Bank statements are written for machines. We translate them for people.

StatementDecoder started with a familiar moment: a charge on a statement from a name that meant nothing — which turned out, after twenty minutes of searching, to be a coffee shop billing through a payment processor. Statement descriptors are written for settlement systems, not for the people who have to read them.

So we built the translation layer: a database of thousands of merchants, banks and processor prefixes, matched against your statement lines with honest confidence levels — backed by AI extraction for the PDFs that defeat rule-based parsers, and AI research for descriptors nobody has catalogued yet. Every identification the AI makes is checked into the database, so the answers compound.

On top of identification sits the analysis people actually want from a statement: subscriptions with real yearly costs and renewal estimates, free trials that quietly converted, duplicate charges, price rises, and bank & FX fees — plus clean exports to CSV, Excel, QIF and OFX for the spreadsheet-and-bookkeeping half of life.