Fair lending exception monitoring that reads every reason and compares every exception with similar loans
Every exception reasoned at the time, not explained a quarter later.
See one case, screen by screen ↓It is the ongoing review of discretionary decisions — rate concessions, fee waivers and underwriting overrides — to check that similar borrowers were treated alike and that each exception has a legitimate, documented reason. Fair Lending Exception Monitoring does it as exceptions are granted, not in a quarterly sample.
How to choose AI solutions for lending →
Updated
Rate concessions, fee waivers and underwriter overrides are where two similar borrowers can be treated differently. Most lenders review them in a quarterly spreadsheet, long after closing: the analyst pulls the exception report, builds a matched-pair sample of a few dozen files by hand, and asks loan officers why they priced a loan the way they did three months ago.
By then the reason field is often blank and the officer no longer remembers. A reason written months later is not documentation made at the time — and without it, a lender cannot show a state examiner that each exception was fair. CFPB examiners found this pattern in competitor-match exceptions: granted to some borrowers on request, without evidence, not offered to others — and the lenders could not show a legitimate reason.
Estimated split for one quarter’s review of discretionary pricing exceptions, by hand and with the solution.
Six specialist agents read every exception's reason and compare it with similar loans; the Fair Lending Officer decides.
Halden Ridge Bank closed 3,186 mortgage loans in Q3 2026 and granted 549 exceptions on them. Here is what Dana Whitlock, its Fair Lending Officer, sees on Monday 12 October, screen by screen, in the working solution.
By 02:10 the agents have collected 549 exceptions — 412 rate concessions, 96 fee waivers and 41 overrides — read every reason against its evidence, matched each loan to a median of 18 peers and tested 22 branches, 25 officers and 5 products. Each dot is one concession: across, what similar loans received; up, what this loan received. Below the diagonal, the borrower got less than its true peers.
“Read 549 reasons against their evidence — 446 supported · 42 written but not supported · 61 missing.”
Three clusters were found overnight; one waits for her decision. Dana picks Branch 14 and the lasso closes around 11 loans at Linden Avenue. The side panel reads the selection in one sentence: Black borrowers (estimated) received 12 bps on average; similar loans received 50 bps. 8 of the 11 were granted by Kyle Brandt.
“Concessions 38 bps smaller for one group, 9 of 11 without a documented reason.”
The investigation lists all 11 loans with score, LTV, the concession each got, its peers and the reason as written — “No reason entered” on nine. It compares them with 14 white borrowers’ loans at the same branch, shows who granted them, and replays how the agents got there: the gap is 38 bps, p = 0.002, and it holds after score, LTV, amount and product.
Of the 14 comparison loans, 10 say “competitor match” — only 4 have the competing offer on file.
Loan 26-14-0871, A. Whitaker: 30-yr fixed, $328,000, score 742, LTV 80%, DTI 34%, locked Aug 14 — 10 bps, no reason. Its closest match, 26-14-0833, G. Holm: $315,000, score 738, LTV 80%, DTI 36%, locked Aug 11 — 50 bps, “Competitor match — borrower has offer”, with no competing offer found. Same branch, same officer.
“No pricing factor explains the 40 bps difference.” The stronger file got the smaller concession: $1,312 on $328,000.
The officers view shows the average concession gap to matched peers for every loan officer, by borrower group. Kyle Brandt — 47 exceptions, 9 without a reason — sits at −38 bps on 8 loans to Black borrowers, and −2 on 33 loans to white borrowers. Only the compliance team sees this view.
The feed shows every concession, waiver and override as it is granted, with its reason read. A reason is due within 3 business days. A competitor match today at Branch 14 came in with the Loan Estimate on file — “Reason in · 6 min”. A 40 bps relationship-pricing concession came in with no deposits found, so the reader asked for them.
The reader never writes or suggests a reason for an officer — it classes what the officer wrote.
Three actions, each with its working: credit the 11 borrowers 38 bps of the loan amount by principal reduction, $13,060 in all; for 90 days, any concession by Kyle Brandt above 10 bps needs Marcus Lindqvist, the pricing desk manager, before lock; and amend Policy 4.2 so no competitor-match exception is granted without the competing Loan Estimate, bank-wide from 1 Nov 2026.
“Nothing is paid or changed until both sign.”
She reads the three actions and the per-loan credits, then signs. The confirmation says exactly what she is approving and what still has to happen before any money moves.
“You are approving 3 actions for Branch 14, including $13,060 in credits to 11 borrowers. Harold Bisset co-signs before anything is paid or changed.”
Harold Bisset co-signs and the remediation is marked signed by both. Credits are scheduled and borrower letters queued; each signature lands on the audit trail beside the agents’ own steps, from the quarter’s close to the flag.
“Harold Bisset co-signed · credits scheduled, letters queued.”
The Q3 2026 fair lending pricing report is drafted from the same data as the scatter — scope and method, results by exception type, findings, actions and timeliness of reasons — and every number links to its loans. Dana approves it, exports the PDF or sends it to the board pack.
The same quarter, seen by the five people who grant, review and sign off on exceptions — what it looked like before, and what it looks like now.
Six specialist agents read every exception's reason and compare it with similar loans; the Fair Lending Officer decides.
Pulls every rate concession, fee waiver and underwriting override from the origination system and pricing engine the day it is granted, with the loan, its size against the rate sheet of the lock day, the officer and the approver.
Reads the reason the officer wrote and looks for the evidence the pricing policy requires — a competing Loan Estimate for a competitor match, an employer letter for relocation, verified deposits for relationship pricing — and classes it supported, written but not supported, or missing.
