A bank feed is an automated connection that transmits a company's bank transactions and balances directly into its accounting or treasury software, replacing the manual download and import of bank statements.
A bank feed can use four main channels, and the choice depends on the bank, the company size and the data needed:
- EBICS: the secure corporate banking protocol used mainly in France, Germany and Switzerland
- ISO 20022 files: camt.053 for end-of-day statements and camt.052 for intraday reporting, progressively replacing the SWIFT MT940 format
- PSD2 account information APIs: open banking access under Directive (EU) 2015/2366, usually through an aggregator
- SFTP file drops: statement files deposited on a server, common with ERPs and older treasury systems
Bank feeds break for predictable reasons. PSD2 connections require the account holder to renew consent periodically, and a lapsed consent silently stops the flow. Overlapping sources create duplicates when a feed and a manual import cover the same days. Transaction labels are truncated, and a SEPA transfer carries only 140 characters of unstructured remittance information. Card acquirers settle in net batches that do not match individual sales.
A bank feed moves data, it does not reconcile it. The transaction arrives in the accounting system with a date, an amount and a label, and someone still has to decide which invoice, customer or expense it corresponds to. That decision is bank reconciliation on the treasury side and cash application on the receivables side.
Phacet starts where the feed stops. The agent that can reconcile bank transactions and detect unmatched flows matches each feed line with its invoice or ledger entry, reads truncated labels semantically, and raises the lines left unmatched as exceptions for review instead of leaving them in a suspense account.
