NITIUM ADVISORY: REPORTED ESG DISCLOSURES
Edition 0.6.0-sample1661
Prepared 6 October 2026. A frozen historical sample, not a live feed.

WHAT THIS EDITION CONTAINS

1,661 eligible records: 112 company/reporting-period directory entries, 1,422 reported metric observations and 127 original disclosure-example analyses. The six metric collections contain 246 energy, 145 greenhouse-gas, 279 water, 298 waste, 349 workforce/safety and 105 governance observations.

The sample covers 30 Indian issuers in 18 broad project industry groups. The earliest selected period starts 1 April 2022 and the latest ends 31 March 2026. Coverage differs by company, year and metric. These endpoints do not mean that every issuer has four complete years of metrics or that the most recently available report has been extracted for every issuer. See coverage.csv/json for the actual company-period and collection coverage.

This is a purposive source-availability and source-use-screened sample. It is not an exchange-index universe, probability sample or representative account of Indian business. Official annual-report/BRSR sources were prioritised, with sector additions and genuine multiple-period directory evidence. Source-access and reuse limitations affect selection. Neither inclusion nor exclusion is a sustainability ranking or a finding about compliance.

HOW RECORDS ARE COUNTED

One company and reporting period is one directory entry. An annual report, BRSR and date-only supporting document can support that same entry. A comparative column alone does not create an additional directory period. The 112 entries are supported by own-year report evidence; some other metric periods appear only as comparatives.

Each stable observation ID counts once. Source versions, restatements, metadata amendments, copies of a report, translated documents, repeated comparative cells and unit conversions add no observations. A reported total and its explicitly reported component are different concepts, but they overlap and must not be added together. A separately labelled permanent headcount and total headcount can have equal reported values without representing disjoint populations.

The 127 D008 entries are original analyses of specific disclosure methods and limitations, not extra reported quantities. The 139 source documents, company table, dictionaries, coverage rows, period qualifications, 430 paired-source histories and 18 observation-version events add zero eligible records. CSV and JSON are alternative formats of the same rows. The counting script reads only the eight primary collections and deduplicates their stable IDs.

VALUES, UNITS AND MISSING INFORMATION

Reported quantities are decimal strings, retaining original precision and unit spellings. No shared unit is silently substituted for a source unit. In particular, a label such as Million GJ must retain its scale, and tCO2, tCO2e, tCO2eq and ratios carrying an energy denominator must not be assumed equivalent. Normalized fields are separate; a blank normalized field means no normalization is supplied and adds no fact.

The primary numeric collections contain selected, explicitly reported numbers. Raw Nil, NA, N.A., dashes, blanks, not-tracked states and unavailable disclosures are not converted to numeric zero. A source's explicit numeric zero remains a numeric string. The source may not define an abbreviation; do not automatically interpret NA as either not applicable or not available.

Primary CSV blank metadata cells mean unavailable or not supplied, never zero. Lists and objects are JSON-encoded inside CSV cells. In every CSV file, the literal two-character token \N means JSON null. A genuinely empty raw source cell remains an empty string. This distinction prevents a recorded source blank from becoming indistinguishable from metadata that was not retained. See data-schema.json and history-schema.json.

PERIODS AND SOURCE LOCATIONS

Reporting dates, source publication dates, retrieval dates and the data-edition date are different fields. Original publication dates were not established for these 139 source records. Upload-folder years, signature dates and retrieval dates are not substitutes. Source retrieval and row transcription-check timestamps are retained separately.

Some April starts are qualified inferences from a fiscal-year label and explicit March-end/date corroboration, rather than a complete quoted standalone BRSR range. Vardhman, KPR Mill, West Coast Paper, Andhra Paper, MDL FY2026 and several added issuer-periods have relevant qualifications, including Cera, Sundram and Himadri. The full company/period tables control. Read period-qualifications.csv/json and each row's date-source locations. A financial date panel supplies dates only; it does not expand an ESG boundary to consolidated financial statements.

