Customer Segmentation: RFM Analysis & Unsupervised K-Means

Treating all customers identically is a recipe for wasted marketing spend. In this class, we master Unsupervised Customer Segmentation by combining classical RFM Analysis with K-Means Clustering.

The 3 Dimensions of RFM Modeling

• Recency (R): Days elapsed since the customer's last interaction. (Low days = High engagement).
• Frequency (F): Total number of completed orders. (High count = High brand loyalty).
• Monetary (M): Cumulative revenue generated. (High dollars = High commercial value).

End-to-End Python Pipeline: RFM + K-Means

import numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score

# Sample customer transaction dataset
np.random.seed(42)
n_cust = 500
df_rfm = pd.DataFrame({
    'recency_days': np.random.exponential(scale=45, size=n_cust).astype(int),
    'frequency_orders': np.random.poisson(lam=6, size=n_cust) + 1,
    'monetary_value': np.random.exponential(scale=350, size=n_cust) + 25
})

# Step 1: Preprocessing & Scaling (Mandatory for distance-based clustering!)
scaler = StandardScaler()
rfm_scaled = scaler.fit_transform(df_rfm)

# Step 2: Determine optimal k using Silhouette Analysis
for k in range(2, 6):
    km = KMeans(n_clusters=k, init='k-means++', random_state=42)
    labels = km.fit_predict(rfm_scaled)
    sil_score = silhouette_score(rfm_scaled, labels)
    print(f"k = {k} → Silhouette Score: {sil_score:.4f}")

# Step 3: Fit final 3-cluster segmentation model
final_kmeans = KMeans(n_clusters=3, init='k-means++', random_state=42)
df_rfm['cluster_id'] = final_kmeans.fit_predict(rfm_scaled)

# Step 4: Cluster Profiling
profile = df_rfm.groupby('cluster_id').agg({
    'recency_days': 'mean',
    'frequency_orders': 'mean',
    'monetary_value': 'mean',
    'cluster_id': 'count'
}).rename(columns={'cluster_id': 'customer_count'})

print("\n--- Segment Business Profiles ---\n", profile.round(2))