The most expensive file in your pipeline is the one that looks fine. A doctored bank statement, an inactive vendor or a borrower who is not who they claim to be can cost you the loan, the asset and weeks of recovery work. The files behind those deals can look clean: the statement balances, the business has a perfect online rating, and the identity matches on name, address and date of birth.
This article walks through six patterns from recent application reviews, anonymized, where a file passed a surface check and one cross-check said otherwise. For each, you will see what caught it and what to add to your own intake process, so you can stop these deals before funds go out without slowing down good borrowers.
Generative tools have made it cheap to edit documents and build a believable business footprint, so the question is “do independent sources agree with each other?” Equipment finance adds exposure most lending does not: you often fund a vendor’s invoice for a physical asset, so the borrower, the seller and the equipment must all hold up.
1. Documents can be edited and still add up
A recent business bank statement showed monthly deposits of $63,250. An earlier version embedded in the same PDF showed $6,300, ten times less. In another file, someone had lowered the withdrawals and raised the ending balance by the same amount, about $7,200, so the statement still added up. A third statement carried a different account holder name, placed in the field through overlapping text layers.
Arithmetic checks catch sloppy edits, not careful ones. Look at provenance as well as content: version history, hidden or overlapping text, fonts and metadata. Where possible, verify reported inflows against an independent source such as a direct bank connection or records issued by the bank. This matters most on broker and dealer submitted packages, where a file may pass through several hands before it reaches your credit team.
2. Identity can match and still be wrong
A submitted driver license matched on name, date of birth, address and license number, yet its issue and expiration dates failed validation when checked against source. In a separate review, an applicant’s name, phone number, date of birth and SSN failed to match in several different combinations.
A partial match is neither a pass nor a fail. Treat each disputed field as a specific question, and verify those anomalies before asking the applicant for more paperwork. That keeps friction low for legitimate borrowers and focuses effort where the risk sits.
3. A good reputation is not an identity link
One construction contractor showed a 5.0 Google rating from eight reviews, an A+ listing with the Better Business Bureau, a website with a matching name and phone number, and an active LLC. The same report surfaced adverse media: the sole principal reportedly pleaded guilty in 2020 to wire fraud and conspiracy tied to a home repair contracting scheme of roughly $319,000.
Both findings were true at once. Reputation signals are cheap to build, and eight reviews is a thin sample. Positive indicators should not cancel a negative finding, and a negative finding should not be acted on until the identity link is confirmed. Verify first, then decide.
4. The name matches, the entity does not
A recent secretary of state search came back as a strong match, including a perfect match on business name. But the owner could not be confirmed, the city did not match, and the company had been administratively dissolved for tax reasons. A strong name match had hidden an inactive company.
The same issue showed up on the vendor side, where the stakes are higher in equipment finance. On a $50,000 equipment financing application, the seller’s registry status came back inactive. Funds go to the vendor, so check entity status as its own step for borrowers and vendors alike, and confirm the legal entity and the intended payment recipient before funding. A registry exception calls for verification. It does not establish fraud.
5. The equipment should fit the business
A landscaping company submitted an invoice for de-icing equipment. The same package also contained an application for a spinal decompression machine, which belonged to a different business. Fraud was not established, but the documents did not tell one coherent story. The equipment is also your collateral, so match the borrower, the equipment and the intended business use, and resolve the document association before treating the use case as settled.
6. Financials tell their own story
In one tax analysis, the numbers did not reconcile. Retained earnings were off by more than $200,000, and the preparer’s own reconciliation showed a $630,000 gap between the books and the tax return. Receivables rose while revenue fell. Shareholder loans and an unexplained intercompany receivable both climbed sharply, with no affiliated entity disclosed.
No single line proves anything. Together they justify a focused request: the schedules that explain the difference, not a blanket request for everything again.
What to take away
- Check provenance, not just content. Version history and hidden text reveal what a clean-looking PDF does not.
- Turn partial matches into specific questions. Verify the disputed detail before burdening the applicant.
- Verify status separately from names, and confirm the vendor and payment recipient before funds move.
- Test consistency across borrower, equipment, vendor and business use.
- Keep a reviewable record: the source evidence, the checks performed and the reviewer’s rationale.
That last point matters most as automated screening becomes common. Software can flag risk quickly, but people should review it, and the record should show why a decision was made. Most flagged items turn out to have an explanation. The goal is not more declines. It is a sharper follow-up question, asked earlier, so good deals move faster and bad ones stop before funding.
Every file above would have passed a quick glance, so the question is what your team checks before the wire goes out.
Shivi Sharma is Co-founder and President of Kaaj, which helps equipment finance lenders turn borrower packages into decision-ready credit analysis. A credit and fraud risk operator, she has worked across American Express, Uber and Varo Bank, where she saw how much high-stakes credit judgment happens outside the system of record. At Kaaj, she focuses on making that judgment structured, visible and reusable for credit teams. Learn more at kaaj.ai.