Matches each exception to loans with the same product, credit score within 20 points, LTV within 5%, the same market and a rate lock within 30 days on the same sheet, and reports the peer median and the closest single match.
Uses HMDA-reported race, ethnicity and sex where the borrower gave it; otherwise estimates from surname and census tract and keeps the probability.
Groups exceptions by branch, officer and product; where a group of at least 8 loans differs from its peers by more than the tolerance, tests significance with score, LTV, amount and product as controls, and looks for documented reasons that explain it.
Drafts the actions for each flagged cluster — the credit to each borrower and how it is computed, limits on discretion, policy changes — and the quarterly fair lending pricing report from the same data.
Decides; the Chief Compliance Officer co-signs. The agents propose; a named person decides.
Ask in plain words about any exception, branch, officer or cluster — or ask for the fix to be drafted. Every answer points to its loans and policy.
Why is Branch 14 flagged?
Across Q3, 11 rate concessions to Black borrowers (estimated) at Branch 14 averaged 12 bps; similar loans at the branch averaged 50 bps — a 38 bps gap, p = 0.002, after score, LTV, amount and product. 9 of the 11 have no written reason, and 8 were granted by Kyle Brandt. Of the 14 comparison loans, 10 say “competitor match” but only 4 have the competing offer on file.
Which exceptions have no written reason?
61 of the 549 Q3 exceptions: 47 rate concessions, 11 fee waivers and 3 overrides. 9 of them are the Branch 14 cluster. Policy 4.2 asks for a reason within 3 business days.
Compare loan 26-14-0871 with its closest match
26-14-0871 (A. Whitaker): 30-yr fixed, $328,000, score 742, LTV 80%, DTI 34%, locked 14 Aug — 10 bps, no reason. Its closest match 26-14-0833 (G. Holm): $315,000, score 738, LTV 80%, DTI 36%, locked 11 Aug — 50 bps, “competitor match — borrower has offer”, with no offer on file. Same officer, same rate sheet. The stronger file got 40 bps less: $1,312.
Draft the remediation for Branch 14
Drafted, for your decision: credit the 11 borrowers 38 bps of their loan amount — $13,060 in all; for 90 days, Kyle Brandt’s concessions above 10 bps need Marcus Lindqvist; amend Policy 4.2 so a competitor match needs the competing Loan Estimate. Nothing is paid until you sign and Harold Bisset co-signs.
13 screens from the working solution, on its sample data. Pick one to see it large.
How peers are matched, what counts as a gap and who decides are settings. Each change is saved as a new rule version.
| Setting | Default | Choose from |
|---|---|---|
| Credit score band for peers | ± 20 points | ± 10, ± 20 or ± 30 points |
| Loan-to-value band for peers | ± 5% | ± 3%, ± 5% or ± 10% |
| Rate lock window | 30 days, same rate sheet | 15, 30 or 45 days |
| Tolerance band | 15 bps | 5 to 40 bps, with the effect shown as you change it |
| Smallest group tested as a cluster | 8 loans | 5, 8 or 12 loans |
| Significance | p < 0.05 | p < 0.01, p < 0.05 or p < 0.10 |
| Reason due | 3 business days | 1, 3 or 5 business days; day-2 reminder to the officer on or off |
| Evidence a reason needs | Competing Loan Estimate · employer letter | Each rule on or off |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample bank data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (DOJ pricing settlement with a Tennessee bank: 13–19 bps, $1.3M relief and a $50K penalty; CFPB Supervisory Highlights on competitor-match pricing exceptions, 2021 and 2023). People, the bank, its branches and loans named on this page are fictional — characters and sample data in the working solution.
It is the ongoing review of discretionary decisions — rate concessions, fee waivers and underwriting overrides — to check that similar borrowers were treated alike and that each exception has a legitimate, documented reason. Fair Lending Exception Monitoring does it as exceptions are granted, not in a quarterly sample.
The peer matcher finds loans with the same product, credit score within 20 points, LTV within 5%, the same market and a rate lock within 30 days on the same rate sheet, and reports the peer median and the closest single match. The borrower group is never used to choose peers.
It checks the reason against the evidence your policy requires. A competitor match needs the competing Loan Estimate on file; relocation needs the employer letter. Each reason is classed as supported, written but not supported, or missing, and the officer is reminded on day 2.
HMDA-reported race, ethnicity and sex are used where the borrower gave them; otherwise the group is estimated from surname and census tract. Estimates are visible only to the compliance team — never to loan officers or the pricing desk — and never used in a pricing decision.
No. The disparity finder reports a gap, its significance and whether documented reasons explain it, and calls an unexplained gap “unexplained”. The Fair Lending Officer reviews the loans and decides what to do.
No. No agent changes a price or a loan, and none sends money or a letter. The report writer drafts the remediation; the Fair Lending Officer signs and the Chief Compliance Officer co-signs before anything is paid or changed.
Yes. The score and LTV bands, market, rate lock window, tolerance band, smallest cluster, significance level, reason clock, evidence rules and approvers are all settings, and each change is saved as a new rule version.
The Agentic Solution Engine builds and deploys it from your requirements — your pricing and override policy, rate sheets and a quarter of past exceptions — and it goes live once every quality gate has passed. We will walk you through it on your own exceptions first.
One question, all fifty states at once, every answer cited to the rule it comes from.
1-min demoRegulatory Change ManagementEvery new rule scoped to products and states, and owned the day it lands.
1-min demoHMDA and 1071 Data ValidationEvery record checked against the file, not a sample, before filing.
1-min demoWe’ll run Fair Lending Exception Monitoring on a quarter of your own pricing exceptions.