Customer Segmentation with RFM Model and K-Means Clustering

🔍 Overview

This project demonstrates how to segment customers using the RFM (Recency, Frequency, Monetary) model combined with K-Means clustering. It helps identify patterns in customer behavior and groups them into actionable segments such as "Loyal Customers", "At Risk", and "Need Attention".

All customer data used in this project comes from OTA (Online Travel Agency) transactional data from kaggle. Customer IDs were randomized, it didn't exist in the data.

---

📁 Dataset

• Source: Proprietary OTA customer booking data
• Anonymized with randomized CustomerID
• Time Period: 1 year snapshot

Fields used:

• Recency: Days since last booking
• Frequency: Number of bookings
• Monetary: Total value of bookings

---

🧪 Step 1: Data Cleaning & Preprocessing

After loading the dataset, I:

• Removed duplicates and nulls
• Calculated RFM values
• Applied square root transformation to reduce skew

---

📊 Step 2: RFM Scoring

Each customer was scored on a scale of 1 to 4 based on R, F, and M quartiles using pd.qcut. These scores were summed into an RFM_Score.

rfm_df['RFM_Score'] = rfm_df['R_Score'] + rfm_df['F_Score'] + rfm_df['M_Score']

Output:

---

🔄 Step 3: Optimal Cluster Selection

To determine the best number of clusters for K-Means, I used:

• Elbow Method
• Silhouette Score

from sklearn.metrics import silhouette_score

Output:

• Based on results, I chose k = 4.

---

📊 Step 4: Clustering with K-Means

I standardized the RFM sqrt values and applied KMeans clustering.

scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)
kmeans = KMeans(n_clusters=4, random_state=42)
rfm_df['Cluster'] = kmeans.fit_predict(X_scaled)

Output:

---

📋 Step 5: Inspect distributions after preprocessing

Before running the clustering algorithms I checked whether the square-root transformation really reduced skewness and brought the three R ∙ F ∙ M features onto comparable shapes.

Output:

📋 Step 6: Summarise the resulting clusters

Once the optimal k was chosen I calculated the mean Recency, Frequency, and Monetary values for each cluster and counted how many customers fell into every group.

Output:

📋 Step 7: Segment Mapping

I assigned human-readable labels to each cluster based on their average RFM behavior:

cluster_labels = {
    0: 'Champions',
    1: 'At Risk',
    2: 'Loyal Customers',
    3: 'Need Attention'
}
rfm_df['Segment'] = rfm_df['Cluster'].map(cluster_labels)

Output:

---

📈 Step 8: Final Analysis & Insights

I aggregated cluster statistics and visualized them using bar plots.

cluster_summary = rfm_df.groupby('Segment')[['Recency', 'Frequency', 'Monetary']].mean().round(1)

Output:

---

🔺 Segment Takeaways

Segment	Insight
Champions	Recently active, high value, and frequent bookers
Loyal Customers	Consistent bookers with solid spend
At Risk	Haven't booked in a while but used to spend a lot
Need Attention	Low engagement and low spend

---

🪧 Tools Used

• Python (Pandas, Scikit-learn, Seaborn, Matplotlib)
• JupyterLab (GitHub Codespaces)

---

📚 How to Use This Repo

git clone https://github.com/13Saksham/rfm-customer-segmentation-python.git
cd rfm-customer-segmentation-python

---

🙏 Acknowledgements

• Kaggle Datasets
• Concepts inspired by DataCamp, StackOverflow, and self-practice