How a credit-data provider populated its ESG product with structured emissions and sustainability data.

ESG metrics hide in unstructured PDFs

The customer's ESG product surface carries ESG-related risk signals on businesses, used by lenders, insurers, and corporate-credit teams to factor sustainability into credit and risk decisions. The challenge isn't whether the data exists. The challenge is the form it lives in.

Most ESG-relevant data on private and listed companies sits in unstructured documents (Carbon Reduction Plans, annual sustainability reports, ESG disclosures, supplier questionnaires), published as PDFs in formats that vary by company, industry, region, and regulator. Manual extraction by sustainability analysts produces structured records on a handful of companies at a time, but doesn't scale to the universe an ESG-credit product needs to cover. Inconsistency between extraction templates undermines comparison across companies. Outdated data, especially on emission targets and reduction progress, erodes the product's credibility for users running risk decisions on it.

The brief: extract a defined set of ESG metrics from CRP and sustainability documents into structured records compatible with the ESG product's schema, and validate the approach against a 50-company sample drawn from the customer's UK customer base.

Specialty extraction against the ESG product schema

Veridion built a specialty-extraction pipeline targeting the metric set the customer's ESG team scoped. The pipeline ingests CRP and sustainability PDFs from each company's website, extracts structured fields using narrative-extraction models against ESG-specific schemas, validates the output, and resolves each company to a stable identifier in Veridion's company knowledge graph so records join cleanly to the rest of the customer's data.

The metric set spans both quantitative emissions data and qualitative sustainability signals. On the quantitative side, the pipeline pulls baseline and current emissions across scopes 1, 2, and 3 in metric tonnes CO₂e, reduction targets expressed as a percentage by year, carbon intensity in tonnes CO₂e per million dollars of revenue, and operational-footprint data (countries of operation per entity and per group) drawn from the company knowledge graph. On the qualitative side, the pipeline surfaces signals on resource use and circular-economy practices, waste-management policies, human-rights commitments, diversity-and-inclusion policies, and living-wage commitments, extracted from policy text and ESG news where available, tagged for review rather than auto-scored.

Output ships in two layers. A red-green risk framework gives the ESG product's users a first-pass directional signal at the company level. Beneath that, the structured-metric record carries the underlying numbers and policy signals for analysts running deeper diligence.

First ESG use case under the Reseller Agreement

The 50-company validation set has been delivered to the customer's ESG team and forms the foundation for the next iteration of the product. The pipeline is the first ESG-specific use case under the Reseller Agreement, anchoring an upsell path into the customer's UK Business Information customer base for ESG data that registry feeds and statutory accounts don't carry.