Where to Find Free Datasets to Practice With (And How to Pick One)

Finding datasets to practice data analysis can seem daunting, but it doesn't have to be. Before you dive into a dataset, make sure it has three key components: a date column, a category column for grouping, and a numeric column for calculations. If a dataset lacks any of these three elements, most analytical techniques will be impractical, regardless of how interesting the topic may be.

Government open data portals are the best source for raw material. Websites like data.gov, data.gov.uk, and the World Bank Open Data provide large collections of real-world, uncleaned data that closely resemble what analysts encounter in their day-to-day work. These portals are messy, often containing inconsistent column names, various date formats, and sometimes renamed categories, requiring cleaning, reconciliation, and documentation skills that are essential for real-world data analysis.

Kaggle is another valuable resource, offering a massive searchable collection of datasets uploaded by users. It's the quickest way to find datasets on specific subjects, thanks to its fast search functionality. However, remember to check for the "Free account needed to download" requirement before downloading.

Other notable sources include Our World in Data, which offers long global time series on various subjects, and FiveThirtyEight, which provides a collection of tidy CSV files behind published stories. The UCI repository hosts a long-standing academic collection of heavily used teaching datasets, many of which are small and already tidy.

When evaluating a dataset, consider the following three-column screen: does it contain a date column, a category column for grouping, and a numeric column for calculations? If it does, you'll likely be able to practice a wide range of analytical techniques. If not, the dataset may be too simplistic to provide meaningful practice opportunities.