I've spent the last few months building a bank statement converter, and the thing that surprised me most wasn't technical it was that accuracy is basically unverified across the whole category.
Upload a statement to any of the popular tools and you get a tidy spreadsheet back. It looks correct. But if the parser dropped a row, misread a 3 as an 8, or lost a page, nothing tells you. You find out when a client does.
The fix turns out to be simple and nobody does it: the statement already contains the answer. Opening balance plus credits minus debits must equal the printed closing balance. If it doesn't, something was extracted wrong. It's a checksum that ships with every bank statement, and converters just ignore it.
Two things I'd genuinely like input on, especially from anyone doing bookkeeping:
1. When you get converted statement data, what do you actually check before trusting it? Do you spot-check rows, or eyeball the total, or just... not?
2. Where else does this pattern exist data with a built-in verification signal that tools ignore? Invoices have totals. Payroll has gross-to-net. Feels like a general gap.
I'm launching the thing I built on PH soon, so declaring the interest up front. But I'm more curious whether the "check the output" gap is as widespread as it looks from in here.
