Smart Bank Reconciliation: How the Matching Logic Works

Short answer: we match by amount first, then by reference, then by date window, then by counterparty. If three of those four agree, we auto-accept. If two agree, we suggest. If one agrees, we park it.
The four signals we score
- Amount. Exact match wins. A ±0.50 PKR tolerance handles bank rounding.
- Reference. Invoice number, cheque number, or our generated tag — whichever the bank printed.
- Date window. Default is ±5 banking days from the source document.
- Counterparty. Vendor or customer name matched fuzzy (the bank loves abbreviations).

The matching ladder
- 3+ signals agree → auto-accept, green row, ready to post.
- 2 signals agree → suggested match, you click once.
- 1 signal agrees → parked in the Suspense queue with the top three candidates.
- 0 signals → "create new" button — the line becomes a fresh expense or receipt.
Why we don't just AI it
Auditors want a rule they can read. Our matching engine prints the score and the reason right next to every accepted line: "matched by amount + reference + date window". You can defend it in 10 seconds. An AI-only system that says "I'm 92% sure" doesn't pass that test.
Tips that actually save time
- Use your invoice number in the bank reference. One auto-match per line, every time.
- Set per-vendor defaults. If HBL → Acme always means "pay vendor bill X", save the rule once.
- Run reco weekly, not monthly. Five minutes a week beats a six-hour close.

Want the deeper how-to? Read the Smart Banking step-by-step, or peek at the full features list. Pricing is on the pricing page.
FAQs
What if two invoices are the same amount?
We show both as candidates and rank by date proximity. You pick once and the rule sticks for next month.
Can it learn my matching habits?
Yes. Accept a match three times for the same vendor and ONE ERP starts auto-applying it — with the audit trail intact.
Does it work with scanned PDF statements?
Drop the PDF in. We OCR it, parse the columns, and feed it into the same matching engine.