Sample Database Resources: 12 Free Datasets for SQL Practice

TL;DR: A sample database is a free, ready-made schema with realistic data you can load in minutes to practice SQL, test a tool or demo a dashboard. Start with DVD Rental or Pagila for PostgreSQL, Sakila or classicmodels for MySQL, AdventureWorks or Northwind for SQL Server, and Chinook if you want one dataset that runs on every engine.

A sample database is a pre-built database, schema plus data, published so people can learn and test SQL without inventing their own tables. The most used are Pagila and DVD Rental (PostgreSQL), Sakila, Employees and classicmodels (MySQL), AdventureWorks, Northwind and Wide World Importers (SQL Server), and Chinook, which ships scripts for SQL Server, Oracle, MySQL, PostgreSQL, SQLite and DB2. Most load with one command (psql -f, mysql <, pg_restore or a SQL Server .bak restore). Below, each one is listed with its engine, its shape, what questions it is good for and the exact load command, grouped by database.

Where Can I Download a Sample PostgreSQL Database?

The fastest PostgreSQL sources are the DVD Rental tutorial file, the Pagila repository and Neon's collection of ready dumps.

• DVD Rental on PostgreSQL Tutorial: 15 tables (film, actor, customer, rental, payment) plus a printable ER diagram. Free. The best first database for a beginner because the schema fits on one page.
• Pagila: the PostgreSQL port of MySQL's Sakila, a DVD rental store with stores, staff, customers and payments. Free (open license). Worth it once DVD Rental feels small: it adds a JSONB schema variant (pagila-schema-jsonb.sql) and partitioned payment tables, so you practice real PostgreSQL features.
• Neon postgres-sample-dbs: one repository with Chinook, DVD Rental, Pagila, Employees, Lego, Netflix, Titanic, World Happiness and more, dumped from Postgres 15. Free. The quickest way to try several datasets on any Postgres host; the largest is 333 MB.

Load Pagila:

createdb pagila
psql -d pagila -f pagila-schema.sql
psql -d pagila -f pagila-data.sql

Load DVD Rental (it ships as a .tar archive, so it needs pg_restore, not psql):

createdb dvdrental
pg_restore -U postgres -d dvdrental dvdrental.tar

A Neon dump gotcha the tutorials skip: those files use COPY ... FROM stdin, which only works through psql -d "<connection string>" -f file.sql. Pasting them into a GUI SQL editor fails with "No source stream defined".

MySQL Sample Database Options: Sakila, Employees, classicmodels

MySQL's own documentation hosts two official samples, and MySQL Tutorial hosts a third that is the best for business reporting.

• Sakila: Oracle's official MySQL sample, a DVD rental business with films, actors, inventory, rentals and payments. Free. Good for joins across many-to-many tables (film to actor, film to category).
• Employees (test_db): about 300,000 employee records and 2.8 million salary rows in a 167 MB export. Free. The one to use when you need volume: indexes, EXPLAIN and date-range queries only show their effect on real row counts. Runs on MySQL 5.0+ and PostgreSQL 12+.
• classicmodels on MySQL Tutorial: a retailer of scale model cars with customers, products, product lines, orders, order details, payments, employees and offices. Free. The closest MySQL sample to a real sales database: revenue by month, top customers and sales rep performance all work out of the box.

Load Sakila:

mysql -u root -p < sakila-schema.sql
mysql -u root -p < sakila-data.sql

Load Employees on MySQL 9.5 or later (the dump uses SOURCE, which now needs the --commands client flag), then verify the checksums:

mysql --commands < employees.sql
mysql -t < test_employees_sha2.sql

Which SQL Server Samples Does Microsoft Publish?

Microsoft publishes three samples, from small to enterprise-shaped.

• AdventureWorks: a bicycle manufacturer with sales, production, purchasing and HR schemas. Free. Ships as .bak files per SQL Server version (AdventureWorks2025, 2022, 2019 and older) in three variants: OLTP (the full transactional database), DW (a star-schema data warehouse) and LT (a lightweight cut). Start with the OLTP file matching your SQL Server version; use DW to learn dimensional modeling.
• Northwind and pubs: the classic trading company (customers, orders, products, suppliers, employees) as instnwnd.sql, with a separate script for Azure SQL Database. Free. Small and familiar: half the SQL tutorials on the web use it.
• Wide World Importers: a wholesale novelty goods importer with purchase orders, supplier payments, stock and customer orders. Free. Use it to practice newer SQL Server features such as temporal tables and JSON.

