Inside Earthmark’s Data Engine: Powering Corporate Sustainability Intelligence

Sustainability performance can no longer be assessed using surface-level scores or static annual surveys. Modern financial institutions, enterprises, and conscious consumers require granular, audit-ready data extracted directly from official corporate disclosures, verified third-party targets, and historical trajectory models.

As Technical Lead at Earthmark, my core mission is ensuring our data pipeline delivers accuracy, speed, and coverage. 

To demonstrate the scope of Earthmark’s proprietary IP, we have recently conducted an audit of the data we have automatically extracted from ESG disclosures.

This audit was conducted both by human experts and AI agents.

1. Scale Overview: Millions of Datapoints Across Global Enterprise

Earthmark’s database captures company attributes, financial indicators, historical snapshots, external certifications, and granular emission extractions.

2. Sector Distribution: Comprehensive Coverage Across 11 GICS Industries

Corporate impact varies drastically depending on industry classification. Earthmark categorizes companies across 11 primary GICS sectors and 20+ NAICS sub-sector divisions.

Companies per GICS Sector

• Consumer Discretionary: 8K

• Industrials: 6K

• Consumer Staples: 2.8K

• Information Technology: 2.6K

• Health Care: 1.5K

• Financials: 1.4K

• Materials: 1.4K

• Communication Services: 1K

• Real Estate: 0.7K

• Utilities: 0.4K

• Energy: 0.2K

NAICS Manufacturing & Service Breakdown

In addition to macro GICS groupings, Earthmark maintains a mapping for granular industry analysis:

• Metals, Machinery, Electronics & Transport Manufacturing: 4K companies

• Nonstore & Specialty Retail Trade: 3K companies

• Chemical, Paper & Plastics Manufacturing: 2.7K companies

• Food, Textile & Apparel Manufacturing: 2.5K companies

• Professional, Scientific & Technical Services: 2.4K companies

3. Longitudinal Depth: 25+ Years of Historical Trend Analysis

Understanding a company’s decarbonization trajectory requires multi-year baseline tracking. Earthmark's database spans historical data from 2000 through 2026, with primary report density covering 2018–2026.

By storing 23.5K annual disclosure data, Earthmark enables deterministic trend analysis across:

• Scope 1 Emissions (Direct fuel & operational greenhouse gases)

• Scope 2 Emissions (Location-based & market-based electricity grid factors)

• Scope 3 Emissions (Supply chain, upstream, downstream, and travel)

• Total Energy Consumption & Renewable Energy Share (%)

• Total Waste Generated, Recycled, and Disposed

• Revenue Correlation & Intensity Metrics (tCO₂e per $M revenue)

4. Third-Party Cross-Validation: 19,000+ records

Earthmark does not rely solely on self-reported corporate disclosures. Our pipeline continuously reconciles company profiles against authoritative third-party datasets:

1. Science Based Targets initiative (SBTi): 13.8K companies are mapped directly to verified 1.5°C and well-below 2°C science-based climate targets, near-term commitments, and net-zero target dates.

2. B Corp Certification: 0.8K companies enriched with verified B Corp impact assessment scores.

3. Good Shopping Guide: 0.5K companies cross-referenced with ethical shopping and consumer transparency metrics.

4. Financial Market Data (Yahoo Finance): 4K public companies enriched with market capitalization, financial volume, and revenue metrics.

5. Pro Search: Earthmark’s Proprietary AI Extraction Engine

To extract ESG datatpoints from 100+ pages unstructured disclosures including plots and images, we engineered Pro Search—Earthmark’s proprietary AI document processing engine.

Architecture & Capabilities

• Sub-15 Second Report Processing: Automatically downloads, extracts, normalizes, and structures data from complex PDF impact disclosures in an average of <15 seconds per report for bulk processing.

• Continuous Human-in-the-Loop Review: Extractions are regularly audited against source text for continuous alignment.

Evaluated Benchmark Performance

On our latest automated testing evaluation suite, Pro Search achieved 97% extraction accuracy on complex PDFs, demonstrating industry-leading accuracy across complex multi-table PDF formats, footnoted scope definitions, plots, images, and varying carbon, waste and energy measurement units.

Evaluation of Earthmark's Pro Search extraction accuracy. Correct as of 13th August 2026.

6. How This Audit Was Conducted 

We have randomly selected 100 companies from different sectors and revenue brackets and manually verified automatically extracted datapoints against the original source reports.

FAQs

What makes Earthmark’s database unique?

Earthmark combines automated AI report extraction (Pro Search) with multi-source third-party assessment (SBTi, B Corp, Good Shopping Guide, Yahoo Finance) and historical trend tracking across 27,000+ companies and 2.85 million structured datapoints.

How accurate is Earthmark’s AI data extraction?

In our latest benchmark test, Pro Search achieved 97% extraction accuracy when reading complex corporate PDF disclosures.

Does Earthmark cover private and public companies?

Yes. Earthmark indexes over 27,000 global entities spanning major public enterprise listings as well as private middle-market companies across 11 key GICS sectors.

Conclusion & Next Steps

High-quality climate data requires a combination of architectural scale, multi-source validation, and specialized domain-specific AI. With over 2.85 million datapoints and sub-15-second processing times, Earthmark is building the foundation for real-time corporate sustainability intelligence.

To learn more about Earthmark’s data solutions or integrate our API into your enterprise workflow, visit: https://business.earthmark.io/marketplaces