IntelliProcure/sustainability_criteria

Sustainability Procurement Criteria

This dataset contains sustainability procurement criteria organized by groups of goods and services (GGS) (German: Waren- und Dienstleistungsgruppen; WDG). It originates from validated Excel files and has been converted to JSONL format for easy consumption.

Groups of Goods and Services

Note: This dataset is currently under active development. Additional groups of goods and services will be added in future releases.

The dataset currently covers the following groups of goods and services (GGS/WDG):

WDG ID	English Name	German Name
WDG_1-Product_Independent_Criteria	Product_Independent_Criteria	Produktuebergreifende_Kriterien
WDG_2-Food	Food	Ernaehrung
WDG_3-Furniture	Furniture	Moebel
WDG_4-Road_Transport	Road_Transport	Strassenverkehr
WDG_5-Print_Services	Print_Services	Druckdienstleistungen
WDG_6-Light_Elect_Appl	Light_Elect_Appl	Beleuchtung_und_elektrische_Geraete
WDG_7-Info_Com_Tech	Info_Com_Tech	Kommunikations- und Informationstechnik (inkl. Medientechnik)-Produkte, Geraete und Dienstleistungen
WDG_8-Construction_Buildings	Construction_Buildings	Bau_Hochbau
WDG_9-Construction_Civil_Engineering_Infrastructure	Construction_Civil_Engineering_and_Infrastructure	Bau_Tiefbau_und_Infrastruktur
WDG_10-Construction_Area	Construction_Area	Bau_Areal
WDG_11-Electricity	Electricity	Strom
WDG_12-IT_Software	IT_Software	IT_Software
WDG_13-Green_Spaces	Green_spaces	Grünräume
WDG_14-Paper_and_Stationery	Paper_and_Stationery	Papier_und_Schreibwaren
WDG_15-Textiles	Textiles	Textilien
WDG_16-Wastepaper_Cleaning_Refuse	Wastepaper_Cleaning_Refuse	Altpapier_Reinigung_Abfall
WDG_17-Gloves_Soap_Bandages	Gloves_Soap_Bandages	Untersuchungshandschuhe_Seife_Verbandsmaterial
WDG_18-Services	Services	Dienstleistungen

Dataset Structure

Format

• Format: JSONL (JSON Lines) - one JSON object per line
• Encoding: UTF-8
• Compression: None

Files

The core dataset is a single merged JSONL file (sustainability_criteria.jsonl) containing criteria from all groups of goods and services (GGS/WDG; Waren- und Dienstgruppen). Each record includes GGS/WDG identifiers (WDG_ID, wdg_name_en, wdg_name_de) to distinguish criteria by category.

Two companion files provide LLM-generated summaries at a coarser grain than individual criteria - see Catalog & Area-of-Action Summaries:

• ggs_summaries.jsonl - one row per GGS/WDG catalog
• aoa_summaries.jsonl - one row per Handlungsfeld (area of action) within a catalog

Each is registered as its own Hugging Face dataset config (ggs_summaries, aoa_summaries), loadable independently of the default config.

Fields

Each JSONL line contains the following fields:

Field	Type	Description
WDG_ID	string	Waren- und Dienstleistungsgruppe identifier
wdg_name_en	string	Waren- und Dienstleistungsgruppe (Goods and Services Group) - English name
wdg_name_de	string	Waren- und Dienstleistungsgruppe (Goods and Services Group) - German name
source_file	string	Original Excel filename (e.g., Food_V1.0.xlsx)
Handlungsfeld-ID	string	Area of Action identifier (WDG-specific)
Handlungsfeld	string	Area of Action name (German) (WDG-specific)
Kriterium-ID	string	Criterion identifier (WDG-specific)
Kategorie Kriterium	string	Criterion category (EK, TS, ZK, TB)
Ausschreibungskriterium	string	Procurement criterion description
Ambitionsniveau: Basis	string	Basic ambition level
Ambitionsniveau: Gute Praxis	string	Good practice ambition level
Ambitionsniveau: Vorbild	string	Best practice ambition level
Nachweise	string	Evidence/documentation requirements
Nachhaltigkeitsdimensionen	string	Sustainability dimensions
Quelle	string	Source reference identifiers
Kommentar	list	Additional comments and notes

Notes

• Empty fields: Fields with no value are represented as empty strings
• Forward-filled hierarchical data: Handlungsfeld-ID and Handlungsfeld values are propagated downward within each file
• Source references: Use the Quelle field to look up source details in the corresponding *_metadata.json file

WDG-Specific Fields

Fields marked as (WDG-specific) are unique to each Waren- und Dienstleistungsgruppe (goods and services group). These fields may vary across different WDGs. All other fields are standardized across all WDGs.

