🛒 Customer Segmentation with RFM and KMeans

📌 Project Overview

This project applies RFM (Recency, Frequency, Monetary) analysis and KMeans clustering to segment customers of an online retail store. The goal is to help the business design targeted marketing strategies and improve customer retention by identifying distinct customer personas.

📊 Dataset

• Source: Kaggle E-commerce Data
• Period: Dec 2010 – Dec 2011
• Size: ~540,000 rows, 8 variables
• Key Fields:
  • InvoiceNo - transaction number
  • CustomerID - unique customer ID
  • InvoiceDate - date of purchase
  • Quantity - number of items purchased
  • UnitPrice - price per item
  • Country - customer’s country

⚠️ Note: The dataset is not included in this repository due to size limitations.
Please download it from Kaggle and place it in the data/ folder.

🛠️ Methodology

1. Data Cleaning
   • Removed missing CustomerID values
   • Excluded cancelled orders (InvoiceNo starting with “C”)
   • Kept only positive Quantity and UnitPrice
   • Created TotalPrice = Quantity × UnitPrice
2. Feature Engineering (RFM)
   • Recency: days since last purchase
   • Frequency: number of unique invoices
   • Monetary: total spend
3. Preprocesing
   • Applied log-transform to reduce skewness
   • Standardized features using StandardScaler
4. Clustering
   • Used Elbow Method and Silhouette Score to test cluster counts
   • Fit KMeans clustering for k = 3 and k = 4
   • Compared profiles to choose the most actionable segmentation

📈 Results & Insights

• Cluster Profiles (k = 3)

  Cluster	Persona	Characteristic	Suggested Action
  0	Regular Shoppers	Recency ~46 days, 3–4 orders, ~$1.4k spend	Upsell with bundles or seasonal campaigns
  1	VIP / Loyal High-Value	Very recent (~13 days), 14 orders, ~$8k spend	Reward with loyalty perks, exclusive offers
  2	At-Risk / Dormant	Long inactive (~161 days), ~1 order, ~$350 spend	Win-back discounts, targeted reactivation

• Cluster Profiles (k = 4)

  Cluster	Persona	Characteristic	Suggested Action
  0	New / Low-Spend Recent	Recent (~17 days), 2 orders, ~$545 spend	Onboarding emails, encourage repeat purchase
  1	Established Regulars	Recency ~68 days, 4 orders, ~$1.8k spend	Maintain engagement, cross-sell or upsell
  2	At-Risk / Dormant	Long inactive (~179 days), ~1 order, ~$350 spend	Reactivation campaigns, discounts
  3	VIP / Loyal High-Value	Very recent (~10 days), 14 orders, ~$8.3k spend	Exclusive rewards, maintain strong loyalty

🔹 Visualizations

Heatmaps of z-scores across clusters

Bar plots of average Recency, Frequency, Monetary

Scatter plots (Recency vs Monetary, colored by cluster)

💡 Business Implications

• VIPs → loyalty perks, exclusive rewards
• Regular Shoppers → upsell and seasonal offers
• At-Risk Customers → win-back campaigns
• New Customers → onboarding and engagement

⚖️ Limitations & Next Steps

• KMeans assumes spherical clusters; alternative methods (DBSCAN, hierarchical) could reveal different patterns
• RFM ignores product categories; future work could add product-level or seasonality features
• Deploying a Streamlit app or Tableau dashboard would make this segmentation interactive for business users

🛠️ Tech Stack

• Python: pandas, numpy, scikit-learn, seaborn, matplotlib
• Clustering: KMeans
• Visualization: seaborn, matplotlib

⚙️ Installation

Clone this repository and install dependencies:

git clone https://github.com/aituar17/Customer_Segmentation.git
cd Customer_Segmentation
pip install -r requirements.txt

📂 Project Structure

Customer_Segmentation/
├── data/ # Dataset folder (not uploaded to GitHub due to size; see data/README.md)
├── images/ # Saved plots for README (heatmaps, bar plots, scatter plots)
│ └── zscore_heatmaps.png
│ └── barplots_k3.png
│ └── barplots_k4.png
│ └── scatter_k3.png
│ └── scatter_k4.png
├── notebooks/ # Jupyter notebook with full analysis
│ └── Customer_Segmentation.ipynb
├── .gitignore # Ignore large files (e.g., dataset)
├── README.md # Project documentation
└── requirements.txt # Dependencies for reproducibility