Multi-currency reconciliation is the process of matching flows recorded in different currencies, or converted at different rates, so that two records of the same transaction can be recognised as one. It is a harder problem than single-currency reconciliation for a structural reason: the two amounts being compared are almost never identical.
The gap has several origins. The invoice is issued in dollars and settled in euros. The bank applies its own rate and adds a spread. One side books at the transaction date rate, the other at the payment date rate. A bank charge is deducted from the amount received. None of these is an error, yet each produces a numerical difference on a flow that is otherwise perfectly legitimate.
This is why amount-based matching fails. A rule requiring exact equality rejects almost everything. A rule with a tolerance band accepts the small genuine discrepancies but also masks real errors that happen to fall inside the band, which is worse: a duplicate payment or a wrong rate slips through unnoticed.
A reliable reconciliation works on a combination of signals rather than a single amount: counterparty, date proximity, reference and label, original currency amount, and the plausibility of the implied rate. That is what semantic matching provides, and it is the difference between a rule engine and a system that recognises correspondence.
The point is not to match everything, but to classify what does not match. A residual explained by a rate is closed; a residual explained by nothing is an exception to investigate. Phacet applies that logic across bank reconciliation and multi-currency invoice control, each match documented by an audit trail.
