Algorithm Denied Your Loan? You Have the Right to Know Why

What actually happens when an algorithm reads your loan file?

You submit financials, the lender's model scores you, and a decision comes back in hours or days. If it's a denial or a smaller offer than you asked for, most owners assume the bank just doesn't want their business. That assumption costs money. Under the Equal Credit Opportunity Act (ECOA) and the Fair Credit Reporting Act (FCRA), you are entitled to a written adverse action notice that states the specific reasons for the decision. Not a form letter. Specific reasons.

Algorithmic lending has grown fast. The Federal Reserve's 2024 Small Business Credit Survey found that 44% of small employer firms applied for financing in 2023, and approval rates at online lenders, which use automated underwriting almost exclusively, lagged large banks on full-approval rates. The model made the call. The law still applies.

What is an adverse action notice and what must it include?

An adverse action notice is the document a lender must send you within 30 days of a denial, a counteroffer for less than you requested, or a less favorable term change. The ECOA requires it. The FCRA adds requirements if a consumer or business credit report was used in the decision.

The notice must include:

• The specific reasons for the adverse action (up to four principal reasons)
• The name and address of the lender
• A statement of your rights under ECOA
• If a credit report was used: the name and contact info for the reporting agency and a notice of your right to a free report

The phrase "did not meet our credit standards" is not a specific reason. "Debt-to-income ratio too high" is. "Insufficient time in business" is. If the notice you received reads like a form rejection, you can push back.

How do you actually request more detail?

Start with a written request to the lender's compliance or underwriting department. Reference the adverse action notice by date and cite ECOA (15 U.S.C. § 1691) and, if applicable, FCRA. Ask for the specific model factors that drove the decision.

Lenders using automated systems are supposed to translate model outputs into human-readable reason codes. Those reason codes are what you want. Common ones include:

| Reason Code | What It Usually Means | |---|---| | Derogatory public records | Judgments, liens, or bankruptcies in your record | | Ratio of balances to credit limits too high | High revolving utilization on business or personal credit | | Too few accounts currently paid as agreed | Limited positive trade lines | | Length of time accounts established | Short credit history for the business or owner | | Insufficient cash flow | Revenue doesn't service projected debt at underwriting standards |

Once you have the actual codes, you are no longer guessing. You know what to fix before the next application.

Does this right apply to business loans or just personal credit?

ECOA covers credit extended to businesses, not just consumers. However, there is a threshold: lenders are not required to provide a written adverse action notice to businesses with gross revenues over $1 million in the prior fiscal year, unless the application was for trade credit, factoring, or similar. For most SMBs reading this, the protection applies.

One practical note: the FCRA's adverse action rules focus on consumer credit reports, so if the lender used your personal credit as part of a small business decision (common for businesses under three years old or with thin business credit files), both statutes apply.

"Most applicants never ask for more detail, which means they reapply blind and get denied again."

What can you do with the reason codes once you have them?

This is where the request pays off. The reason codes give you a repair list.

If the issue is fixable in 60 to 90 days: Address it before reapplying. High utilization, for example, can often be reduced. A single late payment in dispute can sometimes be corrected or explained with documentation.

If the issue is structural: Consider a different loan product or lender. A cash flow problem that disqualifies you for a term loan may not disqualify you for a revenue-based product or an SBA microloan. The SBA's 7(a) program has different underwriting criteria than most bank models.

If the reason code seems wrong: It happens. Models pull from data aggregators that sometimes carry errors. Request your business credit reports from Dun & Bradstreet, Experian Business, and Equifax Business and check them against what the lender cited. Dispute inaccuracies directly with the bureau.

If you believe the decision was discriminatory: File a complaint with the Consumer Financial Protection Bureau or the Office of the Comptroller of the Currency. Algorithmic systems are still subject to disparate impact analysis under fair lending law.

How is AI changing underwriting, and does that shift the legal ground?

Lenders are moving fast on this. Machine learning models now incorporate variables well beyond FICO and revenue: account transaction patterns, industry classification, even web presence signals in some platforms. The reason codes generated by these models are sometimes approximations of what the model actually weighted, which regulators have flagged as an explainability problem.

The CFPB issued guidance in 2022 specifically addressing this: lenders cannot use a "complex algorithm" as a shield against providing specific, accurate reasons. The law does not have a "model too complicated to explain" exception. If a lender cannot tell you why you were denied in plain terms, that is a compliance problem on their end.

For SMB operators, this matters because it means the adverse action notice you receive from a fintech or an automated bank portal should be treated as a starting point for a conversation, not the final word.

What we'd actually do

• Request the adverse action notice in writing within 60 days of the decision. Cite ECOA and ask specifically for the principal model reason codes, not just the boilerplate. Keep a copy of every communication.
• Pull your business credit reports from all three bureaus before your next application. Dispute any inaccuracies. A clean file going in removes one category of denial before the model ever scores you.
• If the reason codes point to fixable factors, wait and repair rather than reapply immediately. Multiple hard inquiries in a short window can themselves become a reason code against you. One targeted application after addressing the specific issues beats three fast rejections.