Safe joins
The model starts at deduplicated orders. Items, payments, and reviews are aggregated before order-level analysis to prevent join fan-out.
Olist data workspace
This project starts with your actual source files. It does not seed, simulate, or fabricate marketplace results. Drop the original CSV files together to create the cleaned, order-anchored analytics model.
Use the provided Olist datasets only. The uploader recognizes the standard Olist filenames and checks that every required source is present.
The model starts at deduplicated orders. Items, payments, and reviews are aggregated before order-level analysis to prevent join fan-out.
Raw geolocation records are aggregated by ZIP prefix before use, preserving the grain of commerce facts.
Every KPI and insight is calculated only after your source records load. Observed relationships are labeled as associations.