Same metric. Same year.
Different rulebook.
Method.
Companies.
Compare.
An ESG number is only as meaningful as the rulebook behind it.
Different boundaries, bases, assurance levels, and fiscal years sit underneath numbers that look identical on the page. One platform reads both halves of the disclosure record: Kestrel reads the words, Strix extracts the numbers — and both capture the rulebook, so you compare like with like.
The narrative side of mining reporting — climate strategy, community, governance, human rights — read passage by passage. It tracks how commitments are made, maintained, modified, or quietly dropped, and how much of each report is genuinely new versus recycled from last year.
Turns thousands of pages of factbooks and reports into structured ESG metrics on one canonical standard — and captures the rulebook behind each number: boundary, basis, assurance, methodology. So when two figures were never built to be compared, you know before you compare them.
Kestrel analyses what companies say. Strix analyses what they report in numbers. That is the difference between scraping ESG data and understanding it — every claim and every datapoint traceable to its source.
Anyone can download an ESG report. Understanding all of them is another thing entirely.
Our 2026 State of ESG Disclosure report found 48.5% of directly-comparable metric pairs confounded by a methodology mismatch, and only 37.7% of disclosures carrying any external assurance. Three guarantees are built into every output to deal with exactly that.
Every number is auditable. Each figure points to the exact source cell, page, or paragraph it came from — nothing is asserted that can’t be traced. That is what lets us name companies with confidence.
We model the rulebook, not just the number — boundary basis, JV treatment, assurance level, fiscal period. We never silently “adjust” one company to match another; we expose the difference, so you know when a league table is apples-to-oranges.
We diff every report against its own past. Silent restatements, recycled disclosures, and quietly-dropped targets are invisible to single-report analysis — they only appear when you hold every year at once, which we do.
Last year’s numbers aren’t fixed.
Reported figures get revised in later reports — in our 2026 State of ESG Disclosure report, 26% of detected restatements came with no disclosed reason at all. Because we hold every figure from every year with full provenance, we can diff the “2021” in the 2021 report against the “2021” in a 2024 report — and flag the change the moment it appears.
live examples from the corpus — every one traceable to its source report.
Across 105,227 disclosed passages, just 0.62% describe measurable shared value.
The words get the same scrutiny as the numbers. Creating Shared Value means a mining activity that earns a genuine commercial return and delivers a measurable benefit to host communities or the environment at the same time — real value for the business and for society, not donations or good PR. Kestrel reads every passage and scores it against four explicit criteria, so the 95.18% that creates no measurable value can’t hide in the prose.
Tied to a core operational dependency — not philanthropy bolted on the side.
A disclosed, quantified result: beneficiaries, volumes, money — something you can check.
An explicit mechanism from the initiative to the shared benefit, not a vague association.
Evidence that survives beyond the company’s own narrative.
Mining HQs across six continents.
The 39 mining majors covered by Kestrel are headquartered in 25 cities across six continents.
39 majors. One standard.
Open any company for its numbers, methodology, targets, and restatements — every figure linked to the report it came from, so you can check us against the source.