Datasets Guide¶

Every dataset used across the course projects and capstones — with download links, what's inside, and which project it powers. All are free and publicly available.

How to use this guide

Each project's dataset-guide.md has detailed schema documentation. This page is the master index of where to get the data. Download the dataset for whichever project or capstone you're working on.

Project Datasets¶

Project 1 — Sales Dashboard¶

Dataset: Sample Superstore Link: https://www.kaggle.com/datasets/vivek468/superstore-dataset-final Contents: ~10,000 retail orders, 4 years, with sales, profit, discount, region, category. Why: The classic beginner sales dataset. Clean, rich, and ideal for revenue/profit analysis. Encoding note: load with encoding="latin-1".

Project 2 — Customer Churn Analysis¶

Dataset: Telco Customer Churn Link: https://www.kaggle.com/datasets/blastchar/telco-customer-churn Contents: 7,043 customers, 21 features, churn flag. Why: The standard churn dataset. Clear target variable, strong actionable drivers.

Project 3 — HR Analytics¶

Dataset: IBM HR Analytics Employee Attrition Link: https://www.kaggle.com/datasets/pavansubhasht/ibm-hr-analytics-attrition-dataset Contents: 1,470 employees, 35 features, attrition flag. Why: Clean HR dataset for attrition analysis (and ethical-judgment practice).

Project 4 — COVID-19 Data Analysis¶

Dataset: Our World in Data COVID-19 Link: https://github.com/owid/covid-19-data/tree/master/public/data Contents: Daily cases, deaths, vaccinations per country, with per-capita columns. Why: The gold standard for time-series and per-capita analysis practice.

Project 5 — E-commerce Analytics¶

Dataset: Brazilian E-Commerce (Olist) Link: https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce Contents: ~100k orders across 8 related tables (orders, items, customers, payments, reviews, products, sellers, geolocation). Why: The most realistic relational dataset — ideal for multi-table joins and customer LTV.

Project 6 — Financial Performance Analysis¶

Dataset: Microsoft Financial Sample Link: https://learn.microsoft.com/en-us/power-bi/create-reports/sample-financial-download Contents: Sales transactions by segment, country, product, with cost, discount, and profit. Why: Clean financial data for P&L, margin, and variance analysis.

Project 7 — IPL Sports Analytics¶

Dataset: IPL Complete Dataset (2008–2020+) Link: https://www.kaggle.com/datasets/patrickb1912/ipl-complete-dataset-20082020 Contents: matches.csv (~1,000 matches) + deliveries.csv (~250,000 ball-by-ball records). Why: Large, granular event data — great for aggregation and domain-specific logic.

Capstone Dataset Suggestions¶

Capstone	Suggested Datasets
Sales Performance Dashboard	Superstore, or any retail sales data
Customer Segmentation	Online Retail II (UCI), Olist
HR Attrition Analytics	IBM HR Analytics
E-commerce Business Dashboard	Olist (multi-table)
Financial Analytics System	Microsoft Financial Sample, synthetic P&L
Social Media Analytics	Kaggle social media datasets, or your own exported analytics

General-Purpose Dataset Sources¶

Kaggle¶

https://www.kaggle.com/datasets — 60,000+ datasets across every domain. Best starting point. Includes notebooks showing how others analysed each dataset.

UCI Machine Learning Repository¶

https://archive.ics.uci.edu — classic, well-documented academic datasets. Online Retail II and Bank Marketing are particularly useful.

data.gov / data.gov.uk¶

Government open data — housing, health, transport, economics. Great for portfolio projects with real public-interest data.

Maven Analytics Data Playground¶

https://mavenanalytics.io/data-playground — curated datasets designed for practice, each with a business context and questions.

Google Dataset Search¶

https://datasetsearch.research.google.com — a search engine for datasets across the web.

data.world¶

https://data.world — community datasets with collaboration features.

Tips for Choosing a Dataset¶

For a portfolio project, pick a dataset that:

Tells a business story — not just numbers, but something a stakeholder would care about
Is the right size — 1k-1M rows. Too small is trivial; too big needs infrastructure
Has a clear question — you can state "I'm trying to find out X"
Genuinely interests you — your enthusiasm shows in the final work
Isn't overdone — the Titanic and Iris datasets are so common they signal "beginner." A less-common dataset stands out.

Creating Synthetic Data¶

When you can't find the right dataset, generate one:

import pandas as pd
import numpy as np

rng = np.random.default_rng(42)
n = 5000

df = pd.DataFrame({
    "order_id": [f"O{i:05d}" for i in range(n)],
    "order_date": pd.to_datetime("2024-01-01") + pd.to_timedelta(rng.integers(0, 365, n), "D"),
    "customer_id": [f"C{rng.integers(1, 500):04d}" for _ in range(n)],
    "category": rng.choice(["Electronics", "Clothing", "Books", "Home"], n),
    "quantity": rng.integers(1, 6, n),
    "unit_price": rng.uniform(10, 500, n).round(2),
    "status": rng.choice(["completed", "pending", "cancelled"], n, p=[0.7, 0.2, 0.1]),
})
df["total"] = df["quantity"] * df["unit_price"]
df.to_csv("synthetic_orders.csv", index=False)

Synthetic data is perfect for practising techniques when you control exactly what patterns exist.

← Capstones · Back to Home