Duplicate invoices: why exact-match checks miss most of them

We planted seven duplicate payments in a year of accounts payable history: 583 invoices, $1.27 million paid. A check for the same vendor, invoice number and amount, the test many accounting systems run before payment, found one of them. The other six were the same invoices entered a little differently: a dropped prefix, two digits swapped, the supplier set up twice.

Duplicate invoices rarely look identical by the time they reach the payment run. Finding them takes a check that expects the differences.

What a duplicate invoice payment is

A duplicate payment is paying a supplier twice for the same thing. Usually it happens because the same invoice was entered twice, in slightly different forms, or because the supplier exists twice in the vendor master and the invoice was paid under both records.

Some duplicates come back without anyone asking: the supplier notices the extra payment and refunds it or holds it as a credit. Many don't, especially when the supplier's own records are messy too. And none come back until someone has spent time finding and claiming them.

How the same invoice gets paid twice

In a multi-site business, an invoice can reach accounts payable several ways, and each is a chance to enter it again:

• It arrives twice. The supplier emails the invoice to AP, and the driver also leaves a paper copy at the site, which the manager forwards.
• It is resubmitted. A reminder or statement arrives before the first copy is paid, and someone enters it as new.
• The supplier exists twice. "Coastal Foodservice" and "Coastal Foodservice Inc" were set up by different people at different times, so an invoice already paid under one looks new under the other.
• It is keyed by hand. Leading zeros get dropped, two digits get swapped, or an amount of $4,815 is entered as $4,185.

None of this is new. In 1985, the US Government Accountability Office reviewed duplicate payments at a Department of Justice payment center and recommended steps "to help avoid forwarding more than one supporting document for the same goods or services" (GAO, AFMD-85-72). Invoices still arrive by more than one route, and they still get paid twice.

Seven duplicates, one caught

The Duplicate Invoice Scanner ships with a sample year of paid invoices from a restaurant group, with seven duplicates of different kinds planted in it. Here is each one, what differs between the two copies, and whether an exact check (same vendor name, invoice number and amount) would stop it.

Six of the seven would be paid twice by a system that only blocks identical entries.

How to find the ones that don't match exactly

Fuzzy duplicate detection works in three steps.

1. Normalize invoice numbers. Strip case, spaces and punctuation, common prefixes such as "INV" and "Invoice #", and leading zeros. INV-0010577 and 10577 both become 10577. Invoice #10412 becomes 10412.

2. Normalize vendor names and compare them. Strip punctuation and legal suffixes such as Inc, LLC, Corp and Co. "SYSCO Corp." and "Sysco" both become sysco. "The Coastal Foodservice Co." and "Coastal Foodservice Inc" both become coastal foodservice. Names that are still slightly different are compared with a similarity score (the scanner uses Jaro-Winkler), so near-identical spellings match too.

3. Look for near-identical numbers with the same amount. Two digits swapped (SY54321 and SY54312), one character missing or added, or one character changed. A single changed character is also what sequential invoice numbers look like, so the scanner only trusts that pattern when the two invoices are dated close together.

Each kind of match gets a confidence level, because some are near-certain and some need a person to look:

The Coastal pair scores 99% because, once "Inc" is stripped, the two vendor names are identical and the pair counts as an exact match.

Results on the sample year

The five high-confidence pairs put $9,525.80 at risk, 0.75% of everything paid that year. The two lower-confidence pairs add another $10,425 to check: the BrightTech invoice whose amount differs, and the Apex invoices with different numbers. Either could be legitimate. A supplier statement settles it in a few minutes.

Choose the time window carefully

The weakest signal, the same amount from the same vendor, needs a time limit. Linen, waste collection and rent are billed at the same amount every month. In the sample, a monthly linen contract of $1,450 produces no false alarms with a 14-day window. Widen it to 31 days and it produces eight, one for every pair of consecutive months. Keep the window shorter than your shortest recurring billing cycle.

Compare two invoices

Enter two invoices and see how each is normalized, whether an exact check would flag them, and how the scanner scores the pair. The defaults are the Sysco pair from the sample.

Compare two invoices · same rules as the full scanner

First invoice

Vendor

Invoice number

Date

Amount

$

Compared as: sysco · SY54321

Second invoice

Vendor

Invoice number

Date

Amount

$

Compared as: sysco · SY54312

Exact-match check

Lets it through

Fuzzy match

86% confidence

Why

Same vendor and amount, near-identical invoice number. $1,286.75 at risk.

Scan a year of invoices

Free tool

Duplicate Invoice Checker

Upload a year of paid invoices as a CSV. Every pair is checked with the rules above, results are ranked by confidence and amount, duplicate vendor records are listed, and the scanner drafts a recovery letter for each supplier.

Runs in your browser. No sign-up.

Open the scanner

Recovering duplicates, and stopping new ones

1. Scan a year of paid invoices. You need vendor, invoice number, invoice date and amount. Export from your accounting system or bank payment history.
2. Check each high-confidence pair against the supplier's statement. It confirms whether the supplier received two payments, and whether they already hold the second as a credit.
3. Claim it. Send the supplier both invoice numbers, dates, amounts and payment references, and ask for a refund or a credit against the next invoice.
4. Merge duplicate supplier records. Every vendor-name variant the scan finds is a future duplicate waiting to happen. Merge them, and require a tax ID or bank details check before a new supplier is created.
5. Take invoices through one channel. Ask suppliers to send invoices to one AP address, and tell sites not to forward paper copies the supplier has already emailed.
6. Pay against the order and the receipt. A duplicate invoice usually bills goods that have already been invoiced against the same purchase order and receipt. Matching each invoice line to an order and an accepted receipt, with a sensible three-way match tolerance, stops it before payment instead of after.

Then rescan every quarter. Duplicates that get through tell you which supplier, site or intake route needs attention.