Mortgage loan scenario search, explained.

Testing one household against every mortgage program a lender offers

The Whitfields' scenario, first call to Loan Estimate in five steps

1. The borrower callsThe Whitfields describe their plan to a loan officer's assistant: price, down payment, two incomes, a veteran, a condo they keep.
2. Someone works out which programs fitEvery program at more than one down payment, with income counted each program's way. This is where the VA option and the condo-rent trap are found, or missed.
3. The loan officer presents options, then takes an applicationSix details make it an application under Regulation Z, and a Loan Estimate is due within three business days.
4. Underwriting tests the fileDocuments, a credit report and an appraisal test every assumption the scenario made.
5. Closing, then quality controlA sample of closed loans is re-underwritten. A program chosen on a guess comes back here as a defect.

Step 2 is the subject of this guide. The people, lender and figures in the example are fictional.

Four surprises in matching borrowers to mortgage programs

Four scenario terms: program shelf, overlay, near-miss, entitlement

Fourteen terms are defined in the one-page summary.

The dozen facts that decide which programs a borrower fits

In shortA scenario needs about a dozen facts, and several of them can flip a whole program: occupancy, any VA entitlement in use, what happens to the home being left, how long the self-employment has run, and the county. Collect six particular details and the scenario becomes an application with a legal clock, so a scenario worksheet stops short of that on purpose.

A scenario is a borrower's situation, described before there is a file. Every lender and broker keeps its own worksheet. The layouts differ, but the fields are much the same, because each one feeds a rule. The table below lists the facts a search needs and the rule each one feeds, with the Whitfields' answers.

Which facts can flip a program from yes to no

Where a scenario becomes an application

Regulation Z says that for most mortgages "an application consists of the submission of the consumer's name, the consumer's income, the consumer's social security number to obtain a credit report, the property address, an estimate of the value of the property, and the mortgage loan amount sought."11 Once a lender has all six, the Loan Estimate is due within three business days, and the official commentary closes the obvious workaround: collecting more information does not delay the clock.11,12

A worksheet built for scenarios therefore stops at the county rather than the street address, and works from the scores the borrower reports rather than a credit pull that needs a Social Security number.

Agency, government, portfolio and non-QM: who writes the rules for each kind of loan

In shortEvery program on a shelf answers to someone: Fannie Mae or Freddie Mac for conventional loans, HUD for FHA, the VA and USDA for their own loans, an investor's matrix for jumbo and non-QM, and the lender itself for portfolio loans. Conventional loans were 78% of the 2025 originations lenders reported under the Home Mortgage Disclosure Act (HMDA), FHA 13% and VA 8%. A search has to know each program's owner, because the owner decides which rule wins.

Denials are not spread evenly across programs

HMDA does not record which program a denied borrower would have fitted. The figure that matters to a scenario search, how many borrowers were offered a program they did not fit when another would have worked, is not published anywhere. Section 10 says how to measure it on your own files.

The rule changes since November 2025 that move a scenario's answer

In shortNo single law governs a scenario search. The answer comes from each program's guide, which changes several times a year, and from Regulation Z's rules on applications, estimates and steering. Between November 2025 and October 2026 the conventional score floor went for automated loans, the departing-residence rule was rewritten, every loan limit rose, and a Colorado law on automated decisions was signed.

The Fair Housing Act was not changed by the 2026 Regulation B rule, and the prudential regulators still examine for fair lending.25 Treat "which program was offered to whom" as a fair lending question in both directions: steering a borrower to a costlier program, and failing to mention one they fit.

The 2026 loan limits by county, and the dollar that turns a loan jumbo

In shortThe conforming limit for a one-unit home is $832,750 in most counties in 2026 and up to $1,249,125 in high-cost ones. Lakewood's county, Jefferson, sits at $862,500. The limit decides whether a loan is conforming or jumbo, sets FHA's range and enters VA's formula, so a search looks it up by county every time.

