Scanning and storing an invoice is the easy 80% of invoice archiving. The valuable 20% is knowing which bank transaction actually paid it, without someone scrolling through a bank statement every month.
Why a simple amount match is not enough
A bare amount match breaks constantly. A marketplace payout bundles dozens of orders into one lump sum. A payment is slightly off because of a bank fee or a Skonto discount. Two unrelated invoices happen to share the same amount in the same week. A matching system that only checks “does this number appear in the bank statement” produces false matches and misses obvious cases.
What real matching needs
A priority-ordered strategy, not one check. Try to find the invoice or reference number itself inside the bank transaction’s description first — the strongest signal. Fall back to exact amount within a reasonable date window. Then a looser vendor-keyword match within a wider window. If nothing is confident, flag it for manual review. And catch net-vs-gross mismatches cleanly, so a Skonto discount does not mask a correct match.
What runs today in Vedron
Vedron’s invoice archive starts with OCR extraction on upload, with a fallback for scans that are difficult to read. Automatic bank matching then runs a priority-ordered strategy: extracted reference number against bank transaction descriptions first, then exact amount within roughly two weeks, then a vendor-keyword match in a wider window, then manual review if none of them are confident.
Common questions
What happens when nothing matches automatically? It is surfaced for a manual pick, not left silently unmatched. The goal is confident automatic matches where the evidence supports them, and a clear fallback where it does not.
Does this handle invoices in different currencies? The exact-amount strategy is EUR-specific; other currencies rely more on the reference-number and vendor-keyword strategies.
Can OCR handle a poor-quality scan? There is a fallback chain, so a scan that a regex extraction struggles with gets a second attempt through a more capable extraction method before being flagged for manual entry.
Talk to us about your setup
Book a short call. Beta clients get two months of custom setup on us.