E-Commerce Behavior Analysis

An end-to-end data analysis and business intelligence project built on the Online Retail II dataset. The project cleans 1M+ rows of raw transaction data, performs RFM customer segmentation, and surfaces insights through an interactive business dashboard.

---

The Problem

Raw e-commerce data is noisy — full of cancellations, missing customer IDs, and negative quantities — making it difficult to identify loyal, high-value customers and act on them.

The Solution

A Python pipeline that:

1. Cleans 1.04 million rows down to 407,664 valid transactions
2. Engineers RFM (Recency, Frequency, Monetary) features per customer
3. Segments 4,312 customers into 10 actionable behavioural groups
4. Visualises all insights in an interactive BI dashboard

---

Business Dashboard

An interactive Plotly Dash dashboard with 5 sections:

Section	What it shows
Executive Overview	Total revenue (£8.83M), orders, customers, AOV, monthly trend, top countries
RFM Segmentation	Treemap, scatter plot, revenue-by-segment bar, segment health table
Product Performance	Top 20 products by revenue and units sold
Geographic Analysis	Choropleth world map, country leaderboard across 37 markets
Time Trends	Daily / Weekly / Monthly revenue toggle, order volume heatmap

Run the dashboard

# 1. Create and activate virtual environment
python -m venv venv
venv\Scripts\activate        # Windows
# source venv/bin/activate   # macOS / Linux

# 2. Install dependencies
pip install -r requirements_dashboard.txt

# 3. Launch
python dashboard/app.py

Open http://127.0.0.1:8050 in your browser.

---

Project Structure

ecommerce-behavior-analysis/
├── data/
│   ├── online_retail_II.xlsx        # Raw source data (UCI / Kaggle)
│   ├── cleaned_retail_data.csv      # 407,664 cleaned transactions
│   └── rfm_segments.csv             # 4,312 customers with RFM scores & labels
├── notebooks/
│   └── 01_data_cleaning.ipynb       # Data cleaning & RFM segmentation pipeline
├── dashboard/
│   ├── _app.py                      # Dash application instance
│   ├── data_loader.py               # CSV loading & all pre-computed aggregations
│   ├── figures.py                   # Plotly chart factory functions (9 charts)
│   ├── layout.py                    # Full sidebar layout & 5 section renderers
│   ├── callbacks.py                 # Interactive callbacks (navigation, filters)
│   ├── app.py                       # Entry point
│   └── assets/
│       └── custom.css               # Dark-mode brand stylesheet
├── requirements_dashboard.txt
├── .gitignore
└── README.md

---

Dataset

• Source: Online Retail II — UCI Machine Learning Repository
• Period: December 2009 – December 2010
• Geography: United Kingdom (primary) + 36 international markets
• Raw size: ~1.04 million rows

---

RFM Segments

Customers are scored 1–5 on Recency, Frequency, and Monetary value, then mapped to 10 segments:

Segment	Description
Champions	Bought recently, buy often, spend the most
Loyal Customers	Buy regularly with high spend
Potential Loyalists	Recent customers with growing frequency
New Customers	Bought very recently for the first time
Promising	Recent shoppers, not yet frequent
Need Attention	Above-average recency, frequency & monetary — fading
About to Sleep	Below-average recency — may be losing interest
At Risk	Used to buy often but haven't returned
Can't Lose	Used to buy very frequently but gone a long time
Hibernating	Low recency, frequency, and monetary

---

Tech Stack

Layer	Tools
Language	Python 3.13
Data	Pandas, NumPy
Visualisation	Plotly, Plotly Express
Dashboard	Plotly Dash, Dash Bootstrap Components
Notebook	Jupyter
Version Control	Git / GitHub