Chapter 10: Credit Analysis: Reading Credit Reports, Understanding Scores, and Credit Repair Strategies for Borrowers

"The score is the first thing everybody looks at and the last thing that tells you anything. The report is where the loan is." — constructed; what a twenty-year underwriter says to every new loan officer, eventually

Overview

You are going to pull credit on the Linden Street borrowers this morning, and forty seconds after the report comes back you will know something about this transaction that neither borrower knows, that their agent does not know, and that the seller they are competing against certainly does not know.

Here it is. One of them has a 742. The other has a 706. The loan is going to be priced at 706.

Not the average. Not the higher one. Not the score belonging to whoever earns more, or whose name goes first on the application, or who has been at their job longer. The lower of the two middle scores, and the pricing engine has no opinion about which human being produced it. That single fact will follow this file for fifty-one days. It sets the mortgage insurance factor. It sets the loan-level price adjustment that Chapter 29 will rebuild from the rate sheet. It is the reason the rate this household actually gets is not the rate on the website they were reading last night. And the entire apparatus of it — the report, the three bureaus, the three scores, the merge, the rule for picking one — takes about ninety seconds to explain and almost nobody explains it.

This chapter is about reading the document rather than glancing at the number on top of it. The score is a summary, and like all summaries it is a compression of something with far more information in it. The remaining term on an auto loan decides whether Chapter 4's ten-month rule applies. The credit limit on a card decides whether a \$400 payment moves a score or does nothing at all. The date a collection was opened decides whether it is a live problem or a seven-year-old scar that will fall off on its own. A loan officer who reads only the score is reading the cover of a book and reviewing it.

We will also be honest about the second half of the chapter's title. Credit can be improved, some of it quickly, and doing that work well is one of the highest-value things a loan officer does. But the work has hard limits, and the space between what genuinely helps and what is useless or actively harmful is full of people who charge money to stand in it. You will finish this chapter knowing which side of that line every technique falls on — and knowing the one sentence you may never say to a borrower, no matter how confident you feel.

In this chapter, you will learn to:

• Explain what a tri-merge report is, where it comes from, and why three bureaus disagree
• Say what a score actually measures — and what it contains no information about at all
• Determine the representative score for any file, with one borrower or four
• Read a tradeline completely and derive the remaining term from it
• Classify lates, collections, charge-offs, and public records, and say how each is treated now
• Compute credit utilization and identify which score factors can move in thirty days
• Distinguish a rapid rescore from a credit supplement from a dispute
• Work a thin file without a score, and state what you may never promise anyone

Learning Paths

🎓 Exam — §10.3 and §10.5. Representative-score selection is one of the most reliably tested calculations on the SAFE MLO test, and the derogatory vocabulary (charge-off versus collection versus public record) shows up as definition traps. §10.1's FCRA material is testable too. 🏠 New LO — §10.4 and §10.10. Reading a tradeline is a mechanical skill you can own in an afternoon and most of your competition never learns. §10.10 will keep your license. 🤝 Partner — §10.3 and §10.7. When an agent asks "can't they just pay something off?", §10.7 is the honest answer, and §10.3 is why the pre-approval says what it says. 📊 Operations — §10.8 and §10.1. Rescore timelines, supplement turn times, and report shelf life are calendar items, and every one of them can cost a lock extension.

10.1 The tri-merge and where it comes from

Start with a fact that surprises people who have worked in this business for years: there is no central credit database.

There are three large, competing, privately owned companies — Equifax, Experian, and TransUnion, usually called the three nationwide consumer reporting agencies — and each one maintains its own file on your borrower. They are not branches of anything. They do not share data with each other. They buy information from furnishers — banks, card issuers, auto lenders, servicers, collection agencies — and furnishing is voluntary. A credit union that reports to two of the three is not doing anything wrong. A landlord who reports to none of them is not doing anything wrong either.

The consequence is the whole reason this chapter exists. Three companies, three separately assembled files, three different sets of accounts, three different balances as of three different report dates, and therefore three different scores for the same person on the same morning. The spread on the Linden Street borrowers is thirteen points on one of them and eight on the other, which is ordinary. Spreads of forty points happen and usually mean one bureau is missing a whole account or carrying something that does not belong.

A tri-merge credit report is the document that solves this. It is not produced by a bureau. It is produced by a credit reporting agency — in mortgage usage, a reseller — that pulls all three bureau files simultaneously, reconciles the duplicates into one line per account, formats the result to residential mortgage lending standards, and returns the scores from each bureau alongside the merged data. Your loan origination system orders it, usually in under a minute, and it costs your company real money per pull.

WHERE THE REPORT COMES FROM                            [constructed teaching example]

    BORROWER
       │  signs authorization; you have permissible purpose
       ↓
    LOAN OFFICER / LOS  ──── order ────►  CREDIT REPORTING AGENCY (reseller)
                                                │
                        ┌───────────────────────┼───────────────────────┐
                        ↓                       ↓                       ↓
                    EQUIFAX                 EXPERIAN               TRANSUNION
                  own file, own            own file, own          own file, own
                  furnishers, own          furnishers, own        furnishers, own
                  score model              score model            score model
                        │                       │                       │
                        └───────────────────────┼───────────────────────┘
                                                ↓
                                     MERGE + DEDUPE + FORMAT
                                                ↓
                                    ONE TRI-MERGE REPORT
                                    3 scores per borrower
                                    1 line per account
                                                ↓
                        ┌───────────────────────┴──────────────────────┐
                        ↓                                              ↓
                  YOUR PRICING                              THE AUTOMATED DECISION
                  (representative score, §10.3)             (Chapter 15 — not here)

The furnishers on the left-hand side of that picture never see the report. The borrower does not receive a copy of it automatically from you — they are entitled to disclosures about the score, which §10.1's compliance note covers, but the merged mortgage report is the lender's document. And the report is a photograph taken on a specific morning. Every balance on it is as of whatever date that furnisher last reported, which is usually the statement cycle date, which is usually not today. That lag is not a defect. It is the single most exploitable fact in §10.7.

Two operational facts about the document that new loan officers learn the expensive way.

The report has a shelf life. Agency and lender rules generally require credit documents to be no more than a few months old at the note date — commonly four months, though the figure varies by program and by investor, so verify the current requirement in the applicable guide before you rely on it. On a forty-five-day contract that never matters. On a new-construction file that slips two quarters, it matters enormously, and you will re-pull, and the re-pull will show whatever the borrower has done since.

There is an older, more thorough cousin. A Residential Mortgage Credit Report is a fuller product in which the credit reporting agency independently re-verifies employment and account information rather than simply merging what the bureaus hold. It is expensive, slow, and rarely ordered now; the merged in-file report plus targeted supplements (§10.8) does the same job faster. Know the term exists, because underwriters and older guideline language still use it.

