How five disclosure cases distinguish missing data and estimates

A blank cell, a reported count and an estimated mass can each leave a different question unanswered. Reading them all as “missing data” loses information; treating every visible number as measured loses a different kind. This study asks what must survive when a detailed disclosure becomes a short research note or dataset entry.

We examine five disclosure cases in four issuer reports for 2024–2025. Two cases come from Aurobindo; the others come from City Union Bank, KRBL and Mazagon Dock. This purposive historical sample illustrates distinctions, not their frequency or company reporting quality.

Five cases, five different evidence requests

1. Information explicitly unavailable

Aurobindo’s waste note says construction and demolition waste information is unavailable. Code this as unavailable information, not evidence of no waste. The next request is for quantity evidence and its coverage, rather than confirmation of an assumed zero.

2. A count exists, but the required mass does not

The same report gives battery counts for both years but excludes them from the waste total because mass information is unavailable. Code count available and mass unavailable separately. A count cannot be inserted into tonnes without supported mass evidence; requesting the count again would miss the actual gap.

3. A stated activity-based estimate

City Union Bank’s water note says both years use estimates based on prescribed norms, employee counts and working days. The report explains the basis; it does not turn the result into metered consumption. Preserve the estimate label and request the norm, inputs and calculation if reproduction is needed. This study does not establish whether that method is suitable for another organisation.

4. An estimate addresses a stated measurement concern

KRBL’s waste note explains that water sprayed on rice-husk ash can increase its weighed mass. It calculates ash using an estimated residual content of approximately 18% of rice husk generated and used in electricity generation. Code an estimated conversion with a stated reason, not a direct dry-mass measurement. Reproduction needs the relevant husk quantity and coefficient basis, not merely another wet weighing.

5. A published total has an explicit exclusion

Mazagon Dock’s Scope 1 note excludes certain off-road vehicles because distance data are unavailable and it seeks to avoid mixed methodologies. Code the disclosed total and the exclusion together. A numeric total does not establish complete coverage. The next request concerns the omitted activity and a suitable calculation basis; this study neither quantifies the omission nor judges the stated methodological choice.

Read the case matrix as questions, not grades

Each case above has four parts: the selected disclosure, our data-state description, the inference to avoid and the evidence needed next. The state belongs to a particular metric, unit, period and population. A report can contain several states at once. Do not assign one company-wide label from a single footnote.

We selected explicit explanations, not unexplained blanks, and read headings, years and footnotes together. Aurobindo contributes two information problems within one waste disclosure, not independent observations for statistical inference. No company measurement was independently tested.

A two-question reading workflow

Question one: is a quantity supplied for the exact item you are trying to record? Include the unit and coverage in that question. A count may answer an item-count question while leaving a mass question unanswered. A number in a total row may answer a selected-coverage question while leaving the complete population unresolved. Preserve what exists without silently promoting it to the requested measure.

Question two: what does the source say about how the quantity was obtained, or why it is absent? Record an identified measurement or calculation basis, an estimate and its inputs, an explicit exclusion, or an unresolved explanation. A generic numeric format cannot carry these distinctions. If the passage gives no reason, write that the reason was not established from the examined passage; do not supply one from industry expectations.

Use the two answers to draft a focused request. Ask for the missing relationship: a mass basis for a count, the inputs behind an estimate, or the activity omitted from a total. Requesting “better data” does not tell the owner which uncertainty must be resolved. Equally, a research reader should not approve a substitute method merely because its arithmetic can be reproduced.

Original experiment: can a short export preserve the decision?

The following four records and all quantities are fictional. Assume a research team receives a clearly labelled schedule for four separate site-period observations of the same emissions indicator, in tonnes of carbon dioxide equivalent. This is an information-preservation test. It does not test emissions calculations, reporting rules or the truth of the schedule.

Record A supplies 12 and explicitly labels a calculation from recorded activity, without estimated inputs. Record B also supplies 12, explicitly described as an estimate using a stated model. Record C supplies no quantity and says the value could not be established. Record D supplies no quantity for an activity that the schedule explicitly excludes because a required input is unavailable. The researchers have not independently validated any of these statements.

Export one keeps only the numeric value. A and B both become 12; C and D both become an empty field. A later reader cannot tell which 12 was estimated or whether an empty field represents an unresolved value or an explicit exclusion. These are two concrete information collisions: different source states produce identical exported values. No percentage accuracy or company score is implied.

Fictional export test. Equal values can represent different data states. Retain the state, basis or reason, covered item and source locator; this preserves information without validating the underlying quantities.

Export two retains value, exact unit, data state, basis or reason, covered item and source locator. A keeps its activity-record basis; B keeps its estimate model; C keeps its unresolved-value explanation; D keeps the omitted activity and missing input. The extra fields preserve the distinctions stipulated in the exercise. They do not make any quantity more accurate.

Test the exports by asking a second reader two questions: which populated record needs the estimate model, and which empty record represents an explicit exclusion? Export one cannot determine either answer from its values alone. Export two identifies B and D. The result shows why a source-linked state field is necessary for these reading decisions; it does not establish that this field set is sufficient for every reporting task.

Make the next use safer

Keep the original source representation alongside the interpreted state. An empty cell, dash or footnote marker should remain traceable even when a research database uses a normalised code. If the code changes after clarification, preserve the earlier reading, new evidence and date. A new answer should not erase the limitation that applied to an earlier release.

Keep quantity availability, method description and population coverage as separate questions. A value can be present but estimated; its method can be documented while coverage remains partial. A statement that a number is reported is weaker than a statement that it is measured, complete or independently verified. Use only the stronger description that the examined evidence actually supports.

This research record supports reading public disclosures. For the operational task of recovering records and deciding whether a proposed estimate can be used, see the missing-records guide and the missing data and estimation decision log. Those decisions require the relevant reporting purpose, method and responsible reviewer; this study does not settle them.

Sources checked 15 October 2026. Aurobindo: printed page 258, left side of PDF page 131, waste footnotes.

City Union Bank: PDF page 47, unnumbered, water table and estimate footnote.

KRBL: printed/PDF page 35, waste calculation note.

Mazagon Dock: printed/PDF page 39, Scope 1 exclusion note.

The source reports retain their own rights and qualifications. Selected issuer facts are attributed; report assets and third-party assurance materials are not reproduced. The experiment and diagram are original Nitium material. This historical sample creates no new company-data records and offers no compliance, performance or assurance conclusion. The issuers are not identified as clients or endorsers.