How to Detect Invoice Fraud in Factoring: The Five Patterns and the Controls That Catch Each
By Zolvo Team ยท 9 min read
Most of the money a factor loses is not lost to credit. It is lost to verification failures. A debtor that goes bankrupt is an underwriting problem you can model and reserve against. An invoice that was never owed, was already pledged to another lender, or was billed before the goods shipped bypasses your credit box entirely, and by the time it surfaces in collections the advance is gone.
Factoring is uniquely exposed because the asset you fund against is a piece of paper the borrower controls until you confirm otherwise. The borrower creates the invoice, names the debtor, sets the amount, and presents it for funding. If you advance before you independently verify any of that, you are trusting the most conflicted party in the deal. This playbook covers the fraud patterns that hit factors and the control that catches each one before funding, not after.
Why fraud detection is a funding-stage problem, not a collections problem
The economics are unforgiving. A factor advancing 80 to 90 percent against face value earns a fee measured in points and carries the full advance as exposure, so one fraudulent batch can erase the margin on dozens of clean clients. Industry practitioners consistently rank fraud among the leading causes of concentrated, catastrophic loss for factors, because it defeats diversification: a single bad actor can fabricate an entire receivables base, and the loss arrives at once.
The defense has to live at the moment of funding. Every control below answers one question before the wire goes out: is this receivable real, unencumbered, currently owed, and owed by who the client says it is? That is the job of an invoice verification layer.
The five fraud patterns that hit factors
1. Duplicate and double pledging
The most common and most dangerous pattern is pledging the same receivable more than once. Within a single factor it shows up as a duplicate: the same invoice submitted twice, sometimes with a tweaked invoice number or altered amount to dodge an exact-match filter. Across lenders it becomes double pledging, where a borrower assigns the same invoice to two funders who both advance against it. Each lender file looks clean in isolation, because the fraud is the second assignment. UCC filings are the legal backstop, but a search at onboarding does not catch an invoice double-pledged six months into the relationship.
The control: deterministic duplicate detection that normalizes invoice numbers, amounts, debtor identity, and PO references, then fuzzy-matches against your entire funded history rather than just the current batch. Pair it with notice-of-assignment discipline so the debtor pays you directly, and reconcile the lockbox so any payment landing elsewhere becomes a flag. This is where automated receivables automation earns its keep: a reviewer compares an invoice to the batch in front of them; software compares it to every invoice you have ever funded.
2. Fake and AI-generated invoices
A fabricated invoice has no underlying sale. Historically these were easy to spot because forgers got lazy: round numbers, sequential invoice IDs, mismatched tax math, a logo lifted from a website. In 2026 that tell is disappearing. Generative tools now produce invoices, purchase orders, bills of lading, and even matching email threads that are internally consistent and visually flawless. The document passes a human eyeball test, so it is no longer evidence of the transaction, only a claim; evidence comes from corroboration outside the document set the borrower controls.
The control: independent debtor confirmation through a channel the borrower does not own. A fabricated invoice survives document review but dies the moment you contact the debtor to confirm the obligation. Multi-channel confirmation matters because fraudsters plant fake contacts that route back to the borrower or an accomplice, so cross-check the debtor contact details against an independent registry and require confirmation from a verified domain or a known accounts-payable line. Layer in pattern checks (template reuse across supposedly unrelated debtors, impossible delivery timelines, duplicated line-item language) so the volume you cannot phone-verify still gets risk-scored. This is the core of invoice fraud detection: assume the paper is perfect and verify the transaction anyway.
3. Pre-billing
Pre-billing is the most defensible-looking fraud because the debtor is real, the relationship is real, and the invoice will eventually be legitimate. The borrower simply bills before the work is done or the goods have shipped to pull cash forward. It often starts as a cash-flow stopgap and escalates into a rolling deficit they can never close. These invoices verify as real because the debtor recognizes the borrower, so the fraud is in the timing, not the existence of the receivable.
The control: verify delivery, not just the obligation. Tie confirmation to proof of performance such as a signed delivery receipt, a goods-received note, or debtor acknowledgment that the work is complete and currently payable. Watch the leading indicators: rising dilution, credit memos issued shortly after funding, a lengthening invoice-to-payment lag, and concentration spikes in a single debtor. Continuous portfolio monitoring turns these slow-moving signals into alerts instead of a quarterly surprise.
