For forensic accounting
The analysis is yours. What an investigation needs from a converter is transaction data complete enough to reason about and traceable enough to stand behind.
The job
Statements across accounts, entities and periods have to become one dataset you can interrogate — matching transfers between accounts, finding patterns, establishing what moved where and when. The conversion is the foundation, and everything downstream inherits its errors.
Why extraction errors are corrosive here
An investigation reasons about relationships between transactions — a payment out of one account matching a receipt into another, a pattern of amounts, a sequence of dates. A single dropped or misread row does not just make one figure wrong; it can break a chain you were relying on, or manufacture a discrepancy that costs days to chase before it turns out to have been a conversion artefact all along. Both directions waste the same time.
What this does about it
Errors are localised, not aggregate
Where the running balance breaks, it breaks at the row responsible. That converts "these accounts do not tie out" into "row 47 of the March statement", which is the difference between a day and a minute when a discrepancy turns out to be an artefact.
Exact arithmetic, not floating point
Every amount is an integer number of minor units throughout, so totals across thousands of transactions do not accumulate rounding drift. The tolerance for a reconciliation mismatch is zero rather than a cent, which means a genuine one-cent discrepancy is still visible.
Provenance for every row
Page and line are recorded per transaction and shown on click. A figure in your working papers can be traced to the printed line it came from without going back through the PDFs by hand.
Nothing is silently corrected
Rows that do not reconcile are flagged and left as they were read. No balancing entry is invented, no amount adjusted to make totals agree, and the export carries a note stating what was and was not verified.
There is no fraud detection here
No anomaly scoring, no duplicate-payment detection, no pattern flagging. Reconciliation verifies arithmetic only: it cannot detect a wrong description, a wrong date, or two errors that cancel out. The investigative work is yours, and a converter claiming to do it would be claiming to know intent from a description line.
Try it on a statement
Three pages a day without an account. A digital statement is read in your browser and is not uploaded.
Drop a PDF bank statement here
Only pages with transactions count against this.
Review
Example statementHabib Bank Limited · 0441-79002251-03 · 2026-04-01 to 2026-04-30
- Row-by-row running balance
- Page subtotals
- Opening + movement = closing
- Printed credit / debit totals
- Printed credit / debit totals
- Transaction count
Arithmetic only. Verifies that amounts, balances and totals agree. Does not verify descriptions or dates, and cannot detect two errors that cancel out.
| Date | Description | Debit | Credit | Balance |
|---|---|---|---|---|
| Apr 02, 2026 | IBFT TRANSFER FROM ZAINAB TRADERS / INV-4471 | PKR 185,000.00 | PKR 1,247,350.00 | |
| Apr 03, 2026 | POS PURCHASE 4157 IMTIAZ SUPER MARKET KARACHI | PKR 12,480.00 | PKR 1,234,870.00 | |
| Apr 05, 2026 | UTILITY PAYMENT K-ELECTRIC ACCOUNT 04412297 | PKR 38,640.00 | PKR 1,196,230.00 | |
| Apr 08, 2026 | CHEQUE 000241 CLEARED — AL-NOOR PACKAGING | PKR 220,000.00 | PKR 976,230.00 | |
| Apr 11, 2026 | IBFT TRANSFER FROM MEHRAN ENTERPRISES / PO-8890 | PKR 342,500.00 | PKR 1,318,730.00 | |
| Apr 14, 2026 | SALARY DISBURSEMENT APRIL 2026 — 11 EMPLOYEES | PKR 685,000.00 | PKR 633,730.00 | |
| Apr 17, 2026 | BANK CHARGES — ONLINE TRANSFER FEE | PKR 1,150.00 | PKR 632,580.00 | |
| Apr 21, 2026 | IBFT TRANSFER FROM SIDDIQUI BROTHERS / INV-4488 | PKR 128,750.00 | PKR 761,330.00 | |
| Apr 24, 2026 | POS PURCHASE 9921 TOTAL PARCO FUEL STATION | PKR 18,900.00 | PKR 742,430.00 | |
| Apr 28, 2026 | PROFIT CREDITED — SAVINGS ACCOUNT | PKR 4,318.00 | PKR 746,748.00 | |
| Closing balance10 transactions | PKR 746,748.00 | |||
Export
Preview is free. Create a free account to download.
Guest plan · preview only, no downloads
Questions
- Can it detect a doctored statement?
- No. An altered statement whose balances were altered to match reconciles perfectly. Arithmetic consistency is evidence about extraction quality, not about authenticity, and it should never be offered as the latter.
- Can I get several accounts into one dataset?
- Convert each statement and combine the exports. There is no cross-account matching built in — that analysis is yours to run in whatever you already use.
- What does the export note say?
- Which checks ran, which passed, which were skipped because the statement did not carry the data, and the scope statement: arithmetic only, not descriptions or dates, and not errors that cancel out.