Content authored by practitioners with experience at Amazon India, Intuit QuickBooks, and the Tata Group. Meet the team →
Reconciliation tools ship reports where the sum of matched, unmatched, and exception buckets does not equal the input total — a stray ₹5,000 lost to a rounding cascade, an orphan entry that fell out of the join because of a null key, or a currency-conversion off-by-one that no one notices until the statutory auditor foots the columns. The residual is small enough to escape casual review and large enough over a year to trigger a Section 128 books-of-account concern, a CARO 2020 statutory-dues disclosure exposure, or a qualified audit opinion on inability to substantiate. Most tools have no built-in check that refuses to publish when the numbers do not tie; they leave the discipline to the human reviewer and hope.
Before a reconciliation cycle publishes, the engine computes four totals from the same input universe — input total, matched total, unmatched total, exception total — and tests whether matched + unmatched + exceptions equals input to the paisa. This is money conservation as a hard gate: value that enters the cycle must appear in exactly one of the three output buckets. If the identity fails, the run is refused. The engine surfaces the specific failure — the residual amount, the identity violated, and where possible the input rows implicated in the imbalance. The finance team resolves the underlying data quality issue and re-runs. If the identity holds, the report publishes and the identity outcome is recorded in the audit trail alongside the reconciliation result. The same discipline applies to sub-identities within the cycle — bank versus book column tallies, currency-by-currency conservation on multi-currency cycles, and inflow-equals-outflow tests on closed-loop settlement flows.
Reconciliation cycles carry an input universe (bank statement, GL extract, settlement file, or their combination) with a canonical input total. The engine emits matched, unmatched, and exception buckets and runs the identity test before publish. Currency-scoped identities run per currency on multi-currency cycles; bank-versus-book identities run per side on BRS cycles; closed-loop identities test that inflows equal outflows on settlement or intercompany flows. The identity outcome is recorded in the evidence trail with the input total, each bucket total, the residual (which must be zero for publish), and a timestamp. The gate never runs after the report has published — the check is a pre-publish precondition.
A reconciliation cycle either publishes with a machine-verifiable statement that money conservation held (input equals the sum of matched, unmatched, and exceptions, to the paisa), or it does not publish at all and the finance team receives a specific, actionable failure describing what did not tie. Downstream reports — the CARO 2020 statutory-dues disclosure, the Section 44AB tax audit reconciliation, the GSTR-9C reconciliation, the year-end BRS pack for the statutory auditor — inherit that guarantee. Match-rate figures cannot silently drift because the denominators and numerators are cross-footed by the same identity check on every cycle.
A national auto-component manufacturer’s controller closes the January 2026 books on 5 February 2026 with a bank-versus-book settlement reconciliation covering ₹4,20,15,000 of collections across 312 dealer accounts and 47 marketplace payout cycles. The reconciliation report from her legacy tool shows a matched total of ₹4,18,50,000, an unmatched total of ₹1,60,000, and an exception total of ₹5,000. Match rate 99.62 percent. She almost signs off. Then she foots the columns. Matched plus unmatched plus exceptions comes to ₹4,20,15,000 — but only because she added them in a spreadsheet. The tool’s report does not show the cross-foot. Two months later, when the Section 44AB tax auditor arrives and pulls the January cycle out of the reconciliation archive, he notices that the matched total in the tool is actually ₹4,18,45,000 — a ₹5,000 difference that had crept in from a rounding cascade on a partial-payment split. The unmatched and exception buckets do not compensate. ₹5,000 of collections has vanished from the reconciled universe, and the auditor has to decide whether it is a Section 128 books-of-account concern, a CARO 2020 statutory-dues disclosure exposure, or both. This is what accounting-identity gates prevent. A reconciliation cycle where matched plus unmatched plus exceptions does not equal the input total is not a report; it is a data quality failure wearing a report’s clothes. Accounting identity gates reconciliation money conservation is the discipline that refuses to ship that failure.
The claim in one paragraph
Bank versus book must tie. Currency conservation must hold. Match rates cannot silently regress. Results that do not add up are refused, not shown. Every reconciliation cycle enters the engine with an input universe and a canonical input total. Before any report publishes, the engine tests whether the sum of the three output buckets — matched, unmatched, exceptions — equals the input total to the paisa. If it does, the report publishes and the identity outcome is recorded in the evidence trail. If it does not, the run is refused. The finance team is surfaced the specific residual — the rounding cascade, the orphan entry, the currency-conversion off-by-one — and forced to resolve the underlying data quality issue before a report goes anywhere. Money that enters a reconciliation cycle appears in exactly one output bucket; value cannot be silently lost, invented, or reclassified.