FIGURE 10.1 — "What a tri-merge actually contains"             [the Linden Street file]
  THE DOCUMENT     Merged residential mortgage credit report, both borrowers, joint
                   report, ordered day 1 by the loan officer through the LOS under
                   written borrower authorization. Produced by a credit reporting
                   agency reselling Equifax, Experian, and TransUnion data.
  THE CONTEXT      Day 1. There is a contract deadline at two o'clock. No income has
                   been documented, no assets verified, no application taken beyond
                   the information needed to pull. This is the first verified fact
                   in the file.
  WHAT IT SHOWS    Nine labeled sections, in this order:
                   1. HEADER — borrower names, addresses as reported, Social Security
                      number match indicators, report number, date, ordering lender.
                   2. SCORES — three per borrower, each labeled with the bureau and
                      the model version, each with up to four ranked reason codes.
                   3. ALERTS — fraud alerts, active security freezes, address
                      discrepancies, consumer statements, deceased indicators.
                   4. TRADELINES — one merged line per account: creditor, account
                      type, ECOA responsibility code, date opened, terms, high credit,
                      credit limit, balance, monthly payment, past due, status, and a
                      24-month payment grid.
                   5. PUBLIC RECORDS — bankruptcies (see §10.5 on what is no longer
                      here).
                   6. COLLECTIONS — separately listed, with the original creditor
                      named where reported.
                   7. INQUIRIES — every hard pull, with date and requesting party.
                   8. CONSUMER STATEMENTS — anything the borrower has asked a bureau
                      to attach to their file.
                   9. ADDRESS AND EMPLOYMENT HISTORY as reported by furnishers.
  WHAT IT DOESN'T  It contains no income, no assets, no employment verification you may
                   rely on, no rent history, and no record of anything the borrower pays
                   that is not furnished by a creditor. It does not show accounts opened
                   in the last several weeks that have not yet reported. It does not show
                   judgments or tax liens (§10.5). Sections 8 and 9 are unverified and
                   are informational only — never treat the employment block as a
                   verification of employment.
  THE DECISION     Read it before you say a number out loud. Then order nothing else
                   until you have read sections 2, 4, and 7 line by line.
  THE LESSON       The score sits in section 2 of nine. Eight-ninths of the document is
                   the part that decides whether this loan closes.

Constructed. Section names and ordering vary by credit vendor; the content set is standard.

⚖️ Compliance Check

You may not pull credit because you feel like it. The Fair Credit Reporting Act (FCRA) permits a consumer report to be obtained only for a permissible purpose, and the one you rely on is a credit transaction initiated by the consumer. In practice: get the borrower's authorization, keep it in the file, and never pull for a curious agent, a nervous seller, a spouse who is not applying, or "just to see." A pull without permissible purpose is a federal violation and, in most shops, a terminable offense on the first occurrence.

The borrower gets a score disclosure. When a credit score is used in connection with a residential mortgage loan, FCRA requires the applicant to receive a notice disclosing the score, the range of possible scores, the key factors that adversely affected it, the date, and the credit reporting agency that supplied it. Your system generates it. Read one, once, so you can explain it.

If the file is declined, the notice is not optional. The Equal Credit Opportunity Act and Regulation B require an adverse action notice with the specific reasons, and FCRA adds credit-report disclosures where a report was used. Chapters 24 and 25 cover the mechanics. Do not let a file die quietly in your pipeline without one.

Warn them about the phone calls, today, before they happen. When a mortgage inquiry posts, the bureaus may sell a prescreened lead — the industry calls them trigger leads — to other lenders who then make firm offers of credit. Your borrower can receive dozens of calls within forty-eight hours, some from callers who imply they are working on this loan. Tell them on the day you pull, tell them the calls are not from you, and tell them they may opt out of prescreened offers at the national opt-out service the bureaus jointly operate. Congress has repeatedly considered restricting mortgage trigger leads and the rules in this area have moved; verify the current law with your compliance department.

The report is nonpublic personal information. Gramm-Leach-Bliley Act safeguards apply. Do not email a full credit report to an agent, ever, for any reason.

Requirements change, state law varies, and this is not legal advice. Verify current rules with your compliance department and your regulator.

10.2 What a score is and what it is not

A credit score is a number produced by a statistical model that reads a credit report and estimates the relative likelihood that this consumer becomes seriously delinquent on a credit obligation within a defined future window — commonly the next twenty-four months. That is the whole claim. It is a rank-ordering device: it says that people who look like this go ninety days late more often than people who look like that.

FICO score is the common name for the family of models built by the Fair Isaac Corporation, which have been the dominant models in mortgage lending for decades. FICO's base scores run 300 to 850. VantageScore, a competing model developed jointly by the three bureaus, uses the same 300–850 range in its current versions. Both numbers look identical on a page and are not the same number.

Here is what a score is not, and every item on this list is a conversation you will have:

• It is not a measure of income, wealth, or ability to repay. The model never sees an income figure. A household earning \$400,000 a year can carry a 610. This is why debt-to-income (DTI) and the score are independent tests in Chapter 4's arithmetic, and why passing one tells you nothing about the other.
• It is not a judgment about a person. It is a probability statement about a population.
• It does not consider ECOA-prohibited bases. Race, color, religion, national origin, sex, marital status, and age are not model inputs, and neither are where the borrower lives or how much they earn.
• It is not one number. It is a specific model version, run against a specific bureau's data, on a specific day.
• It is not the number on the borrower's phone.

That last one deserves its own treatment, because it produces more friction on day one than anything else in this chapter.

Why the borrower's app says something different

Mortgage lending has long used older "classic" FICO versions delivered through the three bureaus — commonly identified as FICO Score 2 at Experian, FICO Score 5 at Equifax, and FICO Score 4 at TransUnion. Consumer-facing apps, card issuers, and free credit sites typically show newer FICO versions (8, 9, and later) or a VantageScore. Those models were built later, on different data, with different handling of collections, authorized-user accounts, and medical items.

So a borrower who checks a free app and sees 761 is not lying, is not confused, and did not misread anything. They are looking at a different instrument. A useful analogy: two thermometers in the same room, one calibrated in 1998 and one in 2020, both accurate to their own scale.

Model names and versions in mortgage use are set by agency and investor requirements and have changed. The Federal Housing Finance Agency has announced a transition toward newer models — FICO 10T and VantageScore 4.0 — and a move away from requiring all three bureaus toward a two-bureau requirement, with implementation timelines that have been revised more than once. Verify the current requirement before you tell a borrower which model will be used on their file.**

📞 On the Phone

Borrower: "That's wrong. My credit app says 761. I checked it this morning."

The answer that loses trust: "Those apps aren't accurate."

They are accurate. They measure a different thing, and telling a careful person that the tool they have been diligently monitoring for two years is junk makes you sound defensive and slightly uninformed.