Catalog & Area-of-Action Summaries

In addition to the criteria themselves, two companion files provide concise, sector-focused LLM-generated summaries at a coarser grain: per catalog, and per area of action within a catalog. Both are grounded in the criteria text of the items they summarize, are written in German, and deliberately favor domain/sector terminology over generic sustainability language. Coverage may be partial - not every catalog or area of action necessarily has a summary yet.

ggs_summaries.jsonl - one row per catalog (GGS/WDG)

Field	Type	Description
WDG_ID	string	Waren- und Dienstleistungsgruppe identifier; joins to sustainability_criteria.jsonl on WDG_ID
GGS_short_description	string	~50-100 word German description of what the catalog covers in practice
GGS_signature_terms	list	Characteristic domain/sector terms for the catalog
typical_project_types	list	Typical projects or services in this catalog
excluded_generic_terms	list	Generic/sustainability terms deliberately avoided for this catalog

aoa_summaries.jsonl - one row per Handlungsfeld (area of action)

Field	Type	Description
WDG_ID	string	Waren- und Dienstleistungsgruppe identifier
Handlungsfeld-ID	string	Area of Action identifier; joins to sustainability_criteria.jsonl on (WDG_ID, Handlungsfeld-ID)
handlungsfeld_short_description	string	~50-100 word German description, grounded in that area's criteria, including an explicit Geltungsbereich (scope) and Nicht gemeint (out-of-scope) boundary
signature_terms	list	Characteristic terms for the area of action
typical_activities	list	Typical activities or verification steps for the area of action

Loading and joining the summaries

from datasets import load_dataset

criteria = load_dataset("IntelliProcure/sustainability_criteria")["train"]
ggs_summaries = load_dataset("IntelliProcure/sustainability_criteria", "ggs_summaries")["train"]
aoa_summaries = load_dataset("IntelliProcure/sustainability_criteria", "aoa_summaries")["train"]

# Combine with pandas: attach catalog- and area-of-action-level context to every criterion
df = criteria.to_pandas()
df = df.merge(ggs_summaries.to_pandas(), on="WDG_ID", how="left")
df = df.merge(aoa_summaries.to_pandas(), on=["WDG_ID", "Handlungsfeld-ID"], how="left")

Data Description

Criterion Categories

• EK (Eignungskriterium): Selection criterion
• TS (Technische Spezifikation): Technical specification
• ZK (Zuschlagskriterium): Award criterion
• TB (Zwingende Teilnahmebedingung): Mandatory participation condition

Sustainability Dimensions

• ökologisch: Environmental/Ecological
• sozial: Social
• ökonomisch: Economic

Ambition Levels

• Basis: Basic level
• Gute Praxis: Good practice
• Vorbild: Best practice / exemplary

Usage Examples

Load with Hugging Face Datasets Library

Basic Loading

from datasets import load_dataset

# Load the entire dataset
catalog = load_dataset("IntelliProcure/sustainability_criteria")

# Access the train split
criteria = catalog['train']
print(f"Total criteria: {len(criteria)}")
print(f"Columns: {criteria.column_names}")

Working with the Dataset

from datasets import load_dataset

catalog = load_dataset("IntelliProcure/sustainability_criteria")
criteria = catalog['train']

# Convert to pandas DataFrame
df = criteria.to_pandas()

# Access specific records
first_record = criteria[0]
print(first_record['Ausschreibungskriterium'])

# Get multiple records
first_ten = criteria[:10]

# Filter criteria by category
selection_criteria = criteria.filter(lambda x: x['Kategorie Kriterium'] == 'EK')

# Filter by sustainability dimension
ecological = criteria.filter(
    lambda x: 'ökologisch' in x['Nachhaltigkeitsdimensionen']
)

# Get records from specific action field
food_criteria = criteria.filter(lambda x: x['Handlungsfeld'] == 'Food')

Advanced Filtering and Analysis

from datasets import load_dataset
import pandas as pd

catalog = load_dataset("IntelliProcure/sustainability_criteria")
df = catalog['train'].to_pandas()