Why this matters for auditors and CFOs
The output of a reconciliation cycle feeds three of the most audited surfaces in an Indian company’s financial statements: the bank-reconciliation pack that supports the cash and cash-equivalents balance, the GST reconciliation that supports the GSTR-9C annual return, and the TDS reconciliation that supports the Form 3CD tax audit disclosure. Each of those surfaces is tested by a statutory auditor who foots the underlying reports. If the reports do not tie — if matched plus unmatched plus exceptions does not equal input — the auditor has three options. Request a corrective journal that flows into the audit-year P&L. Qualify the audit opinion on inability to substantiate. Or note the exception in the CARO 2020 report under Clause 3(vii) on statutory dues. All three outcomes carry cost — real P&L, real professional-liability exposure for the audit partner, and real signalling damage to the board audit committee.
Here is what breaks in most systems. A reconciliation tool takes in a bank statement of ₹4,20,15,000, runs its matching logic, and emits three totals. The matched total is computed as the sum of matched pairs’ amounts, with rounding at each line. The unmatched total is computed as the sum of items the matcher did not resolve. The exception total is computed as items flagged for review. None of the three is ever cross-footed against the input. If a rounding cascade drops ₹5,000 out of the matched total — or an orphan entry with a null counterparty ID falls out of both the matched set and the unmatched set — the tool publishes the report anyway. The controller does not notice unless she foots the columns herself. Six months later, the auditor does foot the columns, and the ₹5,000 becomes a discussion the audit committee did not want to have.
The Indian regulatory and accounting framework
Four provisions anchor money conservation as a legal obligation, not a housekeeping preference.
Companies Act 2013 Section 128 requires every company to keep books of account that give a true and fair view of the state of affairs of the company, with proper explanation of every transaction. A reconciliation report where matched plus unmatched plus exceptions does not equal input is by definition not giving a true and fair view of the settlements it reconciles — some value has been silently lost, misclassified, or duplicated. Section 128 is the statutory hook that makes the identity failure not merely embarrassing but non-compliant.
Rule 3(1) proviso to the Companies (Accounts) Rules 2014 requires every company using accounting software to use software that records an audit trail of each transaction and each edit. The Ministry of Corporate Affairs rule was made mandatory from 1 April 2023 and now covers every company across scale. An audit trail whose totals do not internally reconcile is not an audit trail — it is a log the auditor cannot use as evidence under ICAI SA 500. Accounting-identity gates produce a trail whose numbers cross-foot on every entry, which is the precondition for the trail being defensible.
Ind AS 1 paragraph 15 requires fair presentation of financial position, financial performance and cash flows. Paragraph 45 prohibits offsetting of assets and liabilities except where an Ind AS specifically permits — which means a reconciliation report cannot net a matched line against an exception line to make the totals tie after the fact. Paragraphs 46 to 49 on materiality and aggregation govern when a residual is small enough to ignore; in practice for statutory audits, any residual that is systematic (i.e., appears every cycle) is presumed material regardless of amount because it signals a control failure, not an aggregation choice.
CARO 2020 Clauses 3(vi) and 3(vii) require the auditor to comment on cost records and to name outstanding statutory dues (GST, PF, ESI, TDS, income tax, sales tax, service tax, customs, excise, VAT, cess) more than six months old as at the balance sheet date. If the reconciliation of statutory-dues payments to the GL does not cross-foot, the CARO 2020 disclosure cannot be relied on — and the auditor must either qualify or expand the disclosure.
Section 44AB Income-tax Act (retained in scope under the Income-tax Act 2025) requires tax audit with Form 3CD, which reconciles turnover, TDS deducted and deposited, and GST turnover. Every Form 3CD line depends on an underlying reconciliation that must foot. When it does not, the 3CD disclosure is qualified.
The through-line across all four is the same: the numbers in the report must add up. Money conservation is not a nice-to-have; it is the statutory precondition for the report being usable as audit evidence.
A worked example
An illustrative Indian auto-component manufacturer with plants in Pune and Chennai runs monthly reconciliation of dealer collections. In January 2026 the cycle universe covers ₹4,20,15,000 of settlements — a mix of NEFT credits from 312 dealer accounts, marketplace payouts from a B2B platform, and a small volume of counter cash deposits at the Pune branch. The bank statement total for the settlement account, foot-summed, is ₹4,20,15,000. The book side — the dealer collection ledger extract for the same period — foots to the same ₹4,20,15,000. Input to the reconciliation engine is ₹4,20,15,000. The engine runs the matching logic and emits three bucket totals.
