GIVEN-CHINYAMA/zambia-retail-intelligence-platform

An end-to-end data science pipeline for Zambian retail analytics — featuring synthetic data generation, EDA, ARIMA & XGBoost demand forecasting, RFM customer segmentation, stockout risk detection, and an executive BI dashboard. Built with Python, Pandas, Scikit-learn, XGBoost, Statsmodels & Plotly # 🛒 Zambia National Retail Intelligence Platform ### *An End-to-End Data Science Project | Retail Analytics · Demand Forecasting · Business Intelligence* Author: Given Chinyama  |  Date: June 2026  |  LinkedIn  |  GitHub --- ## 📌 Overview Zambia's retail sector is undergoing rapid transformation — driven by urbanisation, mobile money adoption, and a growing middle class across cities like Lusaka, Kitwe, Ndola, and Livingstone. Yet the vast majority of retailers still operate without data-driven insight, relying on intuition over analytics. This project addresses that gap by building a **National Retail Intelligence Platform** — a production-grade, end-to-end data science pipeline that simulates, cleans, analyses, forecasts, and visualises Zambian retail data across 7 cities, 8 store chains, and 24 product categories. The platform is designed to serve three distinct audiences: | Audience | Use Case | |---|---| | 🏪 Retailers & FMCG distributors | Reduce stockouts, optimise inventory, plan promotions | | 📊 Data scientists & analysts | Reproducible forecasting and segmentation pipelines | | 🏛️ Government & trade policy teams | National retail price monitoring and trade analytics | --- ## ✨ Key Features - **Realistic synthetic dataset** — 80,000 transactions across 7 Zambian cities, population-weighted and seasonally adjusted, all priced in Zambian Kwacha (ZMW) - **Dual forecasting models** — ARIMA for seasonal trend interpretation and XGBoost for high-accuracy daily demand prediction - **RFM customer segmentation** — Recency, Frequency, and Monetary analysis combined with K-Means clustering to identify Champions, Loyal Regulars, New Customers, and At-Risk segments - **Stockout and overstock risk detection** — Rule-based flags combined with Z-score anomaly detection on daily revenue streams - **Executive BI dashboard** — A 9-panel interactive Plotly dashboard covering revenue, margins, seasonality, promotions, and Yo …