πͺ Olist Marketplace
Overview
Public dataset from the Brazilian Olist marketplace (Kaggle, 2016β2018): ~96K orders joined across 8 relational tables.
Data Structure
| Table | What It Captures |
|---|---|
| Orders | Order lifecycle: created, approved, delivered, estimated dates |
| Order Items | Products per order: price, freight, seller |
| Sellers | Seller metadata: location, zip code |
| Customers | Customer metadata: location, unique ID |
| Reviews | Customer review scores (1β5) and comments |
| Payments | Payment method, installments, value |
| Products | Product category, dimensions, photos |
| Geolocation | Lat/long for zip codes |
Causal Reasoning Model
Three nested paths converge to explain bad customer reviews (score β€ 2):
1. Logistics Delays: late shipment β delivery past estimate β customer frustration 2. Product Quality Gaps: category mismatch, missing photos, description issues 3. Payment Friction: high installment count, payment method issues
Sample Questions
- "Why did this customer leave a bad review?"
- "How did logistics delays contribute to low satisfaction?"
- "What product quality factors drove 1-star reviews?"
Try It
Ask Gem Logic: Why did this customer leave a bad review?
Start Session →