of RFM by cluster
   • Choropleth map of customer countries

🧩 Segment Profiles

🧠 Business Insight

• Cluster 0: High spenders, but inactive → win back with rewards or early access.
• Cluster 1: Likely one-time buyers → nudge with follow-ups or discounts.
• Cluster 2: Moderate engagement → develop into loyal customers with tailored campaigns.

🛠️ Tools

• Python (pandas, sklearn, seaborn, plotly)
• K-Means, PCA
• Jupyter Notebook

📓 View Full Jupyter Notebook on GitHub