Reported workforce and safety data
Updated: Oct 7
Read the population and time basis before using a workforce count or safety rate. This collection contains 349 reported observations from 29 Indian issuers: selected employee/worker headcounts, lost-time injury frequency rates, recordable-injury counts, fatality counts and source-reported training tallies and percentages. It is an evidence reference, not a safety ranking.
Edition: 0.6.0-sample1661. Hub review: 6 October 2026. Selected metric periods run from 1 April 2022 to 31 March 2026, with uneven issuer, year and metric coverage. Release source retrieval and selected-row checking took place on 5–6 October 2026; row timestamps remain in the files. Original report-publication dates were not established. This is a fixed historical, source-screened sample, not a live feed, representative sample, complete issuer inventory or latest-year coverage claim.
A headcount and a safety rate answer different questions
There are 211 headcount observations, 40 lost-time injury frequency-rate observations, 26 fatality-count observations, 14 recordable-injury observations and 58 training observations (29 tallies and 29 reported percentages). Headcount totals and permanent/non-permanent components overlap. Counts are not full-time equivalents, annual averages or hours worked unless the source explicitly uses such a basis.
Preserve Employees and Workers as source population labels. A contract-workforce note in a safety table does not automatically redefine the headcount table.
Training topics can overlap. SIS’s FY2025–26 employee training figures include repeated courses and people who left during the year; those tallies are not unique people, and the reported percentages above 100% are retained. Preserve source-specific populations, denominators and literal Number or No. units, and do not transfer a Group or committee remit to standalone workforce scope. SIS’s FY2025–26 worker-training rows and Himadri’s FY2025–26 worker health-and-safety training rows remain excluded.
Ten teaching observations: overlapping counts and rate bases
The two shipbuilders remain the unchanged ten-observation teaching selection from edition 0.5.0-sample1040. All selected periods are 2024–25, but table-specific population and method qualifications still matter. Original rate-unit spelling is retained. No cross-company total or performance ordering is supplied.
Cochin Shipyard Limited
Standalone CSL. Headcounts are the financial-year-end snapshot in Q20, not an annual average or full-time-equivalent measure. The separate safety table explicitly includes contract workforce; that note is not silently transferred to the headcount table.
Reporting period: 2024-04-01 to 2025-03-31. Selected source: Cochin Shipyard Limited Business Responsibility and Sustainability Report FY2024-25.
Employees: Permanent (D)
Reported value: 695 number. Printed p02 / PDF p2.
Employees: Other than Permanent (E)
Reported value: 199 number. Printed p02 / PDF p2.
Total employees (D+E)
Reported value: 894 number. Printed p02 / PDF p2.
Total workers (F+G)
Reported value: 2647 number. Printed p02 / PDF p2.
Lost Time Injury Frequency Rate (LTIFR) - Workers
Reported value: 0.0648 per one million person-hours worked. Printed p18 / PDF p18.
No. of fatalities - Workers
Reported value: 0 number. Printed p18 / PDF p18.
Mazagon Dock Shipbuilders Limited
Standalone MDL. Headcounts are at financial-year end. The Workers population in the safety table is retained as classified by the issuer; contractor equivalence is not inferred from the headcount categories.
Reporting period: 2024-04-01 to 2025-03-31. Selected source: Mazagon Dock Shipbuilders Limited Business Responsibility & Sustainability Report 2024-2025.
Workers: Permanent (F)
Reported value: 1679 number. Printed p3 / PDF p3.
Workers: Other than Permanent (G)
Reported value: 3385 number. Printed p3 / PDF p3.
Total workers (F + G)
Reported value: 5064 number. Printed p3 / PDF p3.
Lost Time Injury Frequency Rate (LTIFR) - Workers
Reported value: 0.04 per one million-person hours worked. Printed p27 / PDF p27.
