🛒 E-Commerce Customer Intelligence — RFM Segmentation & Revenue Analytics

A production-grade data science project demonstrating end-to-end customer analytics, statistical hypothesis testing, and actionable business intelligence.

Executive Summary • Key Findings • Statistical Analysis • Visuals • How to Run

---

🗂 Project Structure

ecommerce-rfm-analysis/
│
├── src/
│   └── analysis.py          # Full modular pipeline (PEP 8 compliant)
│
├── visuals/                 # All generated charts (PNG, 150 dpi)
│   ├── 01_segment_distribution.png
│   ├── 02_rfm_distributions.png
│   ├── 03_monetary_by_segment.png
│   ├── 04_rfm_scatter.png
│   ├── 05_monthly_revenue.png
│   ├── 06_correlation_heatmap.png
│   └── 07_confidence_intervals.png
│
├── reports/
│   └── executive_report.txt # Full statistical report
│
├── requirements.txt
└── README.md

---

📋 Executive Summary

Problem: An e-commerce business lacks clarity on which customers drive revenue, which are churning, and where marketing spend should be prioritized.

Solution: This project applies RFM (Recency, Frequency, Monetary) segmentation combined with rigorous statistical testing to quantify customer value, validate segment differences, and generate five data-backed business recommendations.

Dataset Overview:

• ~50,000 raw transactions spanning 2 years (2022–2023)
• 1,200 unique customers across 5 countries
• Intentional data quality issues: duplicates, missing IDs, cancelled orders, price outliers
• Final clean dataset retention: ~87% after robust cleaning

Stack: Python · Pandas · NumPy · SciPy · Matplotlib · Seaborn

---

🔍 Key Findings

Customer Segments at a Glance

Segment	Customers	Avg Spend	Share of Revenue
🏆 Champions	~198	$18,400+	~30%
💜 Loyal Customers	~210	$14,200+	~25%
⚡ Potential Loyalists	~180	$13,500+	~20%
⚠️ At Risk	~195	$11,800+	~18%
💤 Lost	~120	$7,200+	~7%

Statistical Highlights

Metric	Value	Interpretation
Mean Monetary Value	$13,167	Average lifetime spend per customer
Coefficient of Variation	27.6%	Moderate spend dispersion — pricing opportunity
Skewness	+0.40	Right-skewed — high-value outliers pull mean up
Kurtosis	+0.12	Near-normal tails — few extreme spenders
95% CI for Mean	[$12,961 – $13,372]	Tight CI confirms reliable estimate

Hypothesis Tests (α = 0.05)

Test	Result	Conclusion
Welch's t-test: Champions vs Loyal	t=14.89, p≈0.00	✅ Champions spend significantly more
One-Way ANOVA: All Segments	F=129.53, p≈0.00	✅ Segment means are statistically distinct

---

📊 Visual Storytelling

All charts use a dark-theme professional design system for maximum impact in portfolio presentations and client decks.

1 · Segment Distribution

Who are our customers?

---

2 · RFM Feature Distributions

How are Recency, Frequency, and Monetary value distributed?

---

3 · Monetary Value by Segment

Do segments actually differ in spending behavior?

---

4 · RFM Landscape Scatter

Where does each customer sit in the R×F×M space?

---

5 · Monthly Revenue Trend

How does revenue evolve over time?

---

6 · Correlation Heatmap

How do RFM dimensions relate to each other?

---

7 · Confidence Intervals — Forest Plot

How certain are we about segment mean spend?

---

📐 Statistical Deep Dive

Why RFM?

