How often does BRSR XBRL match the company's own PDF?
In a random sample of 114 company-years, the XBRL differed materially from the PDF of the same submission in 10 of them (8.8%).
- Key finding
- 8.8% of sampled company-years (10 of 114; 95% CI 4.8–15.4%) have XBRL that materially differs from their own PDF.
- Data
- BRSR filings on NSE and BSE, FY2021-22 to FY2025-26
- Scope
- Aggregate figures only; no company is named
- From
- Sustainize, Dscvry's sustainability reporting product: it collects ESG data, checks it, and produces GRI, BRSR, ESRS and ISSB disclosures with every number traced to its source.
A BRSR filing exists twice: as a PDF that people read and as XBRL that databases ingest. If the two disagree, the number in every dataset, screener and rating built on XBRL may not be the number the company published. We measured how often that happens.
How we sampled
We drew a simple random sample of 120 company-years from the 4,520 that have both a complete XBRL file and a valid PDF, with no outlier picking. 114 had at least one value we could check. For each, we compared Scope 1, Scope 2, total energy and total waste, for the current and prior year, against the PDF of the same exchange and the same submission. Every candidate difference was read on the PDF page by a person.
What we found
- In 10 of 114 company-years (8.8%; 95% CI 4.8–15.4%) the XBRL value differs materially from the company's own PDF.
- Three were scale or unit errors: a value multiplied by 100, a value multiplied by ten million (a crore), and a figure in gigajoules tagged as kilojoules.
- One was exactly double the PDF figure, in both years.
- The rest were wrong rows, digit slips or values that cannot be reconciled with anything printed in the PDF.
- At value level, 89.7% of values match their own PDF (707 of 788).
The PDF is not always the right one. In at least four company-years the XBRL was correct and the PDF's own arithmetic was wrong, for example a printed total that leaves out one of its components.
Automatic checks need a human
Our automatic scale detector flagged six company-years in the sample. Read on the page, four of the six were artefacts: a typo in the PDF, an OCR misreading, a total the PDF never prints. Automated checks are good at finding where to look; they are not a verdict. This is why Sustainize shows the source page next to every value it is unsure of and asks a person to confirm it.
Why it matters
XBRL is what flows into databases, benchmarks and, increasingly, assurance work. With BRSR Core subject to assurance or assessment for the largest listed companies, a value that differs between XBRL and PDF is a finding waiting to happen. The cheapest time to catch it is before filing.
What to check before filing
- Generate the XBRL and the PDF from the same dataset, not by retyping.
- Reconcile totals with their components in both.
- Check every value against its source page, and keep that link.
See how this works in the live demo: read values from a document, settle the one the system is unsure of, and keep a page link on every number.
Method and caveats
- An earlier version of this analysis compared some XBRL copies with the other exchange's PDF. Those cross-copy comparisons were removed; the 8.8% counts only comparisons with the PDF of the same submission.
- The rate for the second exchange copy of each company-year has not been measured separately.
- Figures are aggregate and no company is named. A difference is a discrepancy, not a judgement on compliance.
- Source: Dscvry analysis of BRSR XBRL and PDF filings on NSE and BSE, FY2021-22 to FY2025-26.