Reconcile the right categories; do not invent exposure hours
For CSL’s employee snapshot, 695 permanent + 199 other-than-permanent = 894 employees. The total-minus-components diagnostic is 0 number. Adding all three rows would count the same employee population twice. For MDL’s worker snapshot, 1679 + 3385 = 5064 workers, again with a zero diagnostic. These arithmetic checks do not establish that Employees and Workers are disjoint across every issuer. CSL Q20, printed p02 / PDF p2; MDL Q20, printed/PDF p3.
CSL’s 0.0648 rate uses the original unit “per one million person-hours worked”; MDL’s 0.04 uses “per one million-person hours worked.” The million-hour number is the rate base, not the actual exposure hours. Do not multiply the rate by headcount, infer an injury count, or infer hours worked without the underlying same-scope numerator and denominator. An explicit fatality count of 0 is still only the source’s specific reported measure and scope.
A population-first reuse checklist
Identify the exact employee/worker category and any contractor coverage, exclusions or organizational changes.
Distinguish a period-end snapshot from an annual flow, average or FTE. Check whether a prior-year figure is a later comparative or an original-source selection.
Keep lost-time injury frequency separate from recordable injuries, high-consequence injuries and fatalities. Related labels do not create interchangeable numerators.
Use the source’s actual hours and rate definition for a calculation. A shared per-million label does not establish identical case definitions or measurement quality.
Keep unavailable cells and explicit zero separate. Absence from this numeric selection is not proof of no incidents or a reporting failure.
Download and use the collection
The 0.6.0-sample1661 complete edition ZIP contains D006-workforce-safety.csv and D006-workforce-safety.json, each with the same 349 reported observations. Extract the format you prefer. The page provides selected teaching cards; it does not provide an on-page search or filtering tool. CSV and JSON are two formats of the same records.
Use the source, coverage, metric, period, history and qualification files inside this ZIP for the expanded collection. Retain mandatory-qualifications, record-qualifications and source-provenance-availability with any subset; shorter source-scope summaries do not replace the full conditions.
Previous-edition example support: the seven links below remain bound to 0.5.0-sample1040, which supports the unchanged teaching examples. Use the current-edition ZIP above for the expanded collection and its supporting files.
Open D006-workforce-safety.csv in a spreadsheet with UTF-8 import. Keep reported_value as text if the original decimal precision matters.
Filter company_id, period_start, period_end, metric_id and original_unit. Keep observation_id with every subset.
Read boundary, method, denominator_if_relevant, footnote_context and comparability_caveat before using the value. Follow source_id, source_url and the printed/PDF locator to the report.
Retain version, source_history_status and the supporting dictionaries. The archive contains data-schema.json, metric-definition-variants.json and the history schema. 68 of this collection’s 349 observations have a structured original/later-source pair; 281 have none in the reviewed history inputs. A missing structured history pair does not mean that a value is unrevised.
Attribution, missing information and maintenance
The numeric file includes explicitly reported numbers. Blank metadata, dashes, Nil, NA and unavailable disclosures are not silently converted to zero. An explicit numeric zero stays zero. A source may not explain an abbreviation. In history CSV files, the token \N represents JSON null; an actual empty source cell remains empty. Histories, amendments, conversions and repeated formats add no observations.
AI-assisted preparation and separate internal source checks support transcription and interpretation. They do not independently verify company measurements, provide professional assurance or establish compliance. The issuers are not identified as Nitium clients or endorsers. No report PDFs, report graphics or third-party assurance materials are distributed. The release does not grant a blanket licence over source reports; retain attribution, exact units, locations and source-specific limits when making a subset.
This frozen edition changes only through versioned corrections or a reviewed successor. Source amendments, broken links, altered reuse conditions and material method changes trigger a review. New company reports are not incorporated automatically. The method file preserves source-specific access limits, including City Union Bank’s direct-PDF availability caveat.
Prepare your own evidence trail
Use the public ESG report directory to locate reporting boundaries and editions. The downloaded files provide historical examples; your own reporting needs its own period, scope and source records.
For help with BRSR workforce and safety-data preparation, review Nitium’s services or discuss the reporting question and evidence gaps.
Comments