Comment by jonahbenton

I have run into issues where CSVs are not correct/aligned with PDF statement data that banks are legally obligated to provide. In addition CSVs almost never have balance data. So I download the PDFs and extract data from them. This is much more painful than it needs to be- providing a sensible machine readable PDF involves just following a few simple rules to ensure the 3 or 4 transaction fields, and the few needed statement fields (dates and balances) are extractable without fragile heuristics. There is no conflict between branded and human readable vs machine readable.