The Olist e-commerce dataset

The Olist dataset is a real Brazilian e-commerce dataset, ~100K orders across 8 normalized tables: customers, orders, order items, payments, sellers, products, reviews, and a category translation. It is the right size: small enough to model on a laptop, big enough to expose every star schema decision you will face.

The dataset has the shapes that matter for analytics. Multiple fact-table candidates (orders, order items, payments). Multiple dimensions with slowly-changing-attribute potential (sellers, products). Real foreign-key relationships you will join. And a translation table that proves you cannot escape lookup tables.