A regulatory reporting review should explain how a reported figure relates to its source, why adjustments were made and what remains unresolved before release. A total that looks plausible is insufficient. The reviewer needs a reproducible chain from the reporting definition and source population to the exact version approved for submission.
This case follows Illustrative Bank, a fictional institution, as it reviews two reporting lines for 30 September 2026. The amounts are in INR lakh, and the line definitions are invented for teaching. The workbook is a reconciliation aid, not a prescribed RBI return. An NBFC can use the same approach after replacing the perimeter, definitions and source controls with those applicable to its own reporting requirements.
Establish the reporting basis first
The reporting owner identifies the legal entity, reporting date, units and relevant line definitions before extracting data. Both example lines use INR lakh at the same reporting date. A figure prepared in rupees or from a different cutoff could produce a difference that looks like a financial exception but is actually a basis mismatch.
The team retains mapping version MAP-v3 and identifies extracts SRC-A and SRC-B. This creates a stable basis for repeating the reconciliation. For real returns, use the applicable RBI directions and instructions, including any relevant amendments and reporting definitions.[2] Do not assume that a familiar field name has an identical meaning across returns or institutions. A reviewer should be able to determine whether the source population is complete for the stated perimeter before spending time investigating individual amounts. Otherwise a perfectly calculated workbook can still omit part of the required population.
Build a visible reconciliation bridge
LINE-01 begins with a source amount of 1,000 and a signed adjustment of minus 20, giving an expected amount of 980. The first draft contains 970, leaving a difference of minus 10. LINE-02 starts at 250, with an adjustment of minus 10 and an expected amount of 240. Its draft amount is also 240.
The expected total is therefore 1,220, compared with a draft total of 1,210. The workbook calculates expected amount as source plus signed adjustment and difference as draft minus expected. It keeps those steps in separate columns so the reader can inspect the arithmetic without decoding a long formula. A negative difference means the draft is below the expected figure under this convention. The sign convention, unit and comparison basis should be stated clearly because another team may otherwise interpret the same number as an excess rather than a shortfall.
| INR lakh | Source | Adjustment | Expected | Draft | Difference |
|---|---|---|---|---|---|
| LINE-01 | 1,000 | −20 | 980 | 970 | −10 |
| LINE-02 | 250 | −10 | 240 | 240 | 0 |
| Total | 1,250 | −30 | 1,220 | 1,210 | −10 |
Distinguish a source correction from a reporting adjustment
The example’s minus 20 adjustment reverses a duplicate amount and is supported by ADJ-01. The minus 10 adjustment for LINE-02 addresses a cutoff issue under the invented case definition and is supported by ADJ-02. Both references identify the reason, preparer, authority and affected line. The adjustment column is not a balancing field to be filled until the total agrees.
The owner should determine whether the underlying source can be corrected or whether a controlled reporting adjustment is necessary for the current cycle. A temporary adjustment may solve the immediate reporting calculation while leaving a recurring source defect unresolved. Record who will correct that defect and how subsequent periods will be checked. Retain the original extract so the reviewer can understand the starting position. Overwriting the source with the adjusted figure removes the distinction between what the system produced and what the reporting process changed.
Make differences traceable to an exception
The minus 10 difference on LINE-01 becomes exception EX-01. The exception log identifies the line, effect, owner and target date, with a status of Open. The reporting preparer investigates whether an approved adjustment was omitted from the draft. The reviewer records that release is held pending resolution rather than accepting a general explanation that the difference is small.
A useful exception record distinguishes an identified cause from an untested hypothesis. At this stage, an omitted adjustment is a possible explanation that needs evidence. The owner must trace the affected records, verify the correct treatment and assess whether similar lines or periods could be affected. The supporting file should explain why the eventual correction is appropriate. Without that discipline, an exception log can become a list of unexplained plugs that merely moves uncertainty away from the main reconciliation sheet.
Check lines as well as totals
The example has a visible net shortfall, but an equally important test is whether offsetting errors could cancel at total level. Suppose one draft line were overstated by 10 while another were understated by 10. The overall total would agree even though both line classifications were wrong. Line-level review therefore remains necessary after a total reconciliation passes.
