The 7 Best Sample Databases for Learning SQL in 2026 (and Their Limitations)

The Classic Problem

Every SQL tutorial starts with a sample database. But most sample databases were designed decades ago for a different era of software development. They're small, they lack realistic business complexity, and they don't reflect modern data engineering practices.

Here's an honest comparison of the most popular options — and where the field is heading.

---

1. AdventureWorks (Microsoft)

Released: 2005 (updated through 2019)

Tables: ~70 (many unused)

Rows: ~120K across all tables

Format: SQL Server backup (.bak)

Strengths

• Well-documented with extensive tutorials
• Covers manufacturing, sales, and HR domains
• Included with SQL Server installations

Limitations

• SQL Server only — not portable to PostgreSQL, MySQL, or SQLite without conversion
• Outdated schema — designed for SQL Server 2005 patterns
• No accounting — no double-entry bookkeeping, no journal entries, no trial balance
• No tax compliance — no GST, VAT, PAYG, or payroll tax calculations
• Complex but shallow — many tables exist but with minimal data
• Not deterministic — no way to regenerate the same data

"AdventureWorks is the default recommendation, but it's a 20-year-old database designed to showcase SQL Server features, not to teach realistic business data modeling." — Common criticism in data engineering communities

---

2. Northwind (Microsoft)

Released: 1997

Tables: 13

Rows: ~3K

Format: SQL Server / Access (.mdb)

Strengths

• Simple and easy to understand
• Great for absolute beginners
• Covers orders, products, customers, suppliers

Limitations

• Extremely small — too little data for meaningful analytics
• No financial data — no invoices, payments, or accounting
• Single currency, single country — no multi-jurisdiction complexity
• No payroll or HR — no employees beyond basic contact info
• Frozen in 1997 — product catalog feels dated

---

3. Chinook (open source)

Released: 2008

Tables: 11

Rows: ~15K

Format: Multiple (SQLite, PostgreSQL, MySQL, SQL Server)

Strengths

• Multi-platform — available for all major databases
• Media/music domain (tracks, albums, artists, playlists)
• Good for teaching joins and relationships

Limitations

• Niche domain — media store, not a general business
• Small — 11 tables is not enough for complex queries
• No financial depth — invoices exist but no accounting behind them
• No temporal complexity — no time-series patterns to analyze

---

4. Sakila (MySQL)

Released: 2005

Tables: 16

Rows: ~47K

Format: MySQL

Strengths

• DVD rental domain — intuitive and fun
• Good for practicing joins, subqueries, and aggregations
• Well-structured with clear relationships

Limitations

• MySQL only — designed for MySQL-specific features
• Obsolete domain — DVD rentals in 2026?
• No business operations — no purchasing, no payroll, no inventory management
• Limited scale — not suitable for data engineering or BI workloads

---

5. TPC-H / TPC-DS (Transaction Processing Council)

Released: 1999 / 2006

Tables: 8 (TPC-H) / 25 (TPC-DS)

Rows: Scalable (1GB to 100TB)

Format: Generator tool

Strengths

• Industry standard for benchmarking
• Scalable — generate any size dataset
• Well-defined queries — comes with standard benchmark queries
• Complex star/snowflake schemas (TPC-DS)

Limitations

• Designed for benchmarking, not learning — schema is abstract and unintuitive
• No business realism — table names like LINEITEM, PARTSUPP don't map to real business concepts
• No accounting or compliance — purely transactional
• Difficult to set up — requires compilation and configuration

---

6. PostgreSQL Sample Databases (dvdrental, pagila)

Released: Various

Tables: 15-16

Rows: ~46K

Format: PostgreSQL

Strengths

• Native PostgreSQL format
• Active community maintenance
• Good documentation

Limitations

• Same domain limitations as Sakila (DVD rentals)
• Small scale
• No financial or operational depth

---

7. Mindweave SME-Sim Datasets (2026)

Released: 2026

Tables: 42

Rows: 39K - 259K per company (up to 825K in bundles)

Format: CSV, PostgreSQL SQL, Apache Parquet, SQLite

Strengths

• 42 tables, 44 foreign keys — full end-to-end business operations
• Double-entry accounting — debits always equal credits
• Real tax compliance — ATO (AU), IRS (US), HMRC (UK) actual brackets
• 3 countries × 3 industries — genuinely different business patterns
• 4 formats — use with any database or analytics tool
• Deterministic — same seed = identical data every time
• Multi-company bundles — test group reporting and consolidation
• Time-series rich — 730+ days of day-by-day simulation

Limitations

• Not free (full datasets) — $19-$199 depending on product, but free samples available
• SME focused — small/medium business, not enterprise or manufacturing
• Three countries — AU, US, UK only (more planned)

---

Comparison Table

Database	Tables	Rows	Accounting	Tax	Multi-format	Deterministic	Free
AdventureWorks	~70	~120K	No	No	SQL Server	No	Yes
Northwind	13	~3K	No	No	SQL Server	No	Yes
Chinook	11	~15K	No	No	Yes	No	Yes
Sakila	16	~47K	No	No	MySQL	No	Yes
TPC-H	8	Scalable	No	No	Generator	Yes	Yes
TPC-DS	25	Scalable	No	No	Generator	Yes	Yes
SME-Sim	42	39K-825K	Yes	Yes	Yes (4)	Yes	Samples

---

Which Should You Use?

Learning basic SQL joins: Chinook or Sakila — small, simple, well-documented.

Learning business data modeling: SME-Sim — 42 tables with real business relationships and accounting.

Benchmarking database performance: TPC-H or TPC-DS — industry standard, scalable.

Testing ERP or accounting software: SME-Sim — the only option with double-entry accounting and real tax compliance.

Building BI dashboards: SME-Sim — 730+ days of time-series data across 7 business domains.

Data engineering pipelines: SME-Sim (Parquet format) or TPC-DS — both offer structured, scalable data.

---

Try It

Free samples are available on multiple platforms:

• GitHub (AU) | GitHub (US) | GitHub (UK)
• Kaggle (AU) | Kaggle (US) | Kaggle (UK)
• Hugging Face (AU) | Hugging Face (US) | Hugging Face (UK)

Browse full catalog →

---

*This comparison is based on publicly available documentation for each database as of April 2026. AdventureWorks and Northwind are trademarks of Microsoft Corporation. TPC-H and TPC-DS are trademarks of the Transaction Processing Performance Council.*