Illustrative — figures are representative of the operating pattern, not actual company data. Cross-verify against your own bank statement and GL extract before action.
| January 2026 reconciliation — identity check | ₹ |
|---|---|
| Input total (bank statement side, foot-summed) | 4,20,15,000 |
| Input total (book side, foot-summed) | 4,20,15,000 |
| Matched pairs total | 4,18,50,000 |
| Unmatched — bank items with no book counterpart | 1,00,000 |
| Unmatched — book items with no bank counterpart | 50,000 |
| Exceptions — flagged for review (dispute, evidence, timing) | 15,000 |
| Sum of three output buckets | 4,20,15,000 |
| Residual (input minus output-bucket sum) | 0 |
| Identity outcome | PASS |
| Match rate | 99.61% |
The identity holds. Matched plus unmatched plus exceptions equals input, to the paisa. The report publishes. The evidence trail records the input total, each bucket total, a zero residual, and a timestamp. The Section 44AB tax auditor who later foots the columns finds the same numbers the controller signed off on.
Now consider the counterfactual — an alternate January cycle where the exception bucket comes out to ₹10,000 instead of ₹15,000. Matched plus unmatched plus exceptions is now ₹4,20,10,000 against an input of ₹4,20,15,000. Residual ₹5,000. The identity fails. In a legacy tool, the report ships anyway. In a system with accounting-identity gates, the run is refused. The finance team is surfaced the failure — “January 2026 cycle: matched + unmatched + exceptions = ₹4,20,10,000; input = ₹4,20,15,000; residual ₹5,000” — and pointed to the specific input rows implicated. Investigation traces the residual to a partial-payment split where the tool rounded ₹12,345.67 down to ₹12,345 on the matched line and did not flag the ₹0.67 orphan. Multiply across 741 partial-payment splits in the cycle and the residual is exactly ₹5,000. The corrective action is to fix the rounding logic — see paise-exact decimal handling and the CA rounding convention for the specific half-up rounding discipline — re-run, and the identity now passes at zero residual. The report publishes with a match rate that the auditor can foot without finding a ghost.
What this means for you at each altitude
- CFO. No silent variance drift year over year. When the audit committee asks whether the 99.61 percent match rate on last quarter’s settlements is comparable to this quarter’s 99.58 percent, the answer is yes — the two rates share the same denominator discipline, both cycles passed money-conservation identity, and the small quarter-over-quarter movement is a real business signal rather than an artefact of rounding.
- Audit partner. CARO 2020 Clause 3(vii) disclosures based on TransactIG output cannot be undermined by rounding fabrication. When the audit team foots the reconciliation, the numbers tie. When Ind AS 1 fair-presentation is tested, the reconciled pack is internally consistent. The audit trail produced under Rule 3(1) proviso is defensible as audit evidence under SA 500.
- Integration engineer. Reconciliation outputs downstream of the engine — the GL posting file, the GSTR-2B/3B feed, the TDS remittance schedule — inherit the money-conservation guarantee. When an integration test asserts that the sum of matched lines in the output file equals the total from the reconciliation report, it always holds. See the developer envelope specification for the specific output-side contract.
- Operations lead. Exceptions get worked to zero, not to convenient. When the engine refuses to publish a cycle because of a ₹5,000 residual, the ops team cannot round-trip the residual into an ad-hoc write-off; the underlying data quality issue is fixed at source, and the fix compounds across every subsequent cycle.
- Board audit committee. The tone of the exception discussion changes. Instead of the auditor surfacing residuals the tool did not catch, the ops team surfaces the residuals themselves — before the report goes to the auditor. The committee sees a system where the cycle is either clean or explicitly refused, never quietly wrong.
How TransactIG delivers this in practice
Every reconciliation cycle TransactIG runs computes the input total from the bank and book universes, emits the matched, unmatched, and exception buckets, and tests money conservation as a precondition of publish. If matched plus unmatched plus exceptions equals input to the paisa — and, on multi-currency cycles, if the same identity holds per currency — the report publishes and the identity outcome joins the machine-readable evidence trail that supports the why TransactIG case for Indian reconciliation software. If the identity fails, the run is refused. The finance team receives a specific, actionable failure describing the residual and the input rows implicated, resolves the underlying data quality issue — a rounding cascade, an orphan entry, a currency-conversion off-by-one — and re-runs. The deterministic reconciliation property means the second run produces the same numbers as the first once the underlying data is fixed, so the fix is verifiable. Match-rate figures the customer publishes to the board, the audit committee, and the statutory auditor cannot silently drift, because the denominator and the numerator on every cycle are cross-footed by the same identity check.
Five FAQs
The five FAQs above address the operational questions Indian controllers and audit partners ask most often when evaluating whether a reconciliation system enforces money conservation as a gate rather than as an aspiration.
Content authored by practitioners with experience at Amazon India, Intuit QuickBooks, and the Tata Group. Meet the team →