The workbook provides a tolerance field for each line and requires an explicit value, including zero where appropriate. That tolerance is an institutional review parameter, not a permission to submit inaccurate information. A blank tolerance or missing source amount produces a Check inputs prompt. Entering zero is meaningfully different from leaving a required value unknown. Basel risk-data guidance provides context for accurate aggregation, reconciliation and controlled manual processes, but its scope and domestic application must be considered separately.[1] The fictional bridge illustrates a control technique, not a complete implementation of that framework.
Preserve independent challenge
The reporting preparer explains the calculations and supporting adjustments. The reviewer challenges population completeness, mappings, adjustment authority and unresolved exceptions. Merely checking that cells add up is a limited arithmetic review. The reviewer also needs to understand the definition that the calculation is intended to implement and whether the evidence supports the chosen treatment.
The review reference should point to the questions raised, responses received and resulting decision. If the reviewer requests a correction, the final record should show what changed between drafts. Where a small team uses compensating review arrangements, those arrangements need to be explicit and credible for the risks involved. A name entered in the Reviewer column is not evidence of a completed review. The workbook’s prompts can identify missing fields, but they cannot determine whether the reviewer exercised suitable judgement or had sufficient independence from preparation.
Resolve arithmetic and workflow separately
If investigation confirms that LINE-01 should be 980, changing its draft amount from 970 to 980 removes the arithmetic difference. The expected and draft totals would both become 1,220. However, EX-01 remains open until the supporting resolution is recorded and reviewed. The workbook intentionally does not turn a zero difference into automatic exception closure.
This separation matters because a number can be made to agree for the wrong reason. The reviewer should see the corrected draft, the evidence supporting the change and the assessment of any wider impact. Only then can the authorised release decision be made. In the saved fictional example, the first draft and open exception remain visible to teach this point. Users can change the number to test the arithmetic, but should still complete the exception record and follow their actual institution’s review process before treating the work as finished.
Control the final version and submission evidence
After the review is complete, the reporting owner identifies the approved release version and the authority that approved it. Working files, draft exports and approved returns should be distinguishable by a controlled naming or version process. The final file should be preserved with the reconciliation, adjustment evidence, reviewer record and resolved exception log.
Submission creates a further evidence requirement. Retain the acknowledgement or other reliable proof of the actual submission outcome, including relevant timestamps and identifiers. A successful file generation or an attempted upload is not necessarily evidence that the authority received an accepted return. If the submission is rejected, record the reason and corrective action. If a later amendment is required, link the revised submission to the original rather than replacing history. The exact correction and notification obligations depend on the actual return and applicable instructions, which the institution must verify.
Adapt controls to the institution and data flow
A bank may consolidate information across products, branches and several source systems. Its reconciliation design should identify the owner of each source and where transformations occur. An NBFC may rely more heavily on portfolio extracts and controlled spreadsheets, but those files still need clear definitions, access arrangements and review evidence. Complexity changes the design, not the need to explain the final amount.
The two-line example deliberately avoids assuming that every regulatory balance should equal the general ledger. Some differences arise from legitimate scope or measurement distinctions, which need a documented bridge. Others reveal errors. Reviewers should understand the difference rather than demand agreement between incompatible measures. Also consider whether one correction affects related reports or management information. A reconciliation that is accurate in isolation may still be inconsistent with another output drawing from the same records, especially when separate teams apply manual adjustments.
Use the toolkit to rehearse a complete cycle
The free workbook includes blank and fictional reconciliation sheets, exception logs and an evidence checklist. Start with a small set of material lines, establish the reporting basis and enter the source, adjustment and draft amounts. Inspect the line checks, then investigate differences through the exception log. Preserve supporting evidence outside the workbook in the institution’s controlled repository.
Test the process with a missing input, an unexplained adjustment and an offsetting difference before relying on it. The phrase Arithmetic and review recorded describes a limited field and tolerance check; it is not a regulatory approval. The useful outcome is a record that lets a second person reproduce the bridge and understand the release decision. When the source, adjustment, exception and final version remain connected, the next reporting cycle starts with clearer definitions and fewer unexplained carry-forward items.
PRIMARY REFERENCES
Sources & further reading
Primary-source references checked on 13 October 2026. Follow the relevant authority’s current text for applicability, amendments and commencement dates. International standards provide context and are not automatically identical to domestic requirements.