What actually works: "It's not wrong, and neither is mine — they're different models. Yours is the current consumer version. Mortgage lending runs on an older version of the same family, and it's stricter about a couple of things, so it usually reads lower. That's not a technicality I'm inventing to justify a rate; it's a requirement, and every lender you call today is going to pull the same older model and get within a few points of what I got. The number I have to price your loan on is 706. Let me show you exactly what's driving it, because two of the four reasons the model gave me are the same reason, and it's fixable."

Three things happened there. You validated their number, you told them the competition will get the same result — which is true and saves you the shopping conversation — and you moved immediately from the number to the cause. Never leave a borrower alone with a score. Leave them with a task.

What the reason codes are for

Every score comes with up to four or five ranked key factors, sometimes called reason codes: short standardized phrases naming what most held this score down. They are genuinely useful and they are routinely misread.

What they are: a ranked list of the largest negative contributors, in the model's own vocabulary.

What they are not: a to-do list with point values attached. "Proportion of balances to credit limits is too high" does not come with a promise that fixing it is worth thirty points, and nobody — not you, not the bureau, not the vendor — can tell you what it is worth on this file.

On the Linden Street report, Borrower 2's four factors read, in the model's characteristic phrasing:

KEY FACTORS — Borrower 2, 706                            [the Linden Street file]
  1  Proportion of balances to credit limits is too high on bank revolving
     or other revolving accounts
  2  Amount owed on revolving accounts is too high
  3  Too many accounts with balances
  4  Length of time accounts have been established

Read those four again. The first three are the same fact stated three ways. The model is not listing three problems; it is listing one problem — \$8,400 spread across four cards, most of them near their limits — with three different labels on it. Factor four is time, and time is not for sale.

That is the whole diagnostic. One fixable thing, one unfixable thing, and §10.6 puts numbers on both.

10.3 The representative score rules

This section is the most consequential page in the chapter, and possibly in Part II. Chapter 29 rebuilds this file's rate from the number this section produces. Chapter 40's capstone turns on it.

The problem it solves: you have three scores for one borrower, or six for two, and pricing needs exactly one. So there is a rule, and it runs in two steps.

THE TWO-STEP RULE                                       [constructed teaching example]

  STEP 1 — reduce each BORROWER to one score
  ─────────────────────────────────────────────────────────────────────────
    three scores returned  →  use the MIDDLE one
                              (742 / 738 / 751 → sort → 738 742 751 → 742)

    two scores returned    →  use the LOWER one
                              (688 / 701 → 688)

    one score returned     →  use it
                              (subject to program rules on single-score files)

    two identical + one    →  use the DUPLICATED score
    different                 (640 / 640 / 672 → 640)

  STEP 2 — reduce the LOAN to one score
  ─────────────────────────────────────────────────────────────────────────
    one borrower           →  that borrower's score from step 1

    several borrowers      →  the LOWEST of the borrowers' step-1 scores

  ─────────────────────────────────────────────────────────────────────────
  LINDEN STREET
    Borrower 1   742 / 738 / 751   →  middle 742
    Borrower 2   706 / 712 / 698   →  middle 706
    THE FILE                       →  lower of {742, 706}  =  706

Notice what step 1 is doing. Taking the middle of three is a crude but effective way of discarding an outlier — if one bureau is missing an account or carrying something that does not belong, the middle score is the one least likely to be distorted by it. It is not an average. Do not average. Averaging 742 and 706 gives you 724, which is not this file's score and never was.

Notice what step 2 is doing. It is not asking which borrower is more creditworthy, or which one earns more, or which one is listed first on the application. It takes the weakest link, because the loan is a joint obligation and the investor is pricing the risk of the whole. The pricing engine does not care which borrower earned it.

On the Linden Street file, that is 706, not 742. Every number downstream — the mortgage insurance (MI) factor of 0.58%, the loan-level price adjustment, the rate itself — comes off the 706.

📄 Read the File

text FIGURE 10.2 — "Two middle scores, one qualifying score" [the Linden Street file] THE DOCUMENT Tri-merge residential credit report, pulled day 1, both borrowers. THE CONTEXT A $385,000 purchase, 5% down, conventional. Pricing has not been run yet. WHAT IT SHOWS Borrower 1: 742 / 738 / 751 -> middle 742. Borrower 2: 706 / 712 / 698 -> middle 706. For a conventional loan with two borrowers, the representative score is the LOWER of the two middle scores: 706. Four revolving accounts, $8,400 of balances, $212 in minimum payments, no lates in 24 months. No collections. No public records. Two auto loans: $487 with 31 payments remaining, $429 with 19 payments remaining. WHAT IT DOESN'T It does not show the $611 furniture account they will open in six weeks, because it does not exist yet. It does not show whether the $8,400 in revolving debt is seasonal or structural. It does not show income, and it cannot tell you whether either borrower can afford this house. A credit report is a photograph, not a film. THE DECISION Price the file at 706, not 742, and say so out loud on today's call before quoting anything. Then run the balances: paying two cards down below 30% utilization before the next reporting date is worth a rapid rescore -- IF the cash exists after closing costs, which is the arithmetic in §10.7. THE LESSON The representative score is the lower middle, and the pricing engine does not care which borrower earned it. Price what the file is, not what the better borrower is.

Constructed. Frozen figures for this book's running file.

The rule has moved, and you must verify it

Be careful here, because this is a place where a confident textbook can make a reader wrong.

The two-step rule above is the principle, and it is how the Linden Street file is priced throughout this book. But agency treatment of multiple-borrower score selection has varied over time and by program. There have been periods in which an agency used the lowest borrower representative score for eligibility while using an average of the borrowers' representative scores for pricing. Program rules differ — conventional, FHA, VA, and USDA do not all handle this identically — and the transition toward new scoring models and a two-bureau requirement will touch this area again.

So learn the structure, then check the source. Fannie Mae's Selling Guide and Freddie Mac's Seller/Servicer Guide are free, public, searchable, and authoritative; HUD Handbook 4000.1 governs FHA. The disciplined habit: quote conservatively at the lower score. If you quote at 706 and the pricing turns out to use something better, you have delivered good news. If you quote at 724 because you averaged, and the engine takes 706, you have to call your borrower three days before the lock deadline and take money back. One of those conversations builds a career.

🎓 NMLS Exam Watch

Representative-score selection appears on the SAFE MLO test as a two-step calculation, and candidates lose it by rushing step 1.

The classic stem: "Borrower A has scores of 715, 690, and 742. Borrower B has scores of 680, 705, and 699. What is the representative score for the loan?" Sort each: A is 690 / 715 / 742. B is 680 / 699 / 705. Lower of 715 and 699 is 699.

