Why The Right Dataset Matters
A dataset that's too clean teaches you nothing about Power Query. A dataset that's too small doesn't stress-test your model. The best practice datasets are realistically messy and large enough to be genuinely interesting to explore.
Good Sources
- Kaggle — thousands of free public datasets across every topic imaginable, many specifically tagged for BI/dashboard practice
- Microsoft's own sample datasets — the "Financials" and "Contoso" sample datasets are built specifically for Power BI practice and come pre-structured well
- Government open data portals — most countries publish public datasets (census, health, transport) that are free and often genuinely messy in realistic ways
- Your own data — bank statement exports, fitness app exports, or streaming history are all free, real, and personally motivating to explore
Picking A Dataset By Goal
| If you want to practice... | Look for... |
|---|---|
| Power Query / cleaning | Government open data (often messy, inconsistent formatting) |
| Data modeling | Multi-table datasets like Contoso (sales, products, customers as separate tables) |
| DAX | Time-series data with dates, so you can practice year-over-year and running totals |
| Dashboard design | Anything visually rich — sports, retail, or entertainment data tends to be more engaging to design around |
💡
Don't spend too long searching for the "perfect" dataset — any reasonably sized, real dataset is enough to build genuine skill. The learning comes from working with it, not from finding it.