AI transaction matching

Getbeel's matching engine finds connections between bank transactions and invoices. Every match comes with a confidence score and clear reasoning.

Match Scenarios

The AI engine classifies every transaction into one of three scenarios:

• Exact match — A bank transaction is matched to an existing invoice. The most common outcome when your records are up to date.
• Missing invoice — A payment appears in the bank statement but no corresponding invoice exists. This could indicate a missing or uncreated invoice.
• Missing payment — An invoice exists but no matching bank transaction was found. This may mean the payment hasn't been received yet or was missed during import.

Transaction Types

Getbeel recognizes the following transaction types from your bank statements:

• Debit — Outgoing payments and expenses
• Credit — Incoming payments and deposits
• Transfer — Internal transfers between accounts
• Fee — Bank charges, service fees, and commissions
• Interest — Interest income or charges
• Other — Any transaction that doesn't fit the above categories

By default, the matching engine filters to debit transactions only (expenses), since these are the most common reconciliation target. You can switch to all transactions to include income and other types.

Exact Matching

The first pass looks for perfect correlations:

• Precise amount — Transaction and invoice amounts are identical
• Same-day dates — Transaction date matches the invoice date exactly
• Identical references — Reference numbers on the transaction and invoice match

These get the highest confidence scores.

Fuzzy Matching

When exact matches are not found, fuzzy algorithms handle:

• Amount proximity — Partial payments, rounding, bank fees, and small discrepancies
• Date flexibility — Matches transactions within a 7-day window to account for processing delays
• Vendor name similarity — "Amazon Web Services" vs "AWS", "Acme Corp." vs "ACME CORPORATION"

Confidence Scoring

• High — Strong match across multiple signals (amount, date, vendor)
• Medium — Good on some signals, weaker on others
• Low — Possible match, needs manual check

Match Reasons

Each suggested match includes a match_reasons array with detailed text explanations. For example:

• "Amount matches exactly ($1,234.56)"
• "Transaction date is within 3 days of invoice date"
• "Counterparty name is similar to vendor: Acme Corp. (92% similarity)"
• "Reference number partially matches invoice number INV-2026-0042"

These reasons are displayed in the review interface so you can understand exactly why each match was suggested.

Amount Differences

When amounts don't align perfectly, the difference is highlighted — helping you spot bank fees, partial payments, or currency conversion gaps.

Multiple Candidates

When several invoices could match one transaction, all candidates are ranked by confidence. You pick the best one.