The trap: answer choices will include 707 (the average of the two middles), 715 (the higher middle), and 680 (the single lowest number on the page). All three are wrong for a reason, and all three are mistakes real loan officers make.

The other trap: a stem giving only two scores for one borrower. Two scores means take the lower, not the average of the two.

Also know: the score is not part of the report's tradeline data, the borrower is entitled to a score disclosure when a score is used on a mortgage application, and a score contains no income information.

The structural consequence: who goes on the loan

Here is a decision this section forces, and it has to be handled carefully.

If a co-borrower's income is not needed to qualify, leaving them off the loan can produce a better representative score and therefore a better rate. That is real, it is legitimate, and it is arithmetic you should be able to run.

It does not work on this file, and it is worth showing why. Borrower 1 alone brings \$6,300.00 of qualifying income. Borrower 1's own obligations — the \$487.00 auto, the \$318.00 in student loans, and the revolving minimums on the accounts they are responsible for — run \$857.00. Hold the housing payment at \$3,033.72 for the moment, which is generous, since a 742-only file would price a little better:

$$\frac{\$3{,}033.72 + \$857.00}{\$6{,}300.00} = \frac{\$3{,}890.72}{\$6{,}300.00} = 61.76\%$$

Sixty-two percent, against a conventional back-end that Chapter 4 established lives well below that. Borrower 2's \$4,200.00 is not optional. Therefore this is a 706 file, and there was never a version of it that was not. Say that on day one.

Two guardrails on the general technique, and neither is negotiable:

You may present the arithmetic. You may not steer, and you may not discourage. Regulation B prohibits discouraging an applicant on a prohibited basis, and a spouse who wants to apply is entitled to apply. Run both structures, show both numbers, and let the borrowers choose. Never tell someone their spouse "shouldn't be on the loan." Chapter 25 covers this properly.

Leaving a spouse off the note does not always leave them off the paperwork. In community property and homestead states a non-borrowing spouse frequently must still sign the security instrument, and in some programs their debts count anyway. That is state law and program law, and it varies enormously. Verify with your compliance department before you promise anyone anything about it.

One more note, and then we move on. The Linden Street borrowers are also shopping an online lender whose advertised rate is lower than yours. It will be worth remembering, when you get to Chapter 40, that an advertised rate is quoted for a file — and that the file it was quoted for has a considerably higher score than this one.

10.4 Reading a tradeline

A tradeline is one account on a credit report. Everything a loan officer needs from a credit report other than the score lives in tradelines, and reading them completely is a mechanical skill that takes about an hour to learn and that most loan officers never bother to acquire. It is, dollar for dollar, the best hour in this book.

Accounts come in two structural families, and the distinction matters to both the score model and the underwriter:

Revolving credit — a credit limit the borrower may draw against repeatedly, with a minimum payment that floats with the balance. Credit cards, store cards, most lines of credit. There is no fixed payoff date. The account never "ends," so it never falls out of the ratio on its own.

Installment credit — a fixed amount borrowed, repaid in scheduled payments over a fixed term. Auto loans, student loans, personal loans, mortgages. There is a payoff date, and therefore a remaining term — the number of scheduled payments left. That number is the reason Chapter 4's ten-month rule exists, and finding it is the single most valuable piece of tradeline reading you will do.

There is a third category you will meet — open accounts, where the full balance is due each cycle, such as some charge cards — and a handful of specialty types. Learn the two main families first.

Here is what a tradeline gives you, field by field.

