toolathon123/south-america-logistics-quarterly

South America Logistics — Quarterly Dataset

South America regional road / rail / city-delivery shipments, aggregated and published once per quarter. Covers Brazil (BR) and Argentina (AR).

Current quarter: 2026-Q1 · Rows: 13,000 · Format: Parquet (snappy)

---

Coverage

Region	Country	Description
BR	Brazil	Highway (road), rail and urban city-delivery legs, primarily from São Paulo (SP) hub
AR	Argentina	Corridors coordinated with the Buenos Aires (CABA) partner team, plus cross-border BR↔AR

• Time window: shipments from 2026-01-01 to 2026-03-31 (Q1 2026).
• The dataset is sharded per country under data/:
  • data/BR-2026-Q1.parquet (8,900 rows)
  • data/AR-2026-Q1.parquet (4,100 rows)

Quick start

import pandas as pd

df = pd.read_parquet("hf://datasets/toolathon123/south-america-logistics-quarterly/data/BR-2026-Q1.parquet")
print(df.shape)

or with the datasets library:

from datasets import load_dataset

ds = load_dataset("toolathon123/south-america-logistics-quarterly")
print(ds["train"].num_rows)

Fields

Field	Type	Description
order_id	string	Unique shipment order identifier (primary key)
country	string	ISO-2 country: BR (Brazil) or AR (Argentina)
region	string	Derived region tag, BR or AR (matches country)
origin_city	string	Origin city
destination_city	string	Destination city
shipment_date	date	Date the shipment left the origin
delivery_date	date	Actual/expected delivery date
weight_kg	float	Cargo weight (kg)
volume_m3	float	Cargo volume (m³)
carrier_name	string	Transport carrier name
vehicle_plate	string	Vehicle licence plate (Mercosur format)
distance_km	float	Route distance (km)
fuel_liters	float	Estimated/measured fuel consumption (L)
status	string	delivered, in_transit, delayed, pending, cancelled
delay_hours	float	Delay in hours relative to plan (0 = on time)

Data sources

• PostgreSQL database — core shipments table export (Brazil hub + Argentina partner DB).
• Spreadsheet — manual cross-border / carrier reconciliation sheets maintained by the Buenos Aires team.

Data-quality pipeline (per quarter, automated)

1. Merge PostgreSQL export + spreadsheet export.
2. Standardise casing/whitespace, coerce numeric and date types.
3. Derive region from country.
4. Remove duplicate order_id (keep the most complete record).
5. Impute missing key fields (status → delivered, delay_hours → 0, weight_kg/volume_m3/fuel_liters → carrier/country medians or distance-based estimates).
6. Validate (no nulls in key fields, unique order_id, region↔country consistency, date order).
7. Export compressed Parquet shards per country (snappy).

Update frequency

• Quarterly. New Parquet shards replace the previous quarter's files under data/ (file name pattern {COUNTRY}-{YYYY-Qn}.parquet).
• Only the data/ files and this README.md need to be replaced/updated each quarter; the pipeline is fully scripted (scripts/generate_data.py, scripts/prepare_dataset.py, scripts/upload_dataset.py).

License

CC BY 4.0. Internal analytics dataset — share with permission of the São Paulo analytics team and the Buenos Aires partner.