Workforce definitions across five selected reports

A headcount is usable only with a description of the people it represents. Similar labels can hide different category rules, entity coverage or time bases. Before a reporting team combines numbers or compares periods, it needs to know whether the proposed calculation answers the intended question.

This study preserves five purposively selected workforce-definition cases from Nitium’s fixed 0.12.17-local1475-gipcl-cera edition, frozen on 6 October 2026. The unit of analysis is one definition question per issuer. Two accepted observations provide a non-additive headcount illustration. All five cases and both observations were also in the historical 1,040-record public projection; they create no new company-data records here.

We reread the selected report panels and notes using exact official PDFs retrieved on 15 October 2026. These are fixed historical editions, not a latest-workforce survey. The study does not inspect personnel files, determine legal employment status, rank employers or estimate undisclosed populations. Its practical contribution is a reproducible way to test whether a proposed total or comparison is supported.

Five source-linked definition cases

RCF’s standalone FY2024–25 workforce note excludes permanent workers from permanent employees. Keep that exclusion with the employee figure; no combined workforce total is calculated here. Source: RCF BRSR, PDF 2, printed 68.

City Union Bank’s FY2024–25 report places its workforce in the Employees category and treats Workers disclosures as inapplicable. This is the bank’s stated categorisation, not Nitium’s legal conclusion or a numeric zero-worker observation. Source: bank BRSR, PDF 4, unnumbered.

eClerx FY2024–25 BRSR, PDF 2: employee headcount covers offices and subsidiaries in India.

NBCC’s FY2024–25 notes define workers as its non-executive Group C&D employees. The headcount and safety panels use unresolved category wording; the selected evidence supplies no cross-table population mapping. Keep the company definition without treating it as all project or contractor personnel. Source: NBCC BRSR, PDF 2/21/40, printed 175/194/213.

Suprajit’s standalone FY2025–26 year-end table gives both permanent and total employees as 1,016; these overlap. Other-than-permanent cells are dashes, while a footnote describes temporary staffing responding to production needs. Neither equality nor the dash establishes annual absence, an annual average or a temporary-staff count. Source: Suprajit annual report, PDF 60, printed 59.

Specify the proposed use before combining rows

Write the question as a complete sentence. “How many selected employees were on the agreed entity roster at the cutoff?” requires different evidence from “How many people worked at any time during the year?” A reader should not need to infer the intended population from a dashboard heading or a familiar word such as total.

Give each input its source label, included and excluded groups, entity and location coverage, time basis, relationship to other rows and permitted use. These six fields describe the proposed calculation; they do not establish a universal employment taxonomy. Keep an owner and an unresolved-question reference alongside the description so a later clarification can be traced.

For addition, ask whether the selected groups are disjoint and whether they exhaust the intended population. For subtraction, confirm that the smaller group is fully contained in the larger group on the same basis. Equal values alone prove neither relationship. A reconciliation can be numerically neat while combining a total with its own components.

For a comparison, check that the earlier and later observations use the same counting unit, entity coverage and time basis. If one is a snapshot and the other covers everyone who participated during a period, leave them as separate measures. A common reporting year does not make their membership or timing identical.

Original test: one population, three rows

Everything in this exercise is fictional. Cedar Components requests a selected employee snapshot at 23:59 IST on 30 September 2026. Its authorised custodian confirms 90 people in the parent entity and 30 in one included domestic subsidiary. The two groups are disjoint and exhaustive, with each person counted once. No other entity or location is included.

A second approved classification covers exactly those same 120 people: 100 permanent and 20 other-than-permanent employees. Those two categories are also disjoint and exhaustive for this snapshot. The total row and the two classification rows are alternative views of one population, rather than three separate groups to combine.

The entity check is 90 + 30 = 120. The category check is 100 + 20 = 120. Adding the total and its components gives 120 + 100 + 20 = 240, which counts every person twice under the stated assumptions. The discrepancy comes from the selected rows’ relationship; it does not identify 120 additional employees.

Now suppose the analyst receives only rows labelled total 120 and permanent 100, with no category definition or mapping. Their difference is arithmetically 20, but labelling it other-than-permanent employees is unsupported until containment, shared coverage and an exhaustive classification are confirmed. Preserve the arithmetic as a hypothetical diagnostic and leave the requested category unresolved.

In a second variation, the parent’s 90-person count is a year-end snapshot while the subsidiary’s 30 describes everyone engaged at any time during the year. Their sum still equals 120. It no longer answers the original snapshot question. Obtain a matched subsidiary snapshot before presenting the combined result; do not repair the mismatch by changing only the headline.

In a third variation, the custodian confirms the snapshot partition but supplies no annual staffing history. The 120-person snapshot remains usable for its stated purpose. It cannot establish annual average staffing or the number of distinct people who worked earlier and left. Leave those outputs unavailable rather than borrowing the snapshot total.

Original fictional population test. Both partitions cover the same 120 people at one cutoff. Confirm the row relationships before adding or subtracting; arithmetic alone cannot establish them.

Record a decision another reviewer can reproduce

Retain the selected output question, exact input versions and the relationship evidence that permits the operation. In the complete fictional case, the decision records two valid reconciliations to 120, rejects the 240 double count and limits the result to the named cutoff and entities. In the incomplete variants, it records the precise missing mapping or date basis.

Ask a second reader to reproduce the result using only this record. They should be able to identify which rows are alternative presentations, which form a permitted partition and which cannot be combined. If they need an undocumented assumption, return the case to the responsible owner before reusing the number.

Keep clarification separate from correction. An owner may explain an existing population without changing the underlying count; a revised extract may instead change membership or repair a value. Record what actually changed and retain the previous version. Do not overwrite an old definition merely because a later report supplies more detail.

Use aggregate evidence for this reporting review. Person-level matching and sensitive employment records should stay with authorised custodians; the article supplies no permission to collect or publish them. A reviewer can ask for a controlled confirmation of population relationships without copying the underlying roster into a general reporting pack.

Scope and next step

The five cases illustrate defined questions within a small historical selection. They do not measure how common those practices are or establish whether any issuer’s underlying workforce records are correct. No safety rate, contractor total, cross-company workforce total or employer ranking is calculated. Historical source notes remain attached to their own editions.

Use the reported workforce and safety data collection for its separate historical observation records and source qualifications. This study tests whether selected populations support a particular operation; it adds no observations or automatic validation. Retain the responsible owner’s definitions, mapping decisions and disclosure review when applying the exercise to a real reporting question.

Nitium’s reporting support can help organise definitions, evidence and unresolved review questions within an agreed scope. Discuss the intended reporting output and available aggregate evidence through Contact. Avoid sending personnel records in an initial enquiry.

Prepared 15 October 2026 with AI-assisted research and internal source checks. Historical cohort and original report periods are retained. Source checking supports transcription and interpretation, not independent verification of company measurements, legal classification or professional assurance. Issuer endorsement and client relationships are not claimed.