About Bank Statement Parser
Bank Statement Parser is an independent software product that turns a PDF bank statement from any bank, in any country, into structured JSON, CSV or Excel — as both a web tool and a REST API. It is built and operated by a small engineering team, and the infrastructure, OCR included, runs entirely on our own servers.
What we do
Bank statements do not come in a standard format. Every bank places columns differently, buries reference codes in the description, writes dates its own way, and some are scanned images. Our job is to collapse that irregularity into one consistent schema: date, description, amount, balance, reference and value date — verified and ready to export.
The engine has been tested against hundreds of real statement layouts from dozens of countries and directly recognizes over a hundred bank formats. When it meets a layout it does not know, it learns that layout automatically. Full technical detail is in the API documentation.
How the engine works
Three-tier extraction
Every PDF is first matched against a known bank layout — the fastest and most exact path. With no match, a generic table-structure parser kicks in, with OCR and a language model where needed. If that still falls short, the document goes to a learning queue and its layout is inferred automatically, so the next statement in that format resolves instantly.
OCR for scanned PDFs
Image-based statements with no text layer are processed with Tesseract OCR running on our own servers. Neither the images nor the extracted text are sent to any external service.
Balance-chain verification
Each extracted row is checked against the equation opening balance + transaction amount = closing balance. It is compared with printed subtotals, and results with no deterministic proof are cross-checked against an independent second engine. Every response carries a per-row `verified` flag and a document-level confidence score.
Who it's for
- Developers wiring statement data into their own products (REST API + webhooks)
- Accountants and bookkeeping teams processing hundreds of statements a month
- Finance teams doing lending / underwriting and spend analysis
- Anyone who just needs a PDF turned into clean CSV or Excel quickly
What happens to your files
When you upload a recognized format, the PDF is processed in server memory only, never written to disk, and wiped the moment processing finishes. An unknown format is held on encrypted disk until automatic learning completes (typically within 24 hours), then permanently deleted. Your statement content is never used to train any AI model — the only thing learned is the column layout.
Full security & data-handling detail →
An independent product
Bank Statement Parser is built and maintained by a small team, with no outside investment. The free plan is permanent (100 pages a month, no card required); paid plans and one-time page packs are processed through Paddle, and your payment details are never stored with us. Our roadmap is driven by user feedback — if a bank's format won't parse, report it from the dashboard and it is usually added within a few days.