📄 Read the File

```text FIGURE 10.3 — "One tradeline, four decisions" [the Linden Street file] THE DOCUMENT A single installment tradeline from the merged report, day 1.

CREDITOR RIDGEVIEW AUTO FINANCE ACCOUNT # ****7391 RESPONSIBILITY Borrower 2, individual (I) TYPE Installment / auto REPORTED BY EFX EXP TU (all three) STATUS Open / pays as agreed

---

DATE OPENED 53 months before this pull TERMS 72 months @ $429 HIGH CREDIT $26,400 CREDIT LIMIT n/a (installment) BALANCE $7,798 PAST DUE $0 MONTHLY PAYMENT $429 LAST PAYMENT current cycle

---

PAYMENT PATTERN most recent month FIRST, reading left to right, 24 months C C C C C C C C C C C C C C C C C C C C C C C C ^ ^ this month 24 months ago KEY C = paid as agreed 1 = 30 days 2 = 60 3 = 90 4 = 120+ - = nothing reported for that month

THE CONTEXT Day 1, before pricing. Chapter 4 has already established that an installment debt near the end of its term may be treated differently in the ratio. This tradeline is where you find out whether that applies. WHAT IT SHOWS Remaining term = 72 - 53 = 19 payments. The debt is real, current, individual to Borrower 2, and reported by all three bureaus. Twenty-four clean months. $429.00 goes into the ratio. WHAT IT DOESN'T It does not print "19 payments remaining" -- you derived that. It does not show today's balance, only the last reported one. It does not prove the payoff date; only a payoff statement or the servicer's amortization schedule does that, and an underwriter will ask for one before excluding anything. THE DECISION Count the $429.00. Nineteen is not ten. Then flag the OTHER auto -- $487.00 with 31 remaining -- as the one to re-check if this file sits until spring, because 31 payments today is 19 in a year. THE LESSON Two numbers on a tradeline that never appear as a number -- the remaining term and the utilization -- are the two that change the loan. Both are subtraction and division you do yourself. ```

Constructed. Field labels vary by credit vendor; the content set is standard.

Deriving the remaining term — two methods, one of which is a check

Method A — term minus age. The report gives you the original term and the date opened. Subtract.

$$72 \text{ months} - 53 \text{ months elapsed} = 19 \text{ payments remaining}$$

This is the reliable method whenever the report carries both fields, and it is what an underwriter will do.

Method B — balance divided by payment. \$7,798 ÷ \$429 = 18.2. Call it eighteen.

Method B is always a little low, because part of every payment is interest, so the balance is worth fewer payments than it appears. Use it as a sanity check, never as the answer. If the two methods disagree by more than a couple of months, something is wrong — a term reported incorrectly, a deferral, a modification, a skipped payment, a balloon — and you should order a supplement (§10.8) rather than guess.

And there is a diagnostic worth memorizing: if balance ÷ payment comes out HIGHER than method A's remaining term, the tradeline is internally inconsistent. It cannot take fewer months to retire a balance than the arithmetic of the balance itself allows. Something on that line is misreported.

The four revolving accounts

Here is the revolving side of the Linden Street file, with the two numbers the report does not print computed in the last two columns.

Read the responsibility column and then read the scores again. Three of the four accounts are Borrower 2's or joint, and the two accounts sitting near their limits are both Borrower 2's alone. Borrower 1 — the 742 — holds the one card with a \$9,000 limit and a \$1,900 balance. The thirty-six-point spread between these two people is not a story about character. It is a story about which of them has been carrying the household's revolving balances.

Two more field-level warnings.

A missing credit limit is a real problem. Some furnishers report a balance and a high credit but no assigned limit. Scoring models must then estimate utilization from the highest balance ever reported, which typically makes the account look far more heavily used than it is. If a card on your file shows no limit, that is a supplement request (§10.8) with genuine upside and no downside.

Student loans lie in a particular way. Borrower 1's \$318.00 is an income-driven repayment plan payment. Plans like that recertify annually, and the figure a servicer furnishes to the bureaus can be stale, can be \$0 during a deferral, or can differ from what the borrower is actually paying. The credit report is the starting point, not the authority. You document the current plan payment with the servicer's statement — and Chapter 4 owns the rule for which figure enters the ratio.

10.5 Derogatories: lates, collections, charge-offs, public records

A derogatory is any item on a credit report reflecting a failure to pay as agreed. The category runs from a single thirty-day late four years ago to a bankruptcy, and the range of severity inside it is enormous. Before the vocabulary, one instruction that governs everything in this section:

Write and speak about this material without moral content. People miss payments because a hospital billed the wrong insurer, because a marriage ended, because a plant closed, because a business failed in a pandemic, because they were twenty-two and nobody had ever explained a grace period. Bankruptcies, collections, and thin files are ordinary features of ordinary lives. Your job is to read the record accurately and tell the truth about what it does to this loan. Your job is not to have a view about the borrower.

Late payments

Furnishers report delinquency in thirty-day buckets: 30, 60, 90, 120, 150, 180. The 24-month grid shows the month each occurred. Two dimensions matter and they matter separately:

Severity — a 90 is much worse than a 30, and the jump from 30 to 60 is larger than the jump from "clean" to 30.

Recency — a 30-day late last month is worse than a 90-day late in year six. Scoring models weight recent delinquency heavily and the weight decays with time. It decays; it does not vanish. Adverse information generally remains reportable for seven years (§10.5's obsolescence discussion below).

For mortgage underwriting specifically, one late is not like another. A mortgage or rental late is weighted far more heavily by underwriters and by program guidelines than a store card late, because it is the closest available evidence of how this household treats a housing payment. A 30-day mortgage late in the last twelve months can move a file from automated approval to manual underwriting or can trip a program's eligibility requirement outright. Chapters 14 and 16 carry the specific waiting periods and requirements; here, learn to find them: read the grid on every mortgage and rental tradeline before you do anything else.

Read the grid direction indicator every time. Vendors differ — some print the most recent month at the left, some at the right. A loan officer who assumes the wrong direction will place a delinquency twenty-two months from where it actually sits, and will tell a borrower the wrong thing with complete confidence.

Collections

A collection is a debt that has been placed with or sold to a third-party collection agency, which then reports it as its own tradeline. Two features cause constant confusion:

The same debt can appear twice. The original creditor's account may show as charged off with a balance, and the collection agency's account shows the same debt again. That looks like two obligations and is one. Reading the "original creditor" field on the collection tradeline is how you catch it.

A collection can be tiny and still be there. A \$74 unpaid gym membership and a \$14,000 unpaid credit card are the same category of item and are not the same problem.

Treatment differs by program and the rules move. Broadly: conventional automated underwriting frequently does not require collections on a one-unit primary residence to be paid off, while FHA's handbook currently treats aggregate collection balances at or above a threshold specially — requiring payoff, a documented payment plan, or a percentage of the balance in the ratio — and excludes medical collections from that calculation. Those are exactly the rules that get revised, so verify the current Selling Guide and Handbook 4000.1 language rather than relying on this paragraph. Chapter 15 covers what the automated systems say about them; Chapter 16 covers FHA.

Charge-offs

A charge-off is an accounting event at the creditor: after a period of nonpayment — commonly around 180 days — the creditor writes the balance off its books as a loss. Two things follow, and borrowers frequently believe the opposite of both:

The debt is not forgiven. It is still owed. It can still be collected, sold, and sued on, subject to the applicable statute of limitations.

It is not "closed and done." It remains reportable, and it is one of the most severely weighted items a score model reads.

The distinction to hold: a charge-off is what the original creditor did with its accounting. A collection is what a third party is doing about the debt. They are different events and both can appear.

Public records

A public record on a credit report is court-derived information: bankruptcies, and historically civil judgments and tax liens.

