Data Analysis Foundations

Author: Anandi Mahure | MSc Data Science, University of Bath (Dean's Award 2025) LinkedIn: linkedin.com/in/anandirm Tools: Python · Pandas · NumPy · SciPy · Matplotlib · pytest · GitHub Actions Domain: Retail Customer Analytics · EDA · Statistical Inference · Segmentation

---

Overview

This repository combines conceptual foundations with a production-quality analytics pipeline applied to a retail customer behaviour dataset (500 customers, 11 features).

The pipeline demonstrates the full data science workflow: load → clean → describe → detect outliers → test hypotheses → segment → visualise.

Three reference notebooks cover the underlying theory in depth:

• Python_Basics_Advanced.ipynb — NumPy, Pandas, OOP, functional programming
• Statistics_DeepDive.ipynb — Probability, Bayes' theorem, hypothesis testing, decision theory
• EDA_DeepDive.ipynb — Outlier detection, correlation, feature engineering, IQR vs Z-score

---

Business Questions Answered

#	Business Question	Technique
1	How are customer spending scores distributed?	Distribution analysis + descriptive stats
2	Does income predict spending behaviour?	Scatter plot + Pearson correlation + regression
3	Which regions have the most customers and highest engagement?	GROUP BY + bar chart
4	Do high-churn customers spend significantly less?	Independent samples t-test (p < 0.05)
5	Does income differ significantly by gender?	Independent samples t-test
6	Which category drives the most total customer spend?	Aggregation + bar chart
7	How do IQR and Z-score outlier detection methods compare?	Dual-method comparison
8	Which spending tier is most at risk of churn?	Customer segmentation + grouped bar chart

---

Architecture

data/
└── customer_analytics.csv       ← 500-row synthetic retail dataset (11 features)
        │
        └──► analysis.py         ← Python analytics pipeline
                │   load & impute → descriptive stats → outlier detection
                │   → hypothesis testing → segmentation → visualisation
                │
                └──► charts/
                        ├── 01_spending_distribution.png
                        ├── 02_income_vs_spending.png
                        ├── 03_regional_breakdown.png
                        ├── 04_churn_by_tier.png
                        └── 05_category_overview.png

tests/
└── test_analysis.py     ← pytest suite (36 assertions)
                            schema · business logic · stats · outliers
                            hypothesis testing · segmentation · charts

Reference Notebooks:
├── Python_Basics_Advanced.ipynb
├── Statistics_DeepDive.ipynb
└── EDA_DeepDive.ipynb

.github/workflows/ci.yml  ← GitHub Actions CI (Python 3.11)

---

Dashboard Output

Spending Score Distribution

Income vs Spending Score by Churn Risk

Regional Customer Analysis

Churn Risk by Spending Tier

Category Revenue & Outlier Analysis

---

Key Findings

• Spending scores are normally distributed (mean ~52), with slight right skew — majority of customers score between 30–70
• Income and spending score show weak positive correlation — high earners don't necessarily spend more
• London accounts for 35% of the customer base; spending scores are consistent across all regions
• High-churn customers have significantly lower spending scores (t-test p < 0.001) — clear early warning signal
• Low spenders carry >80% high churn risk; high spenders are predominantly low churn
• Clothing drives the highest total customer spend across all categories
• IQR method detects ~2.8% outliers in total spend; Z-score (stricter) identifies ~0.4%

---

How To Run

# Clone the repo
git clone https://github.com/anandi-mahure/data-analysis-foundations.git
cd data-analysis-foundations

# Install dependencies
pip install -r requirements.txt

# Run the analytics pipeline (generates all 5 charts)
python analysis.py

# Run the test suite
pytest tests/ -v

---

Skills Demonstrated

Python Pandas NumPy SciPy Matplotlib Statistical Testing Outlier Detection Customer Segmentation EDA Hypothesis Testing pytest GitHub Actions CI/CD Descriptive Statistics Data Cleaning Feature Engineering Retail Analytics