VynFi/vynfi-sap-showcase

VynFi SAP Showcase

An "all capabilities" SAP-style ERP + audit dataset: one end-to-end run that exercises the full DataSynth surface and ships the complete artefact tree — universal-journal postings (ACDOCA), document chains, subledgers, FX, intercompany + consolidation, per-period financial statements with comparatives, operating-expense breakout, and Sankey funds-flow exports, accounting standards, audit (ISA/PCAOB/SOX) + COSO controls, HR, manufacturing, treasury, tax, ESG, project accounting, and ERP-coupled banking (KYC/AML). Sister datasets: `VynFi/vynfi-journal-entries-1m` (GL at scale) and `VynFi/vynfi-audit-p2p` (audit-focused P2P).

Dataset scope

• —4 SAP company codes × 12 fiscal periods, multi-currency (USD / EUR / GBP / SGD) → real FX translation and group consolidation.
• —~648,801 JE lines / ~174,944 journal entries (Custom per-entity volumes).
• —The headline additions over the sister datasets: per-period financial statements (Balance Sheet / Income Statement / Cash Flow / Changes in Equity), standalone and consolidated, with prior-period comparatives, an income-statement operating-expense sub-category breakout, and Sankey funds-flow exports for the income statement and cash-flow statement.

What's inside

A full ERP/audit artefact tree (JSON unless noted):

Financial statements & Sankey

Statements are produced per period per entity plus consolidated. Income- statement and balance-sheet lines carry prior-period comparatives (amount_prior); the income statement breaks operating expenses into salaries / rent / utilities / depreciation / marketing / other (with an IS-OPEX subtotal).

financial_reporting/sankey/flows.json is a list of self-describing d3-sankey / Plotly-compatible flows:

{ "statement_type": "income_statement", "fiscal_year": 2024, "fiscal_period": 3,
  "company_code": "1000", "currency": "USD", "is_consolidated": false,
  "nodes": [{ "id": 0, "label": "Revenue", "value": "1200000", "line_code": "IS-REV" }],
  "links": [{ "source": 0, "target": 1, "value": "480000", "label": "Gross Profit", "is_contra": false }] }

Explore it interactively in the **Financial Sankey Explorer** Space.

Quick start

from datasets import load_dataset
ds = load_dataset("VynFi/vynfi-sap-showcase")        # the journal-entry table
print(ds["train"].column_names)

# the rich artefacts (financial statements, Sankey, banking, …) are files:
from huggingface_hub import hf_hub_download
import json
flows = json.load(open(hf_hub_download(
    "VynFi/vynfi-sap-showcase", "financial_reporting/sankey/flows.json", repo_type="dataset")))
print(len(flows), "sankey flows")

Generation config

configs/examples/hf/sap_showcase.yaml in `mivertowski/SyntheticData`.

datasynth-data validate --config sap_showcase.yaml
datasynth-data generate --config sap_showcase.yaml --output ./sap_showcase_out
python3 scripts/qa_gate.py ./sap_showcase_out          # publish gate

Reproducibility

Privacy

Fully synthetic. Aggregate statistics only; no client identifiers, no PII — the dataset passes the qa_gate.py PII / denylist scan before publication. Company / counterparty names are fictional or nominative.

Citation

@dataset{vynfi_sap_showcase_2026,
  author    = {Ivertowski, Michael and DataSynth contributors},
  title     = {VynFi SAP Showcase — all-capabilities SAP-style ERP + audit dataset},
  year      = {2026},
  publisher = {VynFi / Hugging Face},
  url       = {https://huggingface.co/datasets/VynFi/vynfi-sap-showcase},
}