Note "historically." Beginning in 2017, under a joint initiative the three nationwide bureaus adopted in connection with settlements with state attorneys general, the bureaus imposed enhanced identification and update-frequency standards on public record data. Most civil judgments and tax liens could not meet those standards, and they were removed from consumer credit reports — judgments beginning in mid-2017, with essentially all remaining tax liens removed by 2018. Bankruptcies remained.

The mortgage consequence is direct and nearly nobody explains it to new loan officers: a judgment or tax lien against your borrower will very likely not appear on your credit report, and it will absolutely appear on your title search. You will find it in week three, from the title company, in the middle of a transaction, attached to a property (Chapter 21). This is why the URLA's declarations in Chapter 9 ask the borrower directly, and why "the credit is clean" is not the same sentence as "there are no liens."

Bankruptcies remain reportable under FCRA's time limits: ten years from the date of entry of the order for relief or adjudication. In practice the bureaus have voluntarily removed Chapter 13 filings earlier than that — commonly at seven years from filing — which is a bureau practice rather than a statutory requirement, so do not promise a borrower a removal date.

The seven-year clock, and the thing it is not

FCRA sets outer limits on how long most adverse information may be reported: generally seven years for accounts placed for collection, charged to profit and loss, and most other adverse items; seven years from the date of payment for paid tax liens; ten years for bankruptcies. Criminal convictions are treated differently, and civil suits and judgments run seven years or until the governing statute of limitations expires, whichever is longer.

Two things this is not:

It is not the statute of limitations on the debt. How long a debt may be reported and how long a creditor may sue on it are different clocks governed by different law, and they routinely differ. Never tell a borrower a debt is "expired." You are not their lawyer.

It is not a clock a collector may restart. Re-reporting an old debt with a fresh date — sometimes called re-aging — is a documented consumer harm and an FCRA problem. If a borrower insists a collection is far older than the report shows and can produce anything supporting it, that is a legitimate dispute under §10.8, and it is one of the rare cases where a dispute mid-transaction may be worth its cost.

📞 On the Phone

Borrower: (after a long pause) "There's a collection on there, isn't there. It's from the hospital. My daughter had surgery in 2022 and I fought the insurance company for eight months and I thought it was handled. I'm sorry. I know how this looks."

What you never say: anything that accepts the apology. There is nothing to apologize for and accepting the apology confirms that there was.

What actually works: "Okay — first thing, you don't have to be sorry, and this is genuinely routine. Medical collections are the most common item I see and the rules treat them differently from other collections for exactly the reason you just described: the amount is usually disputed between two institutions and the patient finds out last. Here's what I'm looking at: it's \$318, it's from two years ago, and on this program it may not need to be paid off at all. Let me confirm that before either of us spends a dollar. What I do need from you is the last statement or letter you have from the hospital, and if you have any of the correspondence with the insurer, send that too."

Then do the thing that matters more than any of it: finish the call on the next fact, not on the collection. People who are ashamed of a credit report stop answering the phone, and a borrower who stops answering the phone loses the house. The fastest way to prevent that is to treat the item as the fourth-most-interesting thing you discussed.

10.6 The five score factors and which ones move fast

FICO publishes the general composition of its scores. The weights below are commonly published as approximate, they describe the population rather than any individual file, and they vary by model version. Treat them as a map of where the mass is, not as a formula.

THE FIVE FACTORS — approximate published weights          [schematic, not to scale]

  PAYMENT HISTORY        ███████████████████████████████████   ~35%
    have you paid on time; how late, how recently, how often

  AMOUNTS OWED           ██████████████████████████████        ~30%
    balances vs. limits (utilization), number of accounts with
    balances, how much of installment debt remains

  LENGTH OF HISTORY      ███████████████                       ~15%
    age of oldest account, average age, age of newest

  NEW CREDIT             ██████████                            ~10%
    recent inquiries and recently opened accounts

  CREDIT MIX             ██████████                            ~10%
    revolving and installment together
  ─────────────────────────────────────────────────────────────────
  Weights are approximate and vary by model version. Never quote
  them to a borrower as a formula that produces points.

Now overlay the only distinction that matters to a loan officer with a forty-five-day contract: how fast can each one change?

WHAT MOVES, AND WHEN

  ONE CYCLE (≈30 days) — and eligible for a rapid rescore with documentation
    │
    ├─ AMOUNTS OWED         pay a balance down; the next furnish reports it
    ├─ a reporting ERROR    wrong balance, wrong limit, not-my-account, duplicate
    └─ a MISSING LIMIT      supplement it and utilization is recomputed
    ─────────────────────────────────────────────────────────────────────
  MONTHS
    │
    ├─ NEW CREDIT           inquiries and new-account effects decay
    └─ recent delinquency   the weight of a recent late decays with time
    ─────────────────────────────────────────────────────────────────────
  YEARS — nothing you can do this quarter
    │
    ├─ LENGTH OF HISTORY    only time; opening accounts makes it WORSE
    ├─ PAYMENT HISTORY      an accurate late stays reportable for seven years
    └─ CREDIT MIX           requires new accounts, which cost you elsewhere
    ─────────────────────────────────────────────────────────────────────
  NEVER
    └─ disputing ACCURATE information (§10.8, and it can freeze the file)

Amounts owed is the entire opportunity. It is roughly thirty percent of the model, it is the only large component that can change within one billing cycle, and — critically — unlike payment history, it has almost no memory. Scoring models read the balances currently reported, not the average balance over two years. A card that has been at ninety percent for four years and is at ten percent on the day the model runs is read as a card at ten percent.

That is the mechanism. Now put numbers on it.

Credit utilization

Credit utilization is reported balance divided by credit limit, computed both per account and in aggregate across all revolving accounts. It applies to revolving credit only. Installment loans have their own, differently weighted measure — the proportion of the original loan still outstanding — and paying an auto loan down does far less for a score than paying a card down.

🧮 Run the Numbers

The Linden Street revolving balances, three ways.

Today, from §10.4's table:

	Balance	Limit	Utilization
Bank card	\$3,850 | \$4,500	85.56%
Retail store card	\$2,100 | \$2,500	84.00%
Credit union card	\$1,900 | \$9,000	21.11%
Department store card	\$550 | \$1,200	45.83%
Aggregate	\$8,400** | **\$17,200	48.84%

$$\frac{\$8{,}400}{\$17{,}200} = 48.84\%$$

Now price three targets. Models look at both the aggregate and the individual accounts, so the discipline is to bring every card under the threshold, not just the total.

Target	What it takes	Paydown	New aggregate
Every card under 50%	bank card to \$2,250; store card to \$1,250	\$2,450** | \$5,950 ÷ \$17,200 = 34.59%**
Every card under 30%	bank card to \$1,350; store card to \$750; dept. store to \$360 | **\$4,040**	\$4,360 ÷ \$17,200 = 25.35%
Every card under 10%	bank \$450; store \$250; credit union \$900; dept. \$120	\$6,680** | \$1,720 ÷ \$17,200 = 10.00%**

Check the middle row, since it is the one people actually do: \$8,400 − \$4,040 = \$4,360, and \$4,360 ÷ \$17,200 = 25.35%. Note also that the credit union card at 21.11% needs nothing at the 30% target, which is why the \$4,040 is smaller than borrowers expect when you first describe it.

What you may say about this: "This moves the biggest single component of the model, and it moves within one reporting cycle."

What you may not say: how many points it is worth. Nobody knows. §10.10.

Four practical notes that separate someone who has done this from someone who has read about it:

The date matters more than the amount. The balance that scores is the one the furnisher reports, which is generally the statement balance on the cycle date — not the balance after the borrower pays mid-cycle. Paying \$2,000 the day after a statement cuts and then waiting three weeks means the model sees the old balance for another month. Ask the borrower for the statement closing date on each card and time the payment to land before it. This is Chapter 4's theme in another costume: every day costs money.

