UberJugaad Enhanced SALT Dataset

Overview

1.9 million ERP transactions + 151,673 business emails + 3,499 supporting documents

Comprehensive enterprise dataset from UberJugaad GmbH, a EUR 14.8B German industrial supplier with 1,467 employees serving 15,606 customers. Enhanced version of the SAP SALT dataset with realistic business communications and documents.

Dataset Structure

UberJugaad Enhanced SALT Dataset/
│
├── 📊 CORE DATA FILES (83.5 MB)
│   ├── all_communications.parquet     [151,673 emails - 6.9 MB]
│   ├── erp_transactions.parquet       [1.9M transactions - 40.8 MB]
│   ├── sales_documents.parquet        [243K headers - 7.1 MB]
│   ├── sales_items.parquet            [1.9M line items - 28.4 MB]
│   ├── supporting_documents.parquet   [3,467 documents - 0.1 MB]
│   └── business_documents.parquet     [32 reports - 22 KB]
│
├── 📧 EMAIL DATABASE (76 MB)
│   └── uberjugaad_email.db           [SQLite: 151K emails + 19K contacts]
│
├── 📖 DOCUMENTATION
│   ├── README.md                      [This file - overview]
│   ├── columns.md                     [Column descriptions]
│   ├── quickstart.md                  [Getting started guide]
│   ├── sample_communications.md       [Email examples]
│   ├── business_docs.md              [Document conversion guide]
│   └── dataset-metadata.json         [Kaggle metadata]
│
├── 💻 CODE EXAMPLES
│   └── uberjugaad-gmbh-notebook-for-data-exploration.ipynb    [Sample analysis code]
│
└── ⚖️ LICENSE                        [MIT License]

Note: All files are in the root directory for easy access. This diagram shows logical groupings.

Dataset Contents

Category	Files	Total Size	Description
Core Data	6 parquet files	83.5 MB	ERP transactions, emails, documents
Email Database	1 SQLite file	76 MB	Complete email system backup
Documentation	6 files	< 1 MB	Guides, examples, metadata
Code	1 Python file	< 1 MB	Starter notebook

Key Features

• Realistic business ecosystem: Complete email threads, documents, and transactions
• Natural language content: 151K emails with realistic subjects and bodies
• Linked data: All files connected via SALESDOCUMENT and customer_id
• Time-synchronized: Chronologically consistent 2019-2020 data
• Discovery-oriented: Patterns embedded in content, not labels

Use Cases

• Email classification and routing
• Sentiment and urgency detection
• Customer behavior analysis
• Document information extraction
• Business process mining
• Anomaly and pattern detection
• Natural language understanding
• Multi-modal analysis (text + structured data)

Quick Start

import pandas as pd

# Load emails
emails = pd.read_parquet('all_communications.parquet')
print(f"Total emails: {len(emails):,}")

# Load transactions
transactions = pd.read_parquet('erp_transactions.parquet')
print(f"Total orders: {len(transactions):,}")

# Find emails mentioning specific orders
order = "0002315309"
order_emails = emails[emails['body'].str.contains(order, na=False)]
print(f"Emails about order {order}: {len(order_emails)}")

# Query email database
import sqlite3
conn = sqlite3.connect('uberjugaad_email.db')
cursor = conn.cursor()
cursor.execute("SELECT COUNT(*) FROM contacts WHERE contact_type = 'external'")
print(f"External contacts: {cursor.fetchone()[0]:,}")
conn.close()

File Descriptions

Core Data Files

• all_communications.parquet: All business emails with subjects, bodies, and metadata
• erp_transactions.parquet: SAP sales transactions with customer and product data
• sales_documents.parquet: Sales document headers from SAP
• sales_items.parquet: Detailed line items for each transaction
• supporting_documents.parquet: Business documents (POs, invoices, shipping notices)
• business_documents.parquet: Meeting agendas and reports

Email Database

• uberjugaad_email.db: SQLite database simulating Exchange backup with emails, contacts, folders

Documentation

• README.md: Dataset overview and structure (this file)
• columns.md: Detailed column descriptions for all files
• quickstart.md: Getting started guide with code examples
• sample_communications.md: Example emails showing variety and realism
• business_docs.md: Guide to converting documents to PDF/HTML
• dataset-metadata.json: Kaggle dataset configuration

Code

• uberjugaad-gmbh-notebook-for-data-exploration.ipynb: Jupyter notebook for comprehensive data exploration and analysis

Author

Patrick Rutledge Enterprise systems specialist and independent data scientist created the fictional UberJugaad GmbH company and enhanced dataset for AI/ML research.

Project Description

This dataset transforms SAP's SALT transactional tables into a full enterprise simulation by layering realistic emails, documents, and embedded patterns. All enhancements are synthetic but modeled on real-world communication and ERP logic.

GitHub: PatrickRutledge/uberjugaad-enhanced-salt Development Year: 2025

---

Citation

If you use this dataset, please cite both the enhanced version and the original SALT dataset:

Enhanced Dataset Rutledge, P. (2025). UberJugaad Enhanced SALT Dataset. Kaggle. https://www.kaggle.com/datasets/PatrickRutledge/uberjugaad-enhanced-salt-dataset

@dataset{rutledge2025uberjugaad,
  title={UberJugaad Enhanced SALT Dataset},
  author={Rutledge, Patrick},
  year={2025},
  publisher={Kaggle},
  url={https://www.kaggle.com/datasets/PatrickRutledge/uberjugaad-enhanced-salt-dataset}
}

Original SALT Dataset (Required) Klein, T., Biehl, C., Costa, M., Sres, A., Kolk, J., & Hoffart, J. (2024). SALT: Sales Autocompletion Linked Business Tables Dataset. NeurIPS 2024 Third Table Representation Learning Workshop.

@inproceedings{klein2024salt,
  title={{SALT}: Sales Autocompletion Linked Business Tables Dataset},
  author={Klein, Tassilo and Biehl, Clemens and Costa, Margarida and Sres, Andre and Kolk, Jonas and Hoffart, Johannes},
  booktitle={NeurIPS 2024 Third Table Representation Learning Workshop},
  year={2024},
  url={https://openreview.net/forum?id=UZbELpkWIr}
}

Citation

If you use this dataset in your research, please cite:

@dataset{rutledge_patrick_2025_17125322,
  author       = {Rutledge, Patrick},
  title        = {{UberJugaad GmbH Enhanced SALT Dataset}},
  month        = sep,
  year         = 2025,
  publisher    = {Zenodo},
  version      = {v1.0.0},
  doi          = {10.5281/zenodo.17125322},
  url          = {https://doi.org/10.5281/zenodo.17125322}
}

APA format:

Rutledge, P. (2025). UberJugaad GmbH Enhanced SALT Dataset (Version v1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17125322

License

MIT License for enhanced content / SAP Sample Code License for original SALT data

Acknowledgments

Built on SAP's SALT dataset. Enhanced with synthetic business communications for AI/ML applications. The original SALT authors are not affiliated with these enhancements.