Bank statement reconciliation,
back to its source.

The same rows now show a chain break.

Consecutive statements have to chain: each expected closing balance must appear among the balances the bank printed. Where they fail to join, banktrail reports the break rather than the total.

• A break is drawn between the two rows it separates, with the date window and the amount nobody can account for. The workbook’s “diff 64,500??” becomes 64,500.00 unaccounted for between 2025-02-06 and 2025-02-14.
• A break has one of two causes. Either an unread row — a row the bank printed that banktrail could not read — or a statement nobody sent. The first break here is an unread row: the cell printed as #####, which stays in the unread list with its file, its line, and the reason it was refused. The second is a missing statement, and no document on hand covers 2025-02-28 to 2025-04-04.
• The row struck out in the workbook is still here. Two payments with the same amount and payee on the same day are two payments, not one. The statement printed both at different positions, and banktrail treats each position as a separate transaction. A parser that discards rows produces a ledger that looks complete but is short. That is the most expensive failure mode, because the output carries the authority of every row that parsed.

A matching balance is only one check. Read the guide to checking for missing bank statements for the full source inventory and coverage steps.

Link your accounting entries to the bank record

Create named ledgers for purchases, sales or other accounting records. Add entries by hand or import CSV, then compare them with bank statement transactions.

• Use the same review tools on every entry. Keep tags, comments, supporting images and edit history with accounting entries. Search descriptions, references, tags and notes within one ledger or across all your accounting ledgers.
• Find and confirm a match. Narrow candidates by text, currency, amount and date range, then record the relationship with a note. You can start from an accounting entry or a bank transaction.
• Follow the link in either direction. Open the bank transaction from its accounting entry, or the accounting entry from the bank transaction. An entry can link to several bank transactions, and a bank transaction can link to several entries. Each reviewer can withdraw their own match.

Reimporting an identical accounting entry with the same external ID keeps the existing entry. A conflicting entry is flagged before the batch is imported.

Real questions, traced to the statement

Three guided reviews use synthetic company data. Each identifies its source files, shows the checks and exception, marks what the evidence can support, and names the next document or action. No client records are used.

Browse all three worked examples.

Every row parsed. The population was still short.

Getting rows out of statement files is the easy part. banktrail came out of a reconciliation across 284 bank statements from 6 banks. Asked how many transfers of one kind had been made, the first answer was eleven. Eleven was right — for one account, out of a population of three, over a date range everyone had assumed was continuous and was not. Every row that had been read had been read correctly, which is exactly what made the number so convincing. No statement parser we could find would have caught it. Accountants, auditors and controllers meet this constantly: every row parses, and the population is still short.

• money

  Money is always handled as Decimal.

  A ledger must foot to the minor unit. It cannot rest on a data type where 0.1 + 0.2 is not 0.3. Every amount is a Decimal from the reader to the ledger, and a string everywhere in between.

• dates

  Adapters declare the date order.

  An adapter is the component that reads one bank's statement format, and it declares that bank's date order. 03/01/2023 is 3 January in Karachi and 1 March in Chicago. Where a date is ambiguous and no order is declared, banktrail warns rather than picks.

• files

  File type is detected from the content.

  Banks send .xls files that are HTML documents and .xls files that are genuine BIFF. The extension does not tell them apart. banktrail reads the bytes and decides from those, which is the only safe way to identify a format.

• footing

  A statement must balance before it imports.

  The import is refused unless opening plus net movement equals closing. That one check catches a misread column and a defective adapter. Neither is imported.

• scans

  A scan is rejected until every token is accounted for.

  banktrail refuses an image-only PDF rather than parsing it to nothing. Reading one is a separate, deliberate step: OCR turns the pages back into positioned words, and a capture report then accounts for every token on the page that looks like money before the import is allowed. A partial read shows up in the report instead of disappearing into the total.

• identity

  Identical payments on the same day are two payments.

  Two payments with the same amount, payee and date are two transactions. banktrail identifies each by its position in the statement rather than by its contents, so both salary payments on the 14th survive an export that shuffles the rows.

• annotations

  Annotations record where they came from.

  Tags, comments and supporting images stay with bank transactions and accounting entries. Each match records who made it and how it was selected, with an optional confidence and note. Reviewers can follow the recorded relationship from either entry.

• audit

  The ledger downloads as a standalone file.

  A SQLite snapshot preserves accounting entries, annotations and bank matches alongside the statement evidence. An auditor can inspect the file with tools they already use, without relying on banktrail being online.

Accounts whose statements carry no balance column are reported as unverifiable, never as passed. Reporting an unchecked account as intact would claim a proof that was never run.