Printed folios and one-based PDF pages can differ. Some reports use two-page spreads; left/right panels are named where relevant. An unnumbered source page is described as a PDF page with an unnumbered-source qualification. Do not invent a printed folio from a PDF offset. The directory retains the legacy REP-CO-JKLAKSHMI-20250331 identifier unchanged.

SOURCE REVIEW AND WHAT IT ESTABLISHES

Every eligible record received a direct check against its source location, followed by a separate internal source-checking pass. Review covered selected values, signs, decimals, units, years, boundaries, labels and relevant notes; metadata corrections were resolved before canonical acceptance. AI-assisted research and data preparation were used. Machine checks establish structure, references and consistency, not the truth of a source disclosure.

This process checks transcription and interpretation of published reports. It does not independently measure emissions or other company quantities, provide assurance of company disclosures, certify legal compliance, or constitute the programme's separate final independent QC. Issuer statements that an external agency assessed a particular table remain attributed issuer statements with their stated limits.

No report PDFs, screenshots, report graphics, logos, third-party auditor/assurance/certificate text, personal stakeholder contact details or private client material are included. Names identify public disclosures, not Nitium clients, endorsers or independently assured companies.

PAIRED HISTORY AND RECORD VERSIONS

source-history.csv/json contains 430 specifically preserved original/later-source relationships tied to accepted observation IDs. observation-version-history.csv/json contains 18 record-version events. They are supporting evidence, not additional metric observations. The history table states whether the accepted observation selects the original or the later side; a later report is not automatically the selected source.

992 of 1422 observations have no structured pair in the reviewed history inputs supplied for this edition. This includes different situations and must not be described as 992 unrevised observations, 992 unverified observations, or 992 source errors. Direct checking of the selected source and the availability of a structured original/later pair are separate questions. Current row qualifiers remain controlling.

Where an original-unit field was not retained in structured history, it stays null with that exact qualification. This does not claim that the source omitted a unit. It is not filled from the later source or a normalized lookup. A raw source blank, an unquantified earlier value and a missing metadata field remain distinguishable.

Selected examples of why the distinctions matter include the CSL comparative waste revision, MDL's originally unquantified prior discharge, RCF's original battery dash versus later numeric comparative, and Dwarikesh's change from a tCO2eq/Mwh label to tCO2eq. Incompatible units prevent a meaningful subtraction or assumed conversion. An arithmetic coincidence is not a causal explanation. Preserve unknown revision reasons and original source labels.

COMPARISON RULES

Read entity/site boundary, period, original unit, method, denominator, population and source version together. A common metric ID groups related concepts; it does not certify comparability. The source-specific dictionary variants must remain available with any subset. Do not sum parent and subsidiary scopes or overlapping table populations.

GHG coverage includes source-specific Scope 1, Scope 2, limited Scope 3 and separately defined combined-intensity concepts. Read the exact metric dictionary and row-specific method labels; unspecified-method Scope 2 rows are not silently relabelled location-based or market-based. A named emissions-factor source alone does not supply that label. Limited business-travel and selected-category Scope 3 figures are not complete value-chain totals. Retain Aurobindo's material-category/commuting limits, eClerx's India/global exceptions, MDL's equipment/travel exclusions and the other row-specific notes.

Withdrawal, consumption, discharge and recycled/reused water are distinct. Some sources report estimates, shared/site-specific flows or differing table boundaries. A reported arithmetic residual is a diagnostic, not a repaired water balance. Dwarikesh's sugar-plant discharge estimate excludes its distillery scope; West Coast's water accounting includes recycling; City Union Bank's consumption estimate has its own basis. Keep those distinctions.

Generated waste, recovered waste and disposed waste are not automatically a closed, disjoint material flow. Preserve category classifications, reporting precision, blank/dash states, stock/timing uncertainty and residuals. Workforce headcounts and Board counts are snapshots, not FTE or annual averages. Rate bases such as per one million-person hours are not the actual hours worked. Contract-workforce coverage and employee/worker population labels must remain attached to safety figures.

