Bank Reconciliation Methodology
Every discrepancy surfaced by reconcilebankstatement comes from explicit matching rules and arithmetic balance equations. This page documents how transactions are normalized, paired, and classified across reports.
1. Parsing & Flow-Direction Normalization
Before comparing lines, both documents are parsed into structured transactions containing date, description, reference, debit, credit, absolute amount, flow ("in" or "out"), and optional running balance.
Because business ledgers sometimes export cash flows using bank statement conventions (where inflows sit in a credit column) and other times using standard double-entry book conventions (where cash receipts are debits), the summary explicitly records ledger_direction_convention (bank or book) so reversals can be diagnosed accurately instead of producing false failures.
2. Pair Ranking & Confidence Levels
- Exact Amount & Flow: Primary match candidates share identical currency magnitude (
amount) and matching cash direction (invs.out). - Configurable Date Window (
date_tolerance_days): Defaults to ±2 days and is adjustable from 0 days (strict same-day clearance) up to 31 days (helpful for delayed check clearing or cross-border settlement). - Description Similarity Scoring (
description_matching): When enabled, payee names, check/reference tokens, and memo strings break ties when multiple transactions share the same dollar/euro amount within the same week. Each matched pair recordsmatch_basisand confidence (high,medium,low).
3. How Differences & Anomalies Are Identified
Unmatched Bank vs. Unmatched Ledger
Every un-paired transaction stays separated by source with an explanatory reason so bank-only charges (wire fees, interest, NSF) are never mixed with book-only timing items.
Amount & Direction Mismatches
Surfaces near-matches classified explicitly as amount_mismatch, direction_mismatch (a debit posted as credit or refund entered backward), or amount_and_direction_mismatch.
Possible Duplicate Clusters
Groups repeated transactions on the same source side (bank or ledger) showing count, transaction_ids, and underlying lines so legitimate split payments are easily distinguished from duplicate imports.
Transparent Amount Corrections
Any OCR or delimiter cleanup adjustment is logged under amount_corrections showing the raw extracted string alongside the normalized figure.
4. Mathematical Balance Verification Checks
Whenever opening, closing, or running balances appear in either document, reconcilebankstatement evaluates three arithmetic proofs (passed, failed, or unverifiable when balance headers/columns are absent):
- Period Equation:
Opening Balance + Net Extracted Cash Flow = Closing Balance - Line-by-Line Running Balance: Checks sequential stated balances (
breaks_countand exact line indices when an OCR row drop causes a discontinuity). - Book-to-Bank Bridge: Tests whether closing balance differences equal explained unmatched/mismatch timing items, flagging any residual
unexplained_difference. Review our Data Accuracy Policy for guidance whenunverifiableappears.
Related Documentation & Trust Resources
Ready to compare your bank statement against your ledger?
Upload PDF, image, CSV, or XLSX files directly in the workbench or test immediately with our sample CSV pair.