Duplicate detection: no document is booked twice

A duplicate document is not carelessness, it is a natural consequence of how documents circulate. A supplier sends an invoice by e-mail and a week later sends it again as a reminder. A business owner photographs a receipt at the till and sends it again in the evening to be safe. A client hands the accountant a folder half full of documents already sent along the way.

The consequences are not cosmetic: an expense booked twice means overstated costs, a wrong VAT deduction and, in an audit, an assessment with penalties. And watching for duplicates by hand is exactly the kind of work in which a person makes mistakes after hundreds of documents and a machine does not.

The key property of a good duplicate check is that it does not rely on the sender's behaviour. Clients do not have to think about whether they have already sent a document, or by which channel. The system compares the documents themselves, so it catches the same receipt regardless of the form and channel it arrived in the second time.

The second key property is that a person has the last word. Two legitimate identical purchases in a row (two identical coffees, two identical tanks of fuel) look like a duplicate but are not. That is why a suspected document is never discarded automatically; it is flagged and waits for confirmation.