Turn Historical Bank Statements Into Cleanup-Ready Rows

Cleanup starts from whatever survived. Digital PDFs from a portal, scans of paper statements, archived months a client downloaded years ago, statements from accounts that no longer exist — a set that is inconsistent in format and incomplete in coverage.

StackExtract converts those documents into structured transaction rows. That is the claim and the whole of it: it turns the files into data you can work with, and the reconstruction of the books is the engagement.

Statements from different institutions and different years will not share a layout, and StackExtract does not require them to. Each document is read on its own terms, which is why the review screen is per document rather than a single batch confirmation: the thing you are checking differs from file to file.

The review screen exists so that the conversion is not something you have to take on trust. Every extracted cell is visible and editable before export, and the rows a check flagged are marked.

Where the statement supplies opening and closing balances, the reconciliation check runs against them and supports the review by telling you whether the transactions account for the movement. Where the statement does not print those figures — and on some card summaries and business formats it does not — the check is reported as unable to run rather than as passing.

This is deliberately the opposite of an opaque conversion. A tool that handed you a spreadsheet with no indication of what it was unsure about would be asking you to re-read the statement anyway, which is most of the work you were trying to avoid.

And nothing is adjusted on your behalf. A figure that disagrees is shown as printed and the row is flagged, which on a cleanup engagement is the only useful behaviour: the discrepancies are what you are there to find, and a tool that resolved them quietly would be hiding the work.

Difficult files are labelled rather than silently handled. A scanned or photographed statement has no text layer, so it goes through the scanned-document OCR beta and is marked as scanned — recognition can produce a plausible wrong character where reading would have produced nothing, and you should know which path a result came from.

For a document that is worth a person rather than more of your own time, Assisted Review is available as a paid add-on on that document. A reviewer checks the flagged fields against the source and reports what they found.

What that review is not is verification of identity. A reviewer can tell you a figure was misread and correct it; nobody can confirm from the document that an account number belongs to whom it says, because the document is the only evidence and it is what is being checked.

Exports are CSV and XLSX, plus QuickBooks-ready CSV for bank statements, and the edits you made in review are in the file.

There is no direct sync with accounting software. You download a file and import it where it needs to go, which keeps the moment data enters the ledger a decision rather than a side effect.

Edits made in review are in the exported file, and each PDF is one document whatever its page count. A cleanup engagement is usually a long tail of long statements, which is the case where per-page metering and manual retyping both cost the most.