RFM is a behavioral segmentation framework proven in direct marketing since the 1990s. It captures three independent dimensions of customer value:

• Recency — customers who bought recently are more likely to buy again
• Frequency — customers who buy often signal brand loyalty
• Monetary — customers who spend more have higher LTV (Lifetime Value)

Scoring Logic

Each dimension is scored 1–5 using quintile binning:

rfm["r_score"] = pd.qcut(rfm["recency"],   5, labels=[5, 4, 3, 2, 1])
rfm["f_score"] = pd.qcut(rfm["frequency"].rank(method="first"), 5, labels=[1,2,3,4,5])
rfm["m_score"] = pd.qcut(rfm["monetary"].rank(method="first"),  5, labels=[1,2,3,4,5])

Note: Recency is scored inversely (lower recency = higher score).
Rank-based binning is used for Frequency and Monetary to handle ties robustly.

Skewness & Kurtosis Interpretation

Metric	Value	What it means
Skewness = +0.40	Moderate right skew	A minority of customers generate outsized revenue — prioritize Champions
Kurtosis = +0.12	Near-mesokurtic	Spend distribution is stable — no extreme tail events distorting averages

---

💡 Business Recommendations

Based on the statistical analysis, five high-priority interventions are recommended:

1 · 🏆 Champion VIP Program

Champions account for ~30% of total revenue. Action: Launch an invitation-only loyalty tier with exclusive early access, dedicated account manager, and quarterly gifting. Projection: 15–20% churn reduction → $12K–18K in protected annual revenue.

2 · 🚨 At-Risk Win-Back Campaign

~195 customers show declining recency with historically strong frequency. Action: Deploy personalized email with 20% "We Miss You" discount within 7 days of inactivity threshold. Projection: 25% reactivation × $180 avg order → +$8,775 in recovered revenue.

3 · 🔁 Lost Segment A/B Test

~120 customers are fully churned (low R, F, M). Action: A/B test SMS vs. email win-back. Segment by original product category for hyper-relevant offers. Projection: Identify the $4–$6 CAC channel with highest ROI for future re-engagement budget.

4 · 📈 Frequency Uplift for Potential Loyalists

This segment buys well per order but infrequently. Action: Introduce a tiered subscription prompt ("Buy 3 times this quarter, unlock 10% off forever"). Projection: +0.8 orders/year per customer across 180 users → +$22,000 incremental annual revenue.

5 · 💲 Dynamic Pricing for High-CV SKUs

CV of 27.6% signals segment-level price sensitivity differences. Action: Audit top-10 revenue SKUs. Apply dynamic pricing (+5%) for Champions and Loyal segments. Projection: 5% price lift on $180K of inelastic SKU revenue → +$9,000 gross margin improvement.

---

⚙️ How to Run

Prerequisites

Python 3.10+

Installation

# 1. Clone the repository
git clone https://github.com/thed700/ecommerce-rfm-analysis.git
cd ecommerce-rfm-analysis

# 2. Create a virtual environment (recommended)
python -m venv venv
source venv/bin/activate        # macOS/Linux
venv\Scripts\activate           # Windows

# 3. Install dependencies
pip install -r requirements.txt

Run the Full Pipeline

python src/analysis.py

This will:

1. Generate realistic synthetic e-commerce data with intentional data quality issues
2. Clean and validate all records (with full audit trail)
3. Engineer RFM features and assign segment labels
4. Compute descriptive statistics, skewness, kurtosis, CV, and confidence intervals
5. Run Welch's t-test and One-Way ANOVA
6. Output 7 publication-quality charts to visuals/
7. Write the full executive report to reports/executive_report.txt

Requirements

pandas>=2.0
numpy>=1.26
matplotlib>=3.8
seaborn>=0.13
scipy>=1.12

---

🔬 Code Quality Standards

• ✅ PEP 8 compliant throughout
• ✅ Fully modular — each pipeline stage is an independent function
• ✅ Type hints on all public functions
• ✅ Docstrings on every function (Google style)
• ✅ Reproducible — fixed random_state / seed throughout
• ✅ No data leakage — RFM scoring computed on clean data only

---

👤 Author

Akmal — Data Analytics & Business Intelligence
📁 GitHub: @thed700
🎓 Specialization: Statistical Modeling · Python · SQL · Business Intelligence

---

📄 License

MIT LICENSE — free to use, adapt, and build upon with attribution.