Turn off the ones your team learns to ignore
Five fire on ordinary document handling, a merged PDF, metadata rewritten on save, the traces a common PDF library leaves behind. They are off by default, and any of them can be switched back on.
Bank statement underwriting starts with three questions about the document itself, asked before any figure is derived from it.



The document you already collected is checked in the four places an alteration shows: on the page, in the file structure, in the metadata underneath, and in the arithmetic. Each finding comes back named, and so does each clean one.
The same check runs at three scopes, and one edited balance fails all three.
A real document for the wrong person is a failure no authenticity check would find, so the record is read against what the application says.
Returned as verified, mismatched, or missing. Missing is its own answer, so a record that carries no name is never scored as agreement.
Provided against observed, reported as matched over total, so a partial match reads as partial rather than as a pass. A file whose latest transaction is 60 or more days old is marked out of date.
A detector that fires on a re-saved PDF is a detector people learn to ignore.
Five fire on ordinary document handling, a merged PDF, metadata rewritten on save, the traces a common PDF library leaves behind. They are off by default, and any of them can be switched back on.
A check that could not run is reported as unavailable rather than clean, so nobody funds on a question that was never answered.
They sit above the income, the obligations and the risk indicators, so an income figure is never read off a file whose document failed.
Want the analysis these checks run alongside?
Cash Flow AnalysisCollecting several documents per deal?
Automate stipsWhat a fraud or risk owner asks before a document check is allowed to hold up a deal.
Bank statements are the main one, the checks run on whatever got uploaded for the analysis. The detection also flags which document types it actually recognized in the file, so if something’s submitted as one thing but isn’t really that, you’ll see it too.
Yes, and part of what we’re checking is where that image actually came from. We can tell whether an embedded photo looks like it came straight from a real capture device, or whether it’s been re-encoded somewhere along the way.
We log the finding along with exactly what triggered it and which indicator it falls under, and it goes right at the top of the report, not buried somewhere inside. What that means for the deal is entirely up to your own rules, not ours.
Yes. We return trust indicators right alongside the risk ones, each with its own ID and description, so a file that passes shows you the evidence for why, not just an absence of red flags.
We report the gap as a number, tag it high, low, or none, and hand you the specific transactions that didn’t reconcile, broken out by account and by period. That way a reviewer’s looking at the actual lines instead of just taking our word for it.
Book a demo and we’ll open one document with you: what each of the three checks tests, what it returns, and where that answer reaches your team.