Content authored by practitioners with experience at Amazon India, Intuit QuickBooks, and the Tata Group. Meet the team →
At 200 transactions per month, a finance team can reconcile in a spreadsheet. At 2,000 transactions, the spreadsheet becomes the problem — not the solution. A reconciliation engine exists for the scale where manual matching breaks down.
Core Components of a Reconciliation Engine
A reconciliation engine has four functional components that distinguish it from a spreadsheet or an ERP module:
1. Data ingestion layer: Accepts inputs from multiple sources — bank statements (MT940, CSV), TRACES Form 26AS downloads, GSTR-2B JSON, payment gateway settlement files — in their native formats. No manual reformatting.
2. Matching rules engine: Applies configurable matching logic — exact match on primary reference, tolerance match on amount, pattern match on known deduction types (TDS, MDR). Rules are configured per reconciliation type and per organisation.
3. Exception classifier: Items that do not match after all passes are classified by exception type — not simply flagged as “unmatched.” The classification determines which team member handles the exception and by what deadline.
4. Audit trail: Every match, every exception classification, and every manual override is logged with timestamp and user. The audit trail is queryable for any period — essential for statutory audit and GST officer inquiries.
Rules vs Patterns: How Engines Work
The Matching Pass Sequence
| Pass | Matching method | Criteria | Typical match rate |
|---|---|---|---|
| Pass 1 | Exact match | Reference number + amount + date | 60–70% of transactions |
| Pass 2 | Tolerance match | Amount ±1% or ±₹100, same reference | Additional 10–15% |
| Pass 3 | Pattern match | Known deduction (TDS 10% = net amount × 10/90) | Additional 8–12% |
| Pass 4 | Fuzzy match | Partial reference, same counterparty, close date | Additional 3–5% |
| Exceptions | Human review | All unmatched after Pass 4 | 5–15% |
A well-configured engine achieves 85–90% auto-match rates. The 10–15% exceptions are where human attention creates value — not in the matching itself.
Why Excel Breaks at Scale
The Structural Limitations
Excel breaks for reconciliation at scale for three reasons specific to Indian finance:
VLOOKUP collision on net-of-TDS amounts: When searching for ₹90,000 (a payment after 10% TDS), Excel will match any other entry of ₹90,000 — including unrelated transactions. A reconciliation engine matches on the combination of counterparty TAN + section code + quarter, not just amount.
No GSTR-2B native matching: Excel has no native connection to GSTN. Each month requires a manual download, a format conversion, and a formula-based comparison that breaks on any format change from GSTN.
No exception routing: Excel can flag a mismatch, but it cannot route it to the correct resolver — TDS mismatches need the accounts receivable team, ITC mismatches need the GST manager, bank exceptions need treasury. Routing in Excel is a manual process.
Multi-Pass Matching Explained
The multi-pass approach matters because different transaction types match on different criteria. A NACH batch credit matches on the NACH presentation date and total amount — not on invoice number. A TDS receipt matches on the deductor TAN and section code — not on the NEFT narration. A platform settlement matches on the gateway settlement ID — not on the bank credit date.
A single-pass matching engine with a fixed criterion fails on all three. A multi-pass engine applies the right criterion for each transaction type.
Variance Taxonomy in Reconciliation Engines
Named variances are the difference between a reconciliation engine and a raw matching tool. When an item cannot be matched, the engine classifies it:
- FEE_DEDUCTION: Platform MDR or commission deducted from settlement — expected, no action needed
- TAX_DEDUCTION: TDS deducted — expected, generates TDS receivable entry
- TIMING_DIFFERENCE: Amount correct, date differs — carry forward to next period
- AMOUNT_MISMATCH: Genuine discrepancy — escalate for investigation
- ROUNDING: Sub-rupee difference — auto-resolve per tolerance rule
A named variance is resolved in one step. An unnamed “unmatched” exception requires the reviewer to diagnose it from scratch — which is what most spreadsheet-based processes produce.
Choosing an Engine vs a Compliance Tool
A compliance tool (like a GST filing portal) helps you file returns. It does not help you match your internal purchase register against GSTR-2B before filing. The reconciliation layer lives between your internal data and the compliance portal.
Reconciliation software India that includes a configurable matching engine handles both the India-specific reconciliation requirements (TDS, GSTR-2B, NACH) and the data integration layer — connecting directly to ERPs like SAP, Oracle, or Tally via API or file export.
For organisations where bank reconciliation is the primary complexity driver, bank reconciliation software with a multi-pass engine handles the bank statement vs ledger matching with the same variance classification approach.
The Reserve Bank of India publishes guidelines on payment system data formats — relevant for organisations implementing MT940 or ISO 20022 integration with reconciliation engines.