Scan Bank Statements into Excel: Capture and QA
A paper bank statement needs two distinct operations: capture a legible document, then reconstruct its transaction rows. A clean-looking Excel file can still contain OCR errors, missing pages, or amounts assigned to the wrong descriptions. Preserve the scan and check the workbook against it.
If a bank-provided digital statement is available, download that before rescanning paper. It avoids an additional capture step, although any extracted data still needs review. This guide concentrates on paper acquisition and quality control, not accounting-software import.
Capture every page clearly
Before scanning, arrange pages in order and check the printed page count. Flatten folds, remove clips, and separate different accounts and statement periods. Include pages showing opening balance, closing balance, and transaction totals, even if those summaries will not become spreadsheet rows.
For a scanner, use 300 DPI as a practical starting setting, then inspect the actual text at normal reading size. Try grayscale for ordinary print and color if faint colored text disappears. These are starting choices, not accuracy guarantees; do not keep a small file at the cost of unreadable decimal points.
For phone capture, place each page flat, use even light, keep the camera parallel, and avoid shadows or glare across amounts. Crop excess background without cutting off dates, page numbers, or the outer transaction column. Review the captured image before moving to the next page.
Rotate sideways pages and correct severe skew before extraction. Save one ordered PDF per account and period, with a filename such as ExampleBank_1234_2026-09.pdf. Keep the unedited capture when making enhanced working copies.
Extract rows, not just text
OCR identifies characters; you still need the correct row and column relationships. Do not assume a line of recognized text equals one transaction. A wrapped merchant description may continue on the next line, while the next amount belongs to a different transaction.
Use an extraction tool that accepts your source and produces a reviewable spreadsheet. Open the converter to review its PDF-to-Excel option, then compare the downloaded rows with the scan. If text is unavailable or extraction misses a column, improve the capture or use controlled transcription rather than repeatedly exporting the same error.
For a fallback text step, Google Drive can convert PDFs and images through Google Docs, but its documentation warns that tables and columns may not transfer.[11] Tabula is intended for text-based PDFs, not raw scans.[12] Neither route should be described as a guaranteed scan-to-transaction conversion.
Worked example: illustrative OCR review
Assume a checking statement shows an opening balance of 1,000.00, a deposit of 900.00, and a withdrawal of 42.50. A checked working table would contain:
| Posted date | Description | Signed amount | Source page |
|---|---|---|---|
| 2026-09-02 | Client payment | 900.00 | 1 |
| 2026-09-03 | Office supplies | -42.50 | 1 |
The illustrative closing balance is 1,857.50. If OCR reads the withdrawal as 425.00, the calculated balance becomes 1,475.00, leaving a 382.50 difference. Return to the amount on the scan; do not change another row to force agreement.
A date printed as 03/09/2026 also needs interpretation. Confirm whether the source means September 3 or March 9. Keep a verified date column separate from raw OCR text until that question is resolved.
Check completeness and exceptions
Review every page, especially the last row before a page break and the first row afterward. Remove repeated headings and summary totals from the transaction table. Confirm that amount columns are numeric and that the chosen checking-account convention is receipts positive, payments negative.
Compare transaction counts, debit/credit totals, and opening-to-closing movement. Check suspicious digits and punctuation directly: a missing minus sign or decimal point can materially change a row. Matching balances are a useful control, not proof that every transaction is correct.
Mark unreadable dates or amounts as unresolved and obtain a clearer copy or bank replacement. Never estimate financial entries to fill an import file. Keep source-page references and reviewer notes so exceptions can be resolved later.
For type-safe workbook cleanup, see CSV to Excel. For already digital sources, use free CSV methods.
Official sources
[11] Google Drive: Convert PDF and photo files to text.
[12] Tabula: Official installation and extraction instructions.