Zero is not always optimal, and this is a footnote, not a strategy. Some model versions treat an all-zero revolving profile slightly differently from a small reported balance. The effect is small enough that it is not worth engineering around and not worth mentioning to a borrower who is trying to get to closing.

Do not close the paid-off cards. Closing a card removes its limit from the denominator, which raises aggregate utilization, and eventually removes the account's age. A borrower who pays off four cards and closes all four can end up worse than when they started. Say this out loud, twice.

Never open a new card to "improve the mix" during a transaction. Which brings us to the other half of the section.

New credit, inquiries, and the thing that undoes all of it

An inquiry is a record that someone accessed the report. Two kinds:

A soft pull — the borrower checking their own credit, an existing creditor reviewing an account, a prescreened offer — is visible to the consumer, is not shown to other lenders on reports furnished for credit decisions, and does not affect the score.

A hard pull results from a credit application the consumer initiated, is visible to other lenders, and can affect the score, typically modestly and temporarily.

Two published FICO behaviors matter here and both are worth telling borrowers:

Rate-shopping de-duplication. Mortgage, auto, and student loan inquiries falling within a shopping window are counted as a single inquiry. FICO publishes the window as 45 days in newer model versions and 14 days in older ones — and mortgage lending uses the older versions. So the practical advice is the tighter number: do all your mortgage shopping inside about two weeks. Shopping four lenders in nine days is one inquiry. Shopping four lenders across two months is four.

A short buffer. FICO's published logic ignores mortgage, auto, and student loan inquiries from the most recent thirty days when scoring. That is why the borrower who applies with three lenders on Monday does not watch their score drop on Tuesday.

⚠️ Where Deals Die

On day 41 of the Linden Street file, the borrowers walked into a furniture store.

They are eight days from closing. They have a conditional approval. They have been told the loan is "approved," because that is the word the industry uses and nobody explained the conditions. So they buy a bedroom set, a sofa, and a dining table with a nine-month promotional financing plan: \$5,200 financed, \$611.00 a month.

Nothing about that decision is unreasonable. They are buying a house; houses need furniture; the store offered a promotion; nobody told them not to.

Watch what it does. This file's monthly obligations were \$4,479.72 against \$10,500.00 of income — a back-end ratio of 42.66%. Add \$611.00:

$$\frac{\$4{,}479.72 + \$611.00}{\$10{,}500.00} = \frac{\$5{,}090.72}{\$10{,}500.00} = 48.48\%$$

The approval's debt-to-income condition is blown. Not close to blown. Blown by nearly six points, on a file that was already at 42.66%.

The mechanism that catches it: lenders re-pull credit before closing. Some run a soft refresh report a few days before the note date; many subscribe to an undisclosed-debt monitoring service that alerts them the moment a new inquiry or tradeline appears between application and closing. This is not exotic and it is not optional — it grew out of agency loan-quality requirements after 2008 and it is now standard. On this file the refresh runs day 44 and finds the account three days after it was opened.

What the disciplined loan officer does instead: you cannot prevent this after it happens, so you prevent it before. At application, in writing and out loud, in the borrower's language: "Between now and closing: no new credit, no new accounts, no cosigning, no large deposits you can't source, no job changes, and no furniture. Not one store card, not one 'no interest for a year' offer, not even an application you don't complete. If you want something, call me first and I'll tell you in two minutes whether it's safe. This is the single most common way a closing gets delayed and it is completely avoidable."

Then send it as an email so it exists in writing. Chapter 19 resolves what happens next on this file.

10.7 Legitimate credit improvement, with its limits

There is real work here. A loan officer who understands §10.6 can, on some files, do something worth thousands of dollars over a loan term. There is also an entire industry selling a fantasy version of this work, and the distance between them is the subject of this section and the next.

Here is the honest inventory.

The authorized user question

An authorized user is someone permitted to use another person's account without being contractually liable for the debt. The tradeline typically appears on the authorized user's report with the primary account holder's full history.

This is legitimate when it is real — a spouse on the household card, an adult child on a parent's long-held account. Two limits:

Scoring models handle it with suspicion. Newer FICO versions include authorized-user accounts but apply logic intended to blunt the effect of accounts added purely to inflate a score.

Underwriters may disregard it entirely. Agency guidance requires the lender to consider whether an authorized-user tradeline is an accurate reflection of the borrower's own credit history, and an underwriter may decline to give it weight — particularly if the borrower's file consists mostly of such accounts. Verify the current Selling Guide language.

And there is a version of this that is not legitimate: paying a stranger to add you to their account for the purpose of importing their history. That is a purchased tradeline, it is a documented fraud pattern, and Chapter 27 covers where it leads. Never suggest it, never facilitate it, and if a borrower's file shows several recently added authorized-user accounts on unrelated parties' credit, treat it as a red flag rather than an asset.

The limit that actually binds: the borrower's cash

Now the part almost no textbook prints, because it is where credit improvement collides with the rest of the file.

The paydown in §10.6 has to come from somewhere, and the somewhere is usually the same money the borrower needs at the closing table. Run the Linden Street file all the way through.

Verified assets: \$38,000.00**. Cash to close: **\$25,376.34. Left over: \$12,623.66, which against a PITI plus MI of \$3,033.72 is 4.16 months of reserves — a compensating factor the underwriter can see and use.

WHAT A PAYDOWN COSTS IN RESERVES                          [the Linden Street file]

  Verified assets                                        $38,000.00
  Cash to close                                        - $25,376.34
  ────────────────────────────────────────────────────────────────
  Reserves if they pay down nothing                      $12,623.66   = 4.16 months

  Option A: every card under 50%     pay $2,450          $10,173.66   = 3.35 months
  Option B: every card under 30%     pay $4,040          $ 8,583.66   = 2.83 months
  Option C: every card under 10%     pay $6,680          $ 5,943.66   = 1.96 months
  ────────────────────────────────────────────────────────────────
  Reserve months = remaining assets / $3,033.72 (PITI + MI)

Every one of those rows is a genuine trade. Option B moves the largest component of the score model and costs the file a third of its reserves. Option C nearly halves them. And not one of the three comes with a promise of a single point.

There is a second cost. Because reserves are a compensating factor and this file already runs a 42.66% back-end ratio at 95% loan-to-value (LTV), spending reserves to chase a score can weaken the file on one axis to strengthen it on another — and the underwriter evaluates the file as a whole. On some files the paydown is obviously right. On a thin-reserve, high-ratio file it can be obviously wrong. You have to actually run it.

There is also a question of which dollars. Part of the \$38,000.00 is a \$10,000.00 gift, and gift funds carry sourcing and permitted-use rules that Chapter 12 owns. Do not assume every verified dollar is available for a debt paydown.

How far is there to go, anyway?

One more piece of honesty. Conventional pricing does not read the score continuously; it reads it in bands. Published price-adjustment matrices have long been organized in bands roughly like this:

SCORE BANDS — the STRUCTURE only     [constructed teaching grid -- modeled on the
                                      structure of published matrices; the grids
                                      were substantially restructured in 2023;
                                      verify current values at the source]

   < 620 │ 620-639 │ 640-659 │ 660-679 │ 680-699 │ 700-719 │ 720-739 │ 740-759 │ 760+
         │         │         │         │         │    ▲    │         │         │
                                                      │
                                            706 is HERE, 14 points
                                            below the next band