One house, two kinds of loan

The limit applies to the original loan amount, so the down payment is the lever that moves a loan across it. Take a $950,000 house in Jefferson County. At 5% down the loan is $902,500, which is $40,000 over the county's $862,500 limit: a jumbo loan, priced and underwritten by an investor's matrix. At 9.25% down, $87,875, the loan is $862,125, which is conforming. Because it is above the $832,750 baseline it is a high-balance conforming loan, which Fannie Mae and Freddie Mac can buy.15,8

Testing one down payment hides that choice. A borrower who says "5% down" may have the cash for 9.25% and not know it matters. Whether to bring it is the borrower's decision, and the loan officer can only offer it after seeing both sides of the line.

How each program counts the same household's income, and the home they are leaving

In shortQualifying income is a calculation, and each program does it its own way. Self-employment is averaged over two years when it is stable or rising. Bank-statement loans count deposits less an expense factor. Rent from a home the buyer is leaving counts three different ways at Fannie Mae, FHA and VA, and that one choice moves the Whitfields' ratio by about nine points.

The Whitfields' income, as a full-documentation program counts it

Fannie Mae generally wants "a two-year history of the borrower's prior earnings" for self-employment, and the lender must track year-to-year changes in gross income, expenses and taxable income to judge whether the business can keep paying.10 Marcus's income rose, so the two-year average is the conservative figure. Had it fallen, a lender would usually use the lower recent year or ask why. Less than two years of self-employment can still qualify when the most recent returns show a full 12 months from the current business and earlier income in the same line of work.10

A bank-statement non-QM loan counts something else: eligible deposits over 12 or 24 months less an expense factor set by the investor's matrix. On the sample lender's matrix, 50% of Marcus's $296,400 of deposits gives $12,350 a month, about $1,950 more than his tax returns show. More income does not make it the better loan; Section 9 shows why.

Three rulebooks for one condo

The Whitfields will rent out their Aurora condo for $2,050 a month. Its payment of principal, interest, taxes, insurance and association dues (PITIA) is $1,640. Each program answers "does the rent count?" differently.

At 52.6%, FHA may still be approved by its automated system, but the loan officer is now presenting a stretched loan, and VA at 49.3% passes only because the household's residual income stays more than 20% above VA's table.6 The rent rule did all of that, before any document was read.

VA partial entitlement worked through, and why the county changes the down payment

In shortA veteran with full entitlement has no VA loan limit. A veteran whose benefit is partly tied up in an earlier loan starts from a quarter of the county's conforming limit, subtracts what is in use, and lenders usually ask for 25% of the price above four times what is left. In Jefferson County that is $78,625 for the Whitfields. Then come the funding fee, the 41% ratio and residual income.

VA's circular for 2026 states the first step: "the amount of remaining entitlement is calculated by taking 25% of the CLL and subtracting the previously used entitlement that was not restored." The next two are lender practice, which the circular also describes: VA loans "are often securitized in pools that require the loan to have a 25% guaranty", so for veterans without full entitlement "lenders often limit the total loan amount to four times the amount of the remaining entitlement."7

The one-unit limit applies even when the veteran buys two to four units.7 With full entitlement there is no loan limit at all, as long as the veteran qualifies and the appraisal supports the price.31

A calculator with one default limit looks right in most counties, because 3,075 of 3,235 sit at the baseline.8 It is wrong in every higher-cost county, including the four around Denver, Jefferson, Denver, Arapahoe and Adams, which all sit at $862,500.8 The four-times line and the 25% down payment are lender and investor practice, which the circular describes but does not set, so a scenario should say which convention it used.7

The funding fee, and the bands that change it

A veteran with partial entitlement is a subsequent user by definition, so the 3.3% band is the one in play at small down payments. At 10% down the Whitfields pay 1.25%: $9,731.25 on a $778,500 base loan, usually financed, for a total loan of $788,231.25.19

After entitlement: the 41% ratio and residual income

Entitlement says how much VA will guarantee. It says nothing about whether the household can carry the payment. VA uses two standards: a debt-to-income ratio, with 41% as the benchmark, and residual income, the money left each month after the housing payment, debts and taxes, compared with a table by family size and region. For loans of $80,000 and up, a family of four in the West, which includes Colorado, needs $1,117 a month. Above 41%, a supervisor's written justification is needed unless residual income "exceeds the guidelines by at least 20 percent."6

Credit score floors, ratio limits and the overlays lenders add on top

In shortAgency score floors run from none to 640: none for Desktop Underwriter loans, 620 or 640 for manual conventional loans, 500 and 580 tiers for FHA, none set by VA. Lenders then add overlays. A good scenario answer says which line decided the result, the agency's or the lender's, because only one of them can be changed by asking a different lender.