4. Fake or colluding debtors
If a fraudster controls both sides of the transaction, every confirmation comes back positive because the same person answers the phone. Fake-debtor schemes range from a shell company with a rented mailbox to a genuine third party colluding for a cut, and the receivable looks fully verified right up until the payment that never arrives.
The control: verify that the debtor is a real, independent, creditworthy entity before you trust any confirmation from it. Check business registration and operating history, confirm the entity exists at a real address rather than a virtual office, and source the contact channel independently. Watch for structural tells: debtors that share an address, bank account, IP, or beneficial owner with the borrower; a single debtor that appears suddenly and dominates the book; payment instructions that route back to the client. A debtor that cannot be independently located should never clear for funding.
5. Altered and inflated invoices
The quieter cousin of fabrication is alteration. A genuine 4,000 dollar invoice is presented as 40,000, line items are padded, a real PO is reused, or amounts are nudged just under a review threshold. Because there is a real transaction underneath, document review tends to wave these through.
The control: confirm the exact amount and terms with the debtor, not just the existence of the invoice, and reconcile against the underlying purchase order and delivery record. Threshold logic that only inspects large invoices is the gap these schemes exploit, so risk-score on pattern (PO reuse, amounts clustering just below limits, terms that do not match the debtor standard) rather than size alone.
Building the control stack: assume the document is perfect
The unifying principle across all five patterns is simple. In 2026 you cannot treat the document as proof. It is a claim made by the most conflicted party in the deal, and the tools to make a fake claim look perfect are now cheap and widely available. Detection comes from corroboration the borrower does not control. A practical pre-funding stack has four layers:
- Independent verification of the debtor existence, independence, and creditworthiness before any single invoice is trusted.
- Multi-channel debtor confirmation of the obligation, amount, terms, and delivery, through channels the borrower does not own.
- Duplicate and double-pledging detection across your full funded history plus UCC and notice-of-assignment discipline.
- Continuous monitoring of dilution, concentration, payment timing, and lockbox reconciliation, so slow-burn fraud surfaces as it develops.
Most factors already do all four manually, which is why fraud review is the bottleneck that slows funding and why under-resourced teams quietly narrow their checks under volume pressure. The fix is not more headcount; it is automating the deterministic parts so reviewers spend judgment only on genuine exceptions. Confidence-scored, exception-based cash application and reconciliation close the loop after funding by catching payments that land in the wrong place, the earliest signal that a receivable was pledged elsewhere, and covenant compliance monitoring extends the same discipline to funder reporting so borrowing-base and concentration limits are tested continuously rather than at quarter end.
This is the layer Zolvo builds for commercial lenders. We automate multi-channel debtor confirmation, duplicate and double-pledging detection, and continuous portfolio and covenant monitoring, on top of the systems you already run such as FactorSoft, LoanPro, QuickBooks, and Plaid rather than replacing them. It is a verification layer that frees your team to underwrite, with SOC 2 Type II controls and a go-live of about two weeks. If you run a factoring or asset-based lending book, see how it maps to your workflow on our factoring use case, or talk to us about a pre-funding fraud review.
Frequently asked questions
What is the single most common type of invoice fraud in factoring?
Pledging the same receivable more than once. Inside a single factor it appears as duplicate submissions; across lenders it becomes double pledging, where a borrower assigns the same invoice to two funders who both advance against it. It is best caught by deterministic duplicate detection across your full funded history combined with UCC searches and notice-of-assignment discipline.
Can you still spot AI-generated fake invoices?
Not reliably by looking at the document, and that is the point. Generative tools now produce invoices, purchase orders, and supporting paperwork that pass human review. Detection has to shift from inspecting the paper to corroborating the transaction through independent debtor confirmation on a channel the borrower does not control, plus pattern checks across debtors.
How is pre-billing different from a normal invoice, and why is it hard to catch?
In pre-billing the debtor and the relationship are real, but the invoice is presented before the goods ship or the work is done. Standard confirmation can come back positive because the debtor recognizes the borrower, so the fraud hides in the timing. You catch it by confirming proof of delivery or completion, not just the obligation, and by monitoring dilution and post-funding credit memos.
Does fraud detection slow down funding?
It does when it is fully manual, which is why teams under volume pressure tend to cut corners. Automating the deterministic checks (duplicate matching, debtor lookups, contact-channel verification) lets reviewers focus only on genuine exceptions, so clean invoices clear faster while the risky ones get more scrutiny.