From shoebox to spreadsheet: digitising paper bank statements

Some clients hand you PDFs. Others hand you a shoebox. Paper statements — posted originals, branch printouts, or the photos a client takes on their kitchen table — are still everywhere in bookkeeping, and typing them in by hand is the single most expensive way to process them. Here's the workflow that turns paper into verified rows.

Capture well and the rest gets easy

OCR quality is decided at capture time. Five rules:

• Flat and square — lay the page flat; shoot from directly above, not at an angle.
• All four corners in frame — cropped edges lose the columns that matter (amounts live at the margins).
• Daylight or bright, even light — shadows across numbers are the classic misread source.
• One page per image — don't photograph a spread of two pages.
• A scanner beats a phone if one's available; 150 DPI is plenty.

A "scan" can be a proper scanner PDF or just phone photos — both work; the rules are the same.

Convert, and demand proof

Upload the captured pages to NoRekey exactly as you would a native PDF — it detects that the document is a scan automatically and routes it through OCR. Two honest things to know:

Scanned pages cost more. A scanned page uses three pages of allowance instead of one, because image pages take genuinely more processing than text ones. You're told before the conversion runs, not after (the details).

Verification matters most here. Reading pixels is exactly where extraction errors live — a smudged 8 becomes a 3, a fold hides a row. This is why every NoRekey conversion, scans included, is checked against the statement's own opening and closing balances: if the extracted transactions don't walk the opening balance to the closing balance to the penny, the conversion is flagged needs review rather than delivered as if it were fine. With hand-typing — or with converters that don't verify — those errors surface weeks later as reconciliation mysteries. Here they surface immediately, labelled.

Then it's just data

From there the shoebox behaves like any other statement: export CSV for working papers, OFX for QuickBooks, CSV for Xero, and reconcile month by month. For a full pile spanning months or years, run it as a catch-up job — inventory first, then convert in batches.

When to rescan instead of squinting

If a conversion comes back flagged and the source photo is dark, angled or cropped, the fastest fix is almost always a better capture, not manual correction. Thirty seconds of re-photographing beats ten minutes of second-guessing a blurry column — and the balance check will confirm the moment the capture is good enough.