How would an insurer premium receipt with a vague reference appear in generated data?

A general insurer receives premium receipt `RCPT-8820` for 1,240.00 through a bank transfer with the reference field reading 'renewal'. A broker holds three renewing covers: `COVER-44091` (1,240.00), `COVER-44107` (1,240.00) and `COVER-44112` (860.00). Two match the amount exactly.

The premium accounts officer cannot choose between them from the receipt alone. The broker's earlier remittance advice, `REM-2093`, listed `COVER-44107` and `COVER-44112` as the covers being paid that day, and it is linked to the same broker account. The officer applies `RCPT-8820` to `COVER-44107` and leaves `COVER-44091` open.

**State changes:** the receipt moves from suspense to applied, the renewal status of `COVER-44107` changes from pending payment to confirmed, and an audit event records the remittance advice as the evidence used. Had the receipt been applied to `COVER-44091`, the wrong cover would have been confirmed and the other would have lapsed at the grace date.

**What it teaches:** a dataset that includes supporting documents and links lets a model learn to look beyond the amount. For evaluation, the tie between two equal-amount candidates is the useful test: does the system find the advice, or does it guess?