Restore AdventureWorks (adjust the path to your data directory):

RESTORE DATABASE AdventureWorks2022
FROM DISK = '/var/opt/mssql/backup/AdventureWorks2022.bak'
WITH MOVE 'AdventureWorks2022' TO '/var/opt/mssql/data/AdventureWorks2022.mdf',
     MOVE 'AdventureWorks2022_log' TO '/var/opt/mssql/data/AdventureWorks2022_log.ldf';

Which Sample Database Works on Every Engine?

• Chinook: a digital media store (artists, albums, tracks, playlists, invoices, customers) with scripts for SQL Server, Oracle, MySQL, PostgreSQL, SQLite and DB2. Free. Built as a cross-engine alternative to Northwind, so the same practice queries run everywhere; the repository now also links a Netflix sample database.

Load Chinook into SQLite, the zero-install option:

sqlite3 chinook.db < Chinook_Sqlite.sql
sqlite3 chinook.db "SELECT COUNT(*) FROM Invoice;"

Which Sample Database Is Best for Practicing SQL Joins?

Pagila or Sakila is the best sample database for practicing joins, because it has every relationship type: one-to-many (customer to rental), many-to-many through junction tables (film_actor, film_category) and a self-referencing chain (rental to inventory to film to store). Chinook is a close second with a clean five-table path from customer to track. Northwind is fine for basic inner and left joins but has few many-to-many cases.

A starter join on Pagila that touches four tables:

-- Top 5 film categories by rental revenue
SELECT c.name AS category, SUM(p.amount) AS revenue
FROM payment p
JOIN rental r         ON r.rental_id = p.rental_id
JOIN inventory i      ON i.inventory_id = r.inventory_id
JOIN film_category fc ON fc.film_id = i.film_id
JOIN category c       ON c.category_id = fc.category_id
GROUP BY c.name
ORDER BY revenue DESC
LIMIT 5;

For join types explained with rows, see our JOIN glossary entry and selecting from multiple tables without JOIN.

Is There a Sample SaaS or Ecommerce Database?

Yes, though fewer of them are official. These are the most realistic for business questions:

• Brazilian E-Commerce Public Dataset by Olist: real anonymized marketplace orders with customers, sellers, payments, reviews and delivery dates, as CSV files. Free (Kaggle account). The best ecommerce sample for delivery time, repeat purchase and review score analysis.
• classicmodels and AdventureWorks: order and sales schemas for B2B and B2C revenue reports.
• TPC-H: the standard decision-support benchmark schema (customers, orders, line items, suppliers, parts, nations) with a generator, dbgen, that builds any size by scale factor. Free. Use it when you need millions of order rows for performance tests.
• Supabase example projects: app-style schemas (auth users, profiles, subscriptions) close to what a real SaaS database looks like. Free.

There is no widely adopted official SaaS sample with subscriptions, plans and MRR. The usual workaround is to generate one: create accounts, plans, subscriptions and invoices tables and fill them with generate_series in PostgreSQL. Our SaaS metrics glossary lists the metrics such a schema should support.

How Do I Load a Sample Database?

Every sample above loads one of four ways, depending on the file type:

Then run a count on the largest table to confirm the data arrived, for example SELECT COUNT(*) FROM salaries; on Employees should return about 2.8 million.

Start Here: A Path for Beginners

1. Week 1: DVD Rental on PostgreSQL. Learn SELECT, WHERE, GROUP BY and two-table joins against a schema small enough to memorize. Keep our SQL cheat sheet open.
2. Week 2: Pagila. Practice many-to-many joins, window functions and JSONB.
3. Week 3: Employees. Same queries at 2.8 million rows; add indexes and read EXPLAIN plans.
4. Week 4: AdventureWorksDW or TPC-H. Star schemas and reporting queries, the shape analytics databases use. Our guide to normalized databases explains why the OLTP and DW versions look so different.

Once your sample is loaded, a useful check on your own answers is to ask the same question in plain English and compare. AI for Database connects to any of these databases (and includes its own SaaS-style sample data at signup), writes the SQL for a question like "top 5 categories by rental revenue" and shows the query, so you can diff it against yours. Try it free.