This sample is published by PrimaForja LLC under the Creative Commons Attribution-NonCommercial 4.0 license (CC BY-NC 4.0). By requesting access you agree that: (1) you will use the data for non-commercial purposes only — commercial licenses for the full dataset are available at https://www.primaforja.com; (2) you will credit "PrimaForja LLC" in any publication, model, or derivative that uses it; (3) no trademark license is granted — the PrimaForja name and marks may be used only for that required attribution; and (4) the data is entirely synthetic and provided as-is, without warranty of any kind. Access is granted automatically once you accept.

PrimaForja Supply — One-Month Sample

A fully synthetic, fully reconciled month of a mid-market industrial distributor: ~40 million rows across 62 interrelated tables, generated at full production scale (27 distribution centers, ~6,800 customers, ~37,000 SKUs, 391 vendors, ~1.7 million orders).

What this is

A free sample of a production-scale synthetic dataset that models a US industrial/office-supply distributor doing roughly $12 million per day in revenue. Those numbers are the generator's default configuration, not a limit — the business scales up or down (rows, revenue, date range) with every table still agreeing with every other table. In its default shape the full product is two years of history (~800 million rows); this sample is one complete calendar month (June 2025), generated by the same pipeline at the same scale — every foreign key resolves and every financial number reconciles, at a size you can explore on a laptop.

Not one row of this is real data. No PII, no scrubbed production data, no licensing entanglements. It is generated by a deterministic, config-driven pipeline (3,300+ tunable parameters) in which the business logic interlocks: facility size drives headcount, headcount drives payroll, payroll rolls into EBITDA; discounts, promotions, freight, and cost flow into a pocket-margin waterfall that reconciles to the dollar.

The month has honest edges, the way a real month does: orders sit strictly inside June, but shipments and invoices for late-June orders spill into July, and June carries backorders and receivables opened earlier.

Using the sample

Each table is one Parquet file — use the dataset viewer above (pick a table from the config dropdown) or load any table directly:

import pandas as pd

orders = pd.read_parquet("hf://datasets/PrimaForja/supply-one-month-sample/orders.parquet")

DuckDB makes the joins effortless:

import duckdb

con = duckdb.connect()
df = con.sql("""
    SELECT c.segment, COUNT(DISTINCT o.order_id) AS orders, SUM(ol.extended_price) AS revenue
    FROM 'hf://datasets/PrimaForja/supply-one-month-sample/order_lines.parquet' ol
    JOIN 'hf://datasets/PrimaForja/supply-one-month-sample/orders.parquet' o USING (order_id)
    JOIN 'hf://datasets/PrimaForja/supply-one-month-sample/customers.parquet' c USING (customer_id)
    GROUP BY 1 ORDER BY 3 DESC
""").df()

data_dictionary.md / data_dictionary.json document every table and column; manifest.json records the generation run (seed 20250 — deterministic: the same seed regenerates this sample byte-for-byte).

What's inside

Orders, order lines, shipments, shipment lines, tracking events, invoices, AR/AP ledgers, returns, backorders, purchase orders, inventory (positions, monthly snapshots, adjustments, slotting), item catalog with costs and price history, vendors with performance and scorecards, carriers, promotions and campaigns, labor/facility/fleet costs, a margin waterfall, and a C-level analytics layer (demand signals, forecast accuracy, customer LTV, price elasticity, DC throughput, rep performance).

Suggested ML tasks: demand forecasting, churn/LTV modeling, price elasticity estimation, backorder risk classification, carrier OTD prediction, anomaly detection on inventory adjustments.

The full product

The complete dataset — two years, ~800 million rows, plus a licensed library of 77 analytics views that power the live executive dashboards — is at https://www.primaforja.com, where you can explore the dashboards built on it and request demo access.

License

CC BY-NC 4.0 — © PrimaForja LLC. Free for research, learning, and evaluation with attribution; commercial use (including resale or bundling into paid products) requires a license — contact sales@primaforja.com.

Downloads last month: 10

Total file size:

837 MB