No default cross-company totals, performance ranking, harmonized emissions inventory or full-coverage time series is supplied. A figure can be useful for illustrating a disclosure method while remaining unsuitable for a peer comparison.

ATTRIBUTION, SOURCE CONDITIONS AND USE

The primary publishers are named in companies.csv/json and sources.csv/json. Nitium Advisory prepared the data structure, factual transcriptions and original analyses. Each main row carries attribution and an official report URL with a precise locator. Source-specific use scopes and discovery limitations are retained in the source table.

CSL and MDL have conditional attributed-reproduction policies; identified third-party material is excluded. Other selected uses follow bounded factual-use checks where no applicable restriction was identified in the examined material. This is not an affirmative open licence. Empty policy pages, unlinked legacy policy labels and document-specific limitations do not become broad permissions. The reports remain on the issuers' websites. Source links should open separately without framing where required.

This edition grants no rights over source reports or third-party protected material. Do not strip attribution, source locations, original units, history qualifications or caveats when making a subset. Do not reproduce report assets merely because the report can be opened publicly.

USING AND MAINTAINING THE EDITION

Start with a reader question and the relevant collection. Filter the downloaded CSV/JSON by company, period, metric and original unit, then inspect scope and method. Spreadsheet filtering is available through downloads; a website-wide live search interface is not asserted by this package.

A direct availability check of City Union Bank’s BRSR returned HTTP 502 at 02:09 UTC on 6 October 2026, while the public web reader still returned cached text. That check does not establish fresh binary or image verification. Source availability can change; the source table and affected rows retain this limitation.

This is a fixed historical edition. Newer issuer reports may exist and are not incorporated automatically. Corrections and expanded selections require a new edition with a change record. The original 100-record two-shipbuilder calibration edition 0.4.3 remains a distinct historical study selection; this larger edition does not expand R001's research sample retrospectively.

For reporting preparation support, visit https://www.nitium.in/services or discuss your period, entity structure and evidence gaps at https://www.nitium.in/contact or contact@nitium.in. No assurance, certification or guaranteed reporting outcome is implied.

SUCCESSOR-SPECIFIC QUALIFICATIONS

Exactly 621 new stable IDs extend the existing 1,040-record selection. The original 100-record two-shipbuilder study and other fixed-cohort studies retain their original samples. Formats, histories, replacements and metadata corrections add no resources or records. The larger sample remains partial and purposive.

All ten accepted Gateway locator corrections are metadata-only. Physical PDF3/19/21 are unnumbered; PDF4/18/20 have printed3/17/19. Four observation versions increment, without changing selected quantities, physical pages or source URLs.

Record-qualifications retains additional raw-token, panel, denominator, actual-hours, source-unit and period qualifications where supplied. Source-history-unit-bindings, history qualifiers and alternative-source support retain exact source vintages. All 18 older West Coast enrichment nulls remain null. Unsupported pair absence never means no revision.

Sundram's FY2026 dependent environmental/safety classes remain excluded. SIS worker-training holds and Himadri worker-health holds remain excluded. Repeated courses/leavers can affect training tallies; topics are not unique people and rates are not capped at100%. Reported Himadri57.81% stays57.81 with its662/1145 rounding caveat. Literal Number and No. remain distinct where required. Audit Committee counts and ESG Committee remits do not imply general workforce, Group or whole-Board scope. See source-scope-and-exclusions and individual rows.

No source PDF, page image, private rights review, assurance statement or certificate is included. Historical scope-document availability is disclosed in source-provenance-availability. Thirty-two source/path bindings refer to18 absent historical documents. Retained scope entries and bounded continuity evidence are different from original files; no missing file is represented as restored and no new licence is inferred. Direct source report URLs remain separately identified.

This fixed package does not itself establish website publication. Exact edition/member checks and independent release approval are separate from source transcription. Author checks alone do not clear publication. Mandatory-qualifications.csv/json retains the full substantive source-specific limitations; shorter source-scope notes are orientation only and do not replace those conditions.