That is the real question to ask before spending \$4,040: how far is the next band, and is the file plausibly within reach of it? At 706, the next band on a conventionally structured grid begins at 720 — fourteen points away. Fourteen points is a plausible outcome from taking two cards from the mid-eighties to under thirty percent. It is not a promised outcome, and the difference between "plausible" and "promised" is §10.10 and the rest of your career.

And note where the pricing lives on this file: at 706, the borrower-paid monthly MI factor is 0.58%, producing \$176.78 a month. That factor is set off the representative score. Chapter 29 rebuilds the whole quote from the rate sheet; Chapter 5 owns MI. The point here is only that the 706 is not a number in a report. It is a line item in a payment.

10.8 Rapid rescore, supplements, and disputes

Three tools, three completely different purposes, and loan officers confuse them constantly — sometimes expensively.

THREE TOOLS, THREE JOBS

  RAPID RESCORE          the data is going to change; make it change FASTER
    you have             documentation from the creditor that something is now
                         different (a balance paid, a limit reported, an error fixed)
    turn time            typically a few business days; varies by vendor
    who pays             the lender, not the borrower
    what it is NOT       not a dispute; not a guarantee of any score change

  CREDIT SUPPLEMENT      the report is silent or stale; go ASK
    you have             a question the merged report does not answer
                         (current balance, current payment, payoff, rating,
                         verification of rent, a missing limit)
    turn time            typically days
    who pays             the lender, as a file cost
    what it is NOT       not a score service; it updates the FILE, not the score

  DISPUTE                the report is WRONG; make the bureau reinvestigate
    you have             information the borrower believes is inaccurate
                         or incomplete
    turn time            statutory; generally 30 days, extendable to 45
    who does it          THE CONSUMER, with the bureau or the furnisher
    what it is NOT       not a tool for accurate information; not fast;
                         and mid-transaction it can freeze the file

Rapid rescore

A rapid rescore is a service in which the lender submits documentary evidence of a corrected or updated account condition to the credit reporting agency, which pushes the update to the bureaus on an expedited basis and returns new scores. It is the only mechanism that converts §10.6's arithmetic into a usable number inside a forty-five-day contract.

What it requires: documentation, from the creditor, of the changed condition — a paid-in-full letter, a zero-balance statement, a corrected limit, a letter confirming an account is not the borrower's. A borrower's screenshot of an app is not documentation. The creditor's letter is.

What it costs: real money per tradeline, and it is standard practice, under the bureaus' agreements with the resellers who provide the service, that the consumer is not charged for it. The lender absorbs it. Verify with your credit vendor; do not invoice a borrower for a rescore.

What it does not do: guarantee anything. It updates data. The model then does whatever the model does.

Two failure modes worth knowing. First, a rescore can move a score down — if the borrower has done something else since the original pull, the fresh calculation reflects all of it, not just the good part. Second, on some files the update simply does not produce a band change, and you have spent your company's money and the borrower's expectations for nothing. Order rescores where the arithmetic is strong and the documentation is clean, not hopefully.

Credit supplement

A credit supplement is a written verification or update of a specific item, obtained by the credit reporting agency directly from the source at the lender's request. It is a workhorse and it is not glamorous.

You will order supplements to: confirm a current balance or payoff, confirm the actual minimum payment when the report is stale, verify an account is closed and paid, verify twelve months of rent (Chapter 12 and §10.9), obtain a missing credit limit, or confirm the status of a collection. Half of Chapter 19's conditions are cleared with supplements.

Supplements update the file, not the score. If you need a score to move, you need a rescore.

Dispute

A dispute is a consumer's formal challenge to the accuracy or completeness of information on their credit report. FCRA gives the consumer this right and obliges the credit reporting agency to conduct a reinvestigation, generally within thirty days — extendable to forty-five when the consumer supplies additional information during the initial period. The agency must contact the furnisher, which has its own duties to investigate and report back. If the information cannot be verified, or is found inaccurate or incomplete, it must be modified or deleted. An agency may decline to reinvestigate a dispute it reasonably determines is frivolous or irrelevant.

Disputes exist for inaccurate information, and for that purpose they are a genuine consumer protection that a loan officer should support wholeheartedly. Accounts that are not the borrower's, balances that are wrong, an account reported as open that was paid four years ago, a collection reported twice, a debt discharged in bankruptcy still showing a balance — those are worth disputing, and a borrower with a real error should be told so plainly.

But a dispute is a legal process with a statutory clock, and it belongs to the consumer, not to you. You do not file disputes for a borrower. Direct them to the bureaus' processes, tell them what the timeline actually is, and document that you did.

⚠️ Where Deals Die

The dispute flag is the quietest file-killer in this chapter.

When a tradeline is under dispute, it carries a comment code saying so. That code has consequences nobody warns the borrower about:

• The automated underwriting system flags it and the lender must resolve the flag before the findings can be relied on. Chapter 15 covers what that message looks like.
• A disputed tradeline may be effectively unusable — some furnishers will not provide a supplement on an account under active dispute, so you cannot even verify the balance.
• Removing the flag is not fast. The consumer has to withdraw the dispute with the bureau and the furnisher has to update, and that can consume most of a thirty-day cycle.

Now the pattern that produces disasters. A borrower applies. Two weeks later, wanting to help, they hire a credit repair company — or a well-meaning relative tells them to "dispute everything." The company files blanket disputes across nine tradelines, several of which are perfectly accurate. Within two weeks the merged report comes back with dispute comments on five accounts, the automated findings are unusable, the underwriter suspends the file, the lock expires, and the contract's financing deadline arrives with no approval.

The disputes did not remove anything. The accurate items were verified by the furnishers and stayed exactly where they were. All the process did was cost the borrower forty days and, potentially, the house.

The prevention takes eleven seconds and belongs in your day-one script: "One more thing. Do not file any credit disputes between now and closing, and do not hire anyone to do it for you. If you see something on the report you think is wrong, tell me first — there's a fast way and a slow way, and the slow way stops the loan for a month."

10.9 Thin files and non-traditional credit

A thin file is a credit report with too few tradelines, or too little history, to support a conventional credit assessment — sometimes too little to generate a score at all.

The most important sentence in this section: no score is not the same thing as bad credit. A file can be thin because the household pays cash, because the borrower recently arrived in the country, because they are twenty-four, because they spent a decade abroad, because they closed everything after a bankruptcy and rebuilt without borrowing, or because they simply never wanted a credit card. None of those is a payment problem. A scoring model has nothing to measure and returns nothing, which is not the same as returning a low number.

The Harlow Street file sits nearby but is a different case, and the distinction is worth drawing.

A single borrower, one income of \$4,150.00 a month, \$395.00 in monthly debts, buying at \$215,000 with FHA financing and a \$10,000 forgivable county down-payment assistance second, at ratios of 41.48% front and 51.00% back — that file has plenty of credit history. It has a 641. Everything in §10.6 and §10.7 applies to it directly, and the fourteen-point question in §10.7 is a live one there in a way it may not be here.

Non-traditional credit

Non-traditional credit is a documented payment history from sources that do not furnish to the credit bureaus, used to establish a borrower's willingness to repay when the credit report cannot. The usual sources:

• Rent — the strongest one available, because it is the closest analogue to a mortgage payment
• Utilities: electric, gas, water, telephone, internet, cable
• Insurance premiums not deducted from payroll: auto, renters, medical
• Childcare, tuition, and other regular