SSolarc Labs
Practical Article 8 min read

Duplicate Invoice Payments: Why Exact-Match Checks Miss Real Duplicates

Published August 2026 by Solarc Labs

A practical AP control pattern for finding duplicate-payment candidates when invoice references, supplier records or document presentation differ, while preserving human review and source evidence.

Duplicate processing is a control problem, not just a duplicate-number problem

HMRC guidance for electronic invoicing explicitly includes prevention of duplication of processing among the internal controls a receiving organization should maintain. In practice, an exact match on supplier ID plus invoice number is useful but narrow. The same commercial obligation can arrive with punctuation changes, spacing changes, supplier-master duplication, a re-keyed reference, a resend, or a document that is associated with a statement or credit note.

Normalize for matching without overwriting the source

A duplicate check can normalize case, punctuation, leading zeros and common reference formatting for comparison, but the original invoice fields should remain independently inspectable. The normalized value is a matching aid, not a replacement for the source record. Keeping both layers makes false positives easier to review and preserves a defensible trail when two documents look similar but are legitimately different.

Use several deterministic signals to create a review candidate

Stronger duplicate detection combines signals such as normalized supplier identity, invoice-reference similarity, gross amount, tax amount, currency, invoice date, purchase-order context and prior payment state. No single signal should silently become a payment decision. The useful output is a bounded reason such as possible duplicate with the matched records and fields visible to the reviewer.

Credit notes, instalments and recurring invoices are where naive rules break

Two records with the same amount can be correct. A credit note can intentionally mirror an invoice amount with the opposite commercial effect. Recurring services can produce the same amount every month. Instalments can share an order context. A review workflow therefore needs document type, sign, period and surrounding evidence before escalating from similarity to a duplicate conclusion.

The safest automation stops before the consequential decision

Automated checks are good at narrowing a large AP queue to a smaller set of explainable exceptions. The final action still needs the exact source document, matched historical record and current payment state in view. That is the same boundary used across Solarc invoice workflows: extraction and deterministic signals can support review, but they do not silently authorize or reject a payment.

Primary sources

Sources used for this article