Customer Analytics Dashboard

Interactive Streamlit dashboard for customer behavior analytics using the UCI Online Retail Dataset.

The app includes:

• sales and product performance views
• RFM customer segmentation
• churn propensity modelling
• probabilistic CLV prediction
• revenue time-series forecasting

Features

• Overview: KPI cards, monthly trends and top countries/products by revenue
• RFM Segmentation: K-Means clustering with elbow/silhouette previews and segment labels
• Churn Prediction: random forest classification with threshold tuning, confusion matrix and precision-recall / ROC curves
• CLV Prediction: BG/NBD + Gamma-Gamma modelling via pymc-marketing
• Revenue Forecasting: SARIMA and Theta-method models (statsmodels) with holdout backtest metrics and 90% confidence intervals

To do

• Enhance segment labelling (currently a static list chosen in order)

Setup

1) Install dependencies

Recommended (uses the lock file for reproducible installs):

uv sync

Or with a standard virtual environment:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2) Run the app

uv run streamlit run app.py
# or, if using a venv:
streamlit run app.py

Project Structure

.
├── app.py
├── app_pages/
│   ├── 1_Overview.py
│   ├── 2_RFM_Segmentation.py
│   ├── 3_Churn_Prediction.py
│   ├── 4_CLV_Prediction.py
│   └── 5_Revenue_Forecasting.py
├── utils.py
├── requirements.txt
├── static/
├── .streamlit/
│   ├── config.toml
│   └── pages.toml
└── data/
    └── online_retail_II.csv

Dependencies

• streamlit for the app UI
• pandas, numpy for data wrangling
• plotly for interactive visualisations
• scikit-learn, scipy for RFM clustering and transformations
• statsmodels for SARIMA and Theta-method forecasting
• pymc-marketing, pytensor, nutpie for CLV modelling

Notes

• The app currently reads data/online_retail_II.csv directly (in utils.py).
• If dependency installation fails, ensure requirements.txt includes all packages imported by the code (notably numpy and scipy).