The Brazilian e-commerce public dataset contains online retail records from Olist. Use it to begin analysis by loading its tables and checking whether the data is structurally trustworthy. section 2.3
Start with a practical data-quality inspection:
- Load customer, order, item, payment, review, product, seller, and category-translation tables into pandas DataFrames. section 2.3
- Inspect previews, shapes, column types, descriptive statistics, missing values, and unique-value counts. section 2.3
- Treat missing-value counts cautiously because sparse fields may be complete by design. section 2.3
A suspicious customer-order relationship
Equal counts of customers and orders could mean one purchase per customer or an incomplete dataset. Compare customer_id with customer_unique_id: 99,441 customer IDs correspond to 96,096 unique customer IDs, with some unique customers linked to as many as 17 IDs. section 2.3