A score still prices the loan even where it no longer decides eligibility. Fannie Mae's credit score topic keeps its section on loan-level price adjustments by score, and rate sheets step by score band.2 Two borrowers at 690 and 760 can both be eligible and pay quite different rates, which is why a small score change can matter even when it opens no new program.

Near-misses: the honest change that turns a decline into an option

In shortWhen a borrower fails a program by a single rule, naming that rule, and the smallest honest change that would pass it with its cost and time, turns a no into a plan the borrower can choose. The changes must be real ones the borrower controls. Anything that changes how the borrower is described, such as occupancy or income, is misrepresentation.

A pricing engine reports a failed program as ineligible, whether it failed by one point of score or by $200,000 of loan amount. A scenario search keeps the distance. The Whitfields fail three programs by one rule each:

Reserves are what is left after closing in checking, savings, investments and the vested part of retirement accounts. Fannie Mae counts the vested amount of a retirement account; the sample lender's bank-statement matrix counts 60% of it, an investor rule the search has to load with the matrix.32

What a near-miss may never suggest

Near-misses and the anti-steering safe harbor

Regulation Z forbids a loan originator to steer a consumer to a loan "based on the fact that the originator will receive greater compensation" unless the loan "is in the consumer's interest."28

Its safe harbor protects an originator who, for each type of loan the consumer is interested in, presents the loan with the lowest interest rate, the lowest rate without risky features such as a prepayment penalty, interest-only payments or a balloon in the first seven years, and the lowest total points and origination fees. The originator must have "a good faith belief" that the consumer "likely qualifies" for each, and must obtain options from a significant number of creditors, defined as "three or more."28

Read closely, that is a scenario search: test the borrower against the shelf, keep the loans they likely qualify for, and pick three by rule. A search that sorts programs into fits, near-misses and non-fits, with the failed rule stated, is also the evidence for leaving a lower-rate loan off the list. The rule is aimed mainly at brokers, who choose between creditors; a retail loan officer who is paid within the pay rules satisfies the steering ban for loans with the employer.28

Where scenario answers go wrong, and where the mistakes surface months later

In shortNobody audits a scenario answer when it is given. Quality control samples closed loans, so a wrong "yes" surfaces as a defect after closing and a wrong "no" never surfaces at all. In the quality control reviews run on ACES software, income and employment caused 20.07% of critical defects in early 2026 and borrower eligibility 7.43%; eligibility defects grew 291.58% in 2025.

The common errors, and why they happen

What happens to the applications that start

Not every denial or withdrawal is a scenario gone wrong; some borrowers change their minds. But each one was a file someone opened, priced and worked, and the MBA names "reduced application pull-through" as one reason production cost rose to $11,094 a loan in 2025.35

Why the first program quoted weighs more at 7% rates, when most buyers ask one lender

In shortThirty-year rates climbed from about 6.1% early in 2026 to 7.40% in October. The MBA expects $2.1 trillion of single-family originations in 2026 and $1.99 trillion in 2027, with refinances down 17% in 2027. At these rates more purchases sit near a ratio or cash limit, and the program that fits can decide whether the purchase happens.

That makes the first loan officer the borrower's main adviser on what to compare, and the loan officer's search the borrower's only search. Federal Reserve researchers found that "time-invariant lender attributes explain little" of the gap between what similar borrowers pay, and that the least financially savvy borrowers tend to overpay.38 Shopping across programs is the same idea one step earlier.

On the Whitfields' sample rate sheet the gap between the VA loan and the conventional one is $492 a month, which is mostly the conventional loan's mortgage insurance and a lower VA rate, and is larger than the 54-basis-point dispersion is worth on the same loan. That sheet is part of the fictional example: its 5.875% to 6.375% sit well below the 7.40% Freddie Mac 30-year average of that week.36

Speed matters too. A 2026 Federal Reserve study of HMDA data puts the average time from application to decision at 50.8 days, measured on lender, state and month groups rather than loan by loan (the median group took 43.5 days).39 A scenario answer that takes until tomorrow adds to that from the start, at the moment a shopping borrower is most likely to call someone else.

