By subscribing you agree to our Privacy Policy and provide consent to receive updates from our company. You can unsubscribe at any time.
© 2026 Polygon Research. All right reserved.
Scan 03 · Underwriting
An underwriting scan compares approval and denial outcomes across borrower groups for the same loan purpose, product and market, then puts the denial reasons beside the result. HMDAVision reports it as the Minority Delta, with odds of denial and disparity odds one click away, for your institution, your peers and the market.
Free tier is three answers a day, fifteen a month, no card.
Start here
Examiners compare denial rates for protected-class applicants with those of a control group for comparable applications, then review the stated reasons and the files behind the difference.
Interagency Fair Lending Examination Procedures, FFIEC
The public record cannot see inside a credit file. It can see every outcome, every stated denial reason and every borrower group for every lender in a market. That is enough to find the delta, name the loan purpose it lives in, and see whether peers show the same one.
01
A delta that survives segmentation
A gap present in every loan purpose, income group and channel points at policy rather than mix.
02
A delta in one purpose only
Purchase level, refinance wide. That is a product or a policy question, and a narrower one to answer.
03
Denial reasons that differ by group
Credit history leading for one group and DTI for another, at similar income, asks why the file was read differently.
04
An odds ratio far from 1.00
Disparity odds compare each group's odds of approval to the control group. A ratio of 0.5 is half the likelihood.
05
Peers approving who you deny
A delta that lenders of similar size in the same market do not show is yours to explain.
06
A delta that widens by release
Distance that grows across HMDA years rather than holding or closing.
The scan itself
The disparity check puts the Non-Hispanic White denial rate beside the minority denial rate for any dimension you pick, with more than sixty available, and reports the difference in points. Three views sit beside it: denial reasons ranked against last year, odds of denial, and disparity odds against a control group.
One lender, one metro, 2025, lender identity removed. Odds of denial for this file: 28.1 percent, or one in 3.6. Every view runs the same way on any peer or on the market, with the same filters.
What the scan says
Purchase denials are level across groups. The delta lives in refinance and home improvement, where it runs 25 to 29 points, and the top two denial reasons are credit history and DTI. That combination points at specific levers rather than a general problem.
Small cells move fast. A purpose with thirty denials can swing ten points on three files. Read the delta with the count beside it.
The same table on peers of similar volume. A refinance delta the whole market shares is a demand and credit-profile question. One only you show is a policy question.
The workflow
Each step is a set of filters in HMDAVision. Keep them and the scan reruns on the next release in minutes.
1
Define the market and loan population
Geography, year, action type, and whether purchased loans are in or out. Net applications is the denominator that matters.
Output: a defined population with a documented rationale.
2
Pick the control group and the dimension
Minority status against Non-Hispanic White is the default. Any of sixty dimensions can be the row: loan purpose, income group, channel, tract type, age.
Output: the delta by the dimension that matters to you.
3
Find where the delta concentrates
Segment until the gap is narrow enough to name a purpose, a channel and an income band.
Output: a specific place to look inside the credit file.
4
Open the denial reasons
Ranked, with last year's rank beside them. A reason that climbed the list in one segment is the first file to pull.
Output: the reason behind the delta.
5
Run the odds views
Odds of denial for the book, then disparity odds by group against the control group. These are the figures a regulator's statistician will compute.
Output: the ratio, before someone else reports it.
6
Repeat on peers, record, repeat next year
The same filters on the peer set separate your delta from the market's. Keep the selections so the trend can be followed.
Output: a peer comparison and an audit trail.
From finding to action plan
Because the scan runs on your institution, your peers and your market in the same session, and is segmented by purpose, channel and income, the reading usually points at one of three moves.
When DTI and insufficient cash lead the purchase denials for a group that peers are approving. The redlining view shows which tracts the assisted loans would land in.
When credit history drives a delta in a purpose where the market delta is small. Overlays, credit score floors and manual underwriting rules are the first place to look.
When the gap sits in a purpose you barely offer. Refinance and home improvement here. The lenders closing those loans nearby show what the product looks like and who is buying it.
Rerun the same filters. A delta that narrows in the segment you acted on is evidence the lever worked.
Start scanning
Ask one question at mortgagedata.ai, or open the disparity check on your own institution with a guided seven-day trial of HMDAVision.
Polygon Research is a research and analytics firm, not a law firm. Nothing here is legal advice, and a fair lending program should be built with counsel.
Keep going
Scan 01
Redlining
Your footprint against tract demographics, and who lends in the tracts you skip.
Scan 02
Marketing
Application mix against the population of the market you serve.
Scan 04
Pricing
Higher-priced lending, rate spread and fees against the market.
Scan 05
Steering
Product, term and channel distribution across borrower groups.