What Is ESG Data? – Definition and How It Works
Learn what ESG data covers, why ratings disagree, which sources are public, and how collection supports verification.
ESG data describes an organisation's environmental, social, and governance characteristics. It is used in investment decisions, supply-chain due diligence, and regulatory reporting.
1. What the Three Pillars Cover
| Pillar | Typical measures |
|---|---|
| Environmental | Emissions by scope, energy use, water, waste, biodiversity impact |
| Social | Labour practices, safety records, diversity data, community impact, supply-chain conditions |
| Governance | Board composition and independence, executive pay, audit arrangements, shareholder rights, ethics record |
Governance data is the most standardised, because much of it is required disclosure. Environmental data is improving under mandatory reporting regimes. Social data is the least comparable, since definitions vary widely between companies.
2. Why Ratings Disagree
The defining problem of the field: two providers can rate the same company very differently, and both be internally consistent.
The reasons are structural rather than accidental:
- Different indicators. Providers measure different things.
- Different weightings. Two providers agreeing on the facts can weight them oppositely.
- Different treatment of missing data — imputed, penalised, or excluded.
- Different materiality views about which factors matter for which industry.
- Different data vintages.
- Self-reported inputs, unverified and inconsistently defined between companies.
The consequence is that an ESG score is a provider's opinion expressed as a number, and comparing scores across providers is not meaningful. Comparing underlying indicators is.
3. Public Sources
A substantial amount of ESG-relevant data is publicly available, which is what makes independent collection viable:
- Regulatory filings — annual reports, sustainability reports, mandatory disclosures.
- Company websites — policies, targets, governance documents.
- Government registers — emissions inventories, environmental permits, enforcement actions.
- Court and regulatory records — litigation, penalties, settlements.
- Standards-body disclosures, where companies report to a voluntary framework.
- News and NGO reporting on controversies.
- Job postings and career pages, which reveal what an organisation is actually building.
4. Where Independent Collection Adds Most
The gap ESG providers struggle with is verification. Self-reported data is not checked against observable behaviour, and that is exactly what collected data can address.
Examples of the pattern:
- A stated diversity commitment against the composition visible in public leadership pages.
- A stated environmental target against enforcement actions in a public register.
- A supply-chain policy against supplier lists and their own public records.
- A safety claim against regulator-published incident data.
- A stated hiring direction against actual job postings.
The value is contradiction detection between claim and record — which is verification rather than another opinion.
5. Collection Considerations
Disclosure documents are PDFs, often long, with the numbers in tables. Extraction quality determines everything downstream, and a table misread is a wrong figure presented confidently.
Definitions vary between companies. Two emissions figures labelled the same way may cover different boundaries. Capture the definition alongside the number, or the comparison is invalid.
Point-in-time matters. ESG data is reported annually with restatements; the reporting period and any restatement must be recorded, not just the value.
Entity resolution is essential. Subsidiaries, joint ventures, and renamed entities must be linked correctly, or data is attributed to the wrong company.
Controversies need dating. An incident from a decade ago is not current performance.
6. Personal Data Appears
Governance data includes named individuals — directors, executives, their compensation and other appointments. Social data can include workforce information.
Named-individual data is personal data with the corresponding obligations, even where the source is a public filing. The disciplined approach is to collect what the analysis requires, record its source, set a retention period, and avoid accumulating personal detail because it happened to be in the document.
7. What Good Practice Looks Like
- Collect indicators, not scores. Scores are opinions; indicators are comparable.
- Record the definition and boundary for every metric.
- Record the reporting period and any restatement.
- Keep the source document, so a figure can be re-verified.
- Separate self-reported from independently verifiable data explicitly.
- Date controversies and decay their weight.
- Resolve entities carefully across subsidiaries and name changes.
Point 5 is the one that gives the dataset its value: knowing which figures are claims and which are records.
8. Summary
ESG data covers environmental, social, and governance characteristics, with governance the most standardised and social the least comparable, and provider ratings disagree because they measure, weight, and treat missing data differently.
Independent collection from filings, registers, enforcement records, and company pages is most valuable for verifying self-reported claims against observable record — so collect indicators rather than scores, capture each metric's definition and reporting period, keep source documents, and mark clearly which figures are claims.