Method
How every number here is checked
Sources
Everything starts from primary sources: BRSR filings retrieved from NSE's and BSE's corporate filing archives (XBRL and PDF), SEBI's published taxonomy and circulars, and government datasets. No number on this site originates from a press report, a vendor database, or an estimate of ours.
Extraction
XBRL filings are parsed deterministically against SEBI's taxonomy — plain code, no AI extraction, so every value is reproducible from the raw filing. The parser, the raw-file manifest and the extracted datasets are published as open downloads on the Data page, so any figure can be re-derived from the same public inputs.
Verification — the standard every finding must pass
- Two sources per figure. A published number must be confirmed by a second independent source: the company's own PDF filing, a prior-year comparative, or an external dataset.
- Findings are triangulated two to three ways before publication. When verification narrows a finding, we publish the narrower version. When it kills one, we don't publish — and where the failed check is itself instructive, we document it (see the Tata Motors exclusion in Finding 02).
- Ambiguity is stated, never smoothed over. Unverified cases are named as unverified and excluded from every summary statistic.
What we do not do
- No ESG ratings or scores — we analyze disclosure, not performance.
- No accusations. Where a filing is compliant, we say so in the same breath.
- No unsourced claims, no adjectives doing the work numbers should do.
Corrections
Errors get corrected in place, with the correction and its date logged publicly on the affected page. We hold ourselves to the standard we read filings by.