# Find all criteria by multiple dimensions
multi_dim = df[
    df['Nachhaltigkeitsdimensionen'].str.contains('ökologisch|sozial', na=False)
]

# Group by action field
by_field = df.groupby('Handlungsfeld').size()
print(by_field)

# Get statistics on ambition levels
print(df[['Ambitionsniveau: Basis', 'Ambitionsniveau: Gute Praxis']].notna().sum())

# Find criteria with all three ambition levels
complete = df[
    (df['Ambitionsniveau: Basis'] != '') &
    (df['Ambitionsniveau: Gute Praxis'] != '') &
    (df['Ambitionsniveau: Vorbild'] != '')
]

Stream Large Datasets

from datasets import load_dataset

# Stream data without downloading entirely (useful for large datasets)
catalog = load_dataset("IntelliProcure/sustainability_criteria", streaming=True)
criteria_stream = catalog['train']

# Iterate through records
for i, record in enumerate(criteria_stream):
    if i >= 100:  # Process first 100
        break
    print(record['Kriterium-ID'], record['Ausschreibungskriterium'])

Access Dataset Information

from datasets import load_dataset

catalog = load_dataset("IntelliProcure/sustainability_criteria")
criteria = catalog['train']

# Dataset info
print(criteria.info)
print(criteria.features)
print(f"Number of records: {len(criteria)}")

# Column names and types
print(criteria.column_names)
for feature_name, feature_type in criteria.features.items():
    print(f"  {feature_name}: {feature_type}")

# Get unique values
print(f"Unique areas of action: {criteria.unique('Handlungsfeld')}")
print(f"Unique categories: {criteria.unique('Kategorie Kriterium')}")

Load with Pandas

import pandas as pd

# Load the merged JSONL file
df = pd.read_json('sustainability_criteria.jsonl', lines=True)

# Filter by WDG
food_criteria = df[df['wdg_name_en'] == 'Food']

# Filter by category
procurement_criteria = df[df['Kategorie Kriterium'] == 'EK']

# Filter by sustainability dimension
ecological = df[df['Nachhaltigkeitsdimensionen'].str.contains('ökologisch', na=False)]

Load with Json Module

import json

with open('sustainability_criteria.jsonl', 'r', encoding='utf-8') as f:
    for line in f:
        record = json.loads(line)
        print(record['Ausschreibungskriterium'])

Resolve Source References

import json

# Load metadata to resolve source references
with open('sustainability_criteria_metadata.json', 'r', encoding='utf-8') as f:
    metadata = json.load(f)

sources = metadata['sources']

# Example: resolve a source reference
quelle_value = "Q-1, Q-2"
source_ids = [s.strip() for s in quelle_value.split(',')]
for sid in source_ids:
    if sid in sources:
        print(f"{sid}: {sources[sid]}")

Filter by WDG

from datasets import load_dataset

catalog = load_dataset("IntelliProcure/sustainability_criteria")
criteria = catalog['train']

# Filter by English WDG name
food_criteria = criteria.filter(lambda x: x['wdg_name_en'] == 'Food')

# Filter by German WDG name
food_criteria_de = criteria.filter(lambda x: x['wdg_name_de'] == 'Lebensmittel')

# Filter by source file
food_v1 = criteria.filter(lambda x: x['source_file'] == 'Food_V1.0')

# Get unique WDGs
df = criteria.to_pandas()
print(df['wdg_name_en'].unique())
print(df['wdg_name_de'].unique())

Sources

Each criterion includes references to source documents. Source details are provided in the metadata files.

Common sources include:

• EU GPP (Green Public Procurement) criteria
• German environmental labels and standards
• Industry-specific guidelines
• Sustainability certifications

License

CC-BY-4.0

Citation

If you use this dataset, please cite:

@dataset{sustainability_criteria,
    author       = { Tilia Renate Ellendorff and Luca Sven Rolshoven and Veton Matoshi and Jann Jeremy Austin and Judith Binder },
    title        = {{Sustainable Procurement Criteria}},
    year         = 2026,
    url          = { https://huggingface.co/datasets/IntelliProcure/sustainability_criteria },
    doi          = { 10.57967/hf/7958 },
    publisher    = { Hugging Face }
}

---

Last Updated: 2026-09-18 Dataset Version: 1.0 Format Version: 1.0