Hari Babu C H

Project Case Study
2014 – 2017

Superstore Analytics:
End-to-End Retail Business
Intelligence Project

9,994
Records Analysed
793
Unique Customers
30+
Business Queries
From Raw Retail Data to Actionable Business Intelligence

This end-to-end data analytics project centres on the well-known Superstore dataset — a four-year (2014–2017) transactional record of a fictional US retail chain spanning 9,994 order lines, 793 unique customers, 1,862 products, and $2.3 M in revenue across four regional markets (East, West, Central, South). The project follows a complete analytical workflow: raw data ingestion, SQL-based data cleaning and exploratory analysis, Python-powered visual exploration, and an interactive Streamlit dashboard — all designed to surface actionable insights around profitability, customer segmentation, product performance, discount governance, and geographic distribution.

  • Clean and standardise raw transactional data using both SQL and Python pipelines
  • Conduct comprehensive exploratory data analysis (EDA) across time, geography, and product dimensions
  • Answer 30+ structured business questions with SQL to drive decision-making
  • Build an interactive multi-chart dashboard with Streamlit and Plotly for real-time filtering
  • Identify loss-making products, discount governance gaps, and regional profit opportunities
  • Deliver customer segmentation insights to support targeted retention and acquisition strategies

The project employs a professional-grade analytics stack spanning SQL, Python, and interactive visualisation libraries to build a fully reproducible, end-to-end pipeline:

Python (Jupyter Notebook) SQL (PostgreSQL-compatible) Pandas & NumPy Plotly & Seaborn Streamlit Dashboard Matplotlib Scikit-learn Microsoft Excel / CSV
  • Data cleaning, type standardisation, and outlier detection (SQL & Python)
  • Feature engineering — profit margin, shipping days, discount bands, CLV, customer value segments
  • SQL Exploratory Data Analysis (EDA) — trends, distributions, correlations
  • 30+ structured business queries covering revenue, profitability, geography, and discounting
  • Interactive Streamlit dashboard with dynamic filters and downloadable outputs
  • Statistical visualisation and business storytelling with Plotly and Seaborn
  • Environmental reporting and communication
Repository Layout
📁 project-root/
├── 📂 Dataset/
│   ├── Raw_Data.csv          (9,994 rows · 21 columns)
│   └── Cleaned_Data.csv      (9,994 rows · 38 columns)
├── 📂 Python/
│   ├── Data_Cleaning.ipynb
│   ├── EDA.ipynb
│   ├── Data_Visualization.ipynb
│   ├── Dashboard.py          (Streamlit app)
│   └── Requirements.txt
├── 📂 SQL/
│   ├── Data_Cleaning.sql
│   ├── EDA.sql
│   └── Business_Queries.sql  (30+ queries)
└── 📄 README.md
Project Visuals

A visual walkthrough of the analysis outputs — from the Streamlit KPI dashboard to exploratory charts and SQL-driven business insight visualisations built in Python and Plotly.

Streamlit Analytics Dashboard
Streamlit Analytics Dashboard
Revenue & Profit Trend
Revenue & Profit Trend
Geographic Treemap
Geographic Revenue Treemap
What the Data Revealed

Across 9,994 order lines and $2.3 M in total revenue, the analysis uncovered a 12.5% overall profit margin with significant variation by category — Technology led in profitability while Furniture lagged, and several sub-categories (notably Tables and Bookcases) operated at a net loss. The West region produced the highest profit, while specific states and cities exhibited structurally negative returns driven by excessive discounting. Orders carrying discounts above 30% were found to be the primary source of avoidable losses, pointing to a clear need for discount governance controls. On the customer side, the Consumer segment generated the most orders, but Corporate customers delivered higher average order values. CLV segmentation identified a small cohort of high-value repeat buyers who merit dedicated retention investment.

This project demonstrates how a rigorous, multi-tool analytics workflow — combining SQL for data engineering and business querying, Python notebooks for EDA and statistical visualisation, and Streamlit for interactive dashboarding — can translate raw transactional data into clear, executive-ready insights. The Superstore dataset, spanning four years and four US regions, served as a realistic sandbox for practicing the full spectrum of a data analyst's craft: from messy raw ingestion through to polished, filterable dashboards. The 30+ business queries and engineered features (profit margin, shipping days, discount bands, CLV, customer segments) collectively form a blueprint that is directly transferable to real-world retail, e-commerce, or supply-chain analytics environments.