Assistant, loan officer, scenario desk, account executive: how long a borrower's scenario question waits for an answer today

In shortAn assistant takes the call, the loan officer prices what they expect will fit, and anything unusual goes to the scenario desk or a wholesale account executive by email or chat. Most of the elapsed time is waiting for that answer. Nobody publishes how long it takes, so the hours below are an estimate to replace with your own.

Who answers a borrower scenario: loan officer, assistant, desk underwriter, secondary

Three people before a borrower hears a scenario answer

Working hours on one non-standard purchase scenario

Memory, pricing engines, scenario desks and more: five ways to match programs

We found no public study of scenario-desk turnaround times or of how often the first program quoted is the best fit, as of October 2026. Your own desk's timestamps and a replay of last quarter's scenarios are the honest baseline.

The Whitfields' $865,000 Lakewood purchase: nine programs at three down payments, VA first, the note saved by 9:06 am

In shortThis is the example from the top of the page, inside the product. Jordan Pike, the assistant, types the call in plain words. Six agents read it, ask the two facts it left out, and test nine programs at three down payments in 38 seconds. Priya Sundaresh, the loan officer, decides what to present; Tom Reyes on the scenario desk confirms the one judgment call; the cited note is saved to the lead at 9:06 am.

1

The call, as Jordan typed it

At 8:12 am on Monday, October 12, Jordan types what the Whitfields said on the phone into the lead for Priya: the $865,000 house in Lakewood, 10% down, Marcus's two years of Schedule C, Elena's salary, "Navy veteran", the scores, the debts and the condo they are keeping. No form and no product codes.

The scenario reader underlines the eleven facts it read, in place, so Priya can see what it took from the words.

2

Two questions the words left out

Two facts the note did not settle would each change several programs. The scenario reader asks them with answers to tap: will the Aurora condo be rented, and is it financed with Elena's VA loan? Priya taps yes to both: $2,050 a month from December 1, and a VA loan from 2019 with $268,400 owed.

It never fills a gap with a guess. A condo financed conventionally would leave Elena's full VA entitlement free; a condo left empty would put its whole payment in the ratio.

3

Nine programs tested at three down payments in 38 seconds

Thirty-eight seconds later the matrix is lit: nine programs against 0%, 5% and the 10% they described. Five cells are green (eligible, with payment, ratio and cash to close), seven amber (eligible with a stated change), fifteen gray (not eligible, with the reason). At their plan, three programs fit, and VA 30-year fixed comes first at $5,266 a month.

Amber cells say what would qualify: "Eligible with 5% down", "Eligible if 720 score", "Eligible if 12 months reserves". The run log beside the matrix shows which agent did what, and how long it took.

4

Switch on an honest change and the matrix re-lights

Under the matrix the near-miss finder has left two changes the Whitfields control: pay Marcus's card to 30% of its limit, or sell the Aurora condo before closing. Priya switches on the condo sale. Three cells turn green, each marked What-if: VA with nothing down at $5,889 a month, VA at 5% down, and the bank-statement loan.

She switches it off again; the Whitfields want to keep the condo. A what-if is a plan to discuss with the borrower; it stays off the file, and each one is logged.

5

The judgment call goes to the desk, then the note is saved

Counting the condo's rent on VA is a judgment rule at Larkmoor, so Priya asks the scenario desk at 8:59 am. The question arrives with the agents' draft and its rule: without landlord experience, Larkmoor wants a lease and three months of payments on both homes in reserve, 3 × ($5,266 + $1,640) = $20,718, and they will have $50,360 after closing. That reserve is Larkmoor's overlay: VA's regulation asks only that the home not be hard to rent, and a lease if one is available.6 Tom Reyes confirms at 9:04 am.

Priya saves the scenario note to the lead at 9:06 and calls the Whitfields with VA at 10% down, and the adjustable loan as a plan B. Nothing reaches them until she sends it.

From Jordan's note to Priya's call took 54 minutes in the sample, including the desk; the search itself took 38 seconds. The solution's design target is 38 seconds from a plain-words scenario to ranked programs, to be measured on your own scenarios.

Inside the program matrix: each cell's rules, near-miss plans, the scenario desk queue and the branch's overlays

In shortOne call is the story. These screens show the rest: each cell's rules opened, the ranked answer and the ten options, the side-by-side comparison, the near-miss plans with cost and time, the scenario desk, the program rules the answers cite, every scenario in the branch, and the settings where the shelf and overlays live.

Click any screen to open it larger. The lender, borrowers, people, rates and figures are fictional sample data.

Six scenario agents from call note to price; the loan officer picks what to present and the desk confirms judgment calls

In shortEach agent has one job and limits it does not cross. None of them sends anything to a borrower, quotes a locked rate, or suggests a change that misstates the borrower. The loan officer chooses what to present; a scenario-desk underwriter confirms every rule the lender marks as judgment, with name and time on the record.

Six scenario agents from call note to priced program matrix

The loan officer chooses options; the desk confirms judgment calls

Every search, what-if, desk confirmation and saved note is on the activity trail with the person's name and the time. Agent steps and human decisions are recorded apart, and scenario notes stay with the lead for seven years.

Priya, Jordan and Tom on the Whitfields' Lakewood purchase

Call notes, investor matrices, overlays and your rate sheet read; a program matrix, near-misses and a cited lead note produced

In shortIt reads the loan officer's own words, applies the agency and government guides, investor matrices and your overlays, prices from your rate sheet, and leaves behind a program matrix, near-misses with their changes, a cited note on the lead and a drafted borrower summary. Your pricing engine, loan origination system and CRM stay where they are.

Your CRM, pricing engine and LOS around a scenario search

Built for scale and for examiners: roles by branch, a maker-and-checker step for judgment rules, versioned overlays, an activity trail of every search and confirmation, and notes kept with the lead for seven years. Design targets from the solution: 38 seconds from a plain-words scenario to ranked programs, and all 27 program and down-payment cells tested and cited on every scenario. Both are targets to measure on your own scenarios.

What a wrong first program and a slow desk answer cost a mortgage lender

In shortMargins are thin: in 2025 a loan cost independent mortgage banks and bank mortgage subsidiaries $11,094 to make and earned them $785. A file that ends in a denial or a withdrawal spreads its cost over the loans that close. A borrower who hears about a better program from another lender is a loan you paid to find and did not make, and a borrower placed in a program they do not fit comes back as a defect.

Timing your own non-standard scenarios and desk questions

Second-lender calls and steering questions you cannot answer

Loan officer hours and loans kept by showing the right program first: the scenario search return

In shortPut in your own numbers. The calculator values the loan officer hours a search would release and the margin on a few extra loans closed because the right program was shown first, then subtracts what the tool and your testing cost. Every default marked illustrative is an assumption to replace.

Loan officer time and desk time on borrower scenarios

Your loan officer hours and loans kept on a year of scenarios

Score your shelf, overlays, rate sheet and judgment rules before software tests your borrower scenarios

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

Scenario desk scores: time scenarios, replay 100, or implement

Sixteen checks for a program-matching tool: missing facts, overlays, income counting and what reaches the borrower

In shortAsk every vendor the same questions, us included. The answers show how the tool reads a scenario, what it does with a missing fact, whose rules and overlays it applies, how it counts income, what it may suggest, what reaches the borrower, and how you would measure it on your own files.

We found no published independent test of any scenario or program-matching tool's accuracy, as of October 2026. A replay of your own scenarios is the fairest comparison, and the same replay measures your current process.

Replay 100 of last quarter's first-contact scenarios on that day's rules, then four weeks beside your loan officers

In shortMeasure first, then replay a quarter of your own scenarios on the rules in force that day, then run live beside your loan officers, then decide with four numbers. None of these steps commits you to a vendor.

1. Time five scenarios and export a month of desk questions. That is your baseline.
2. Write the shelf and overlays down, with the revision and effective date of each.
3. Pull 100 first-contact scenarios from last quarter, every outcome, with the rate sheets and guide versions of each day.
4. Replay them, then run live beside the loan officers for four weeks, with people deciding every presentation.
5. Decide with four numbers: better fits found, false fits, desk questions avoided, and loan officer minutes per scenario.

Measuring program matching: better fits found, false fits

Loan limits, VA entitlement, overlays, near-misses and when a scenario becomes an application: answers loan officers search for

In shortThe questions loan officers, brokers and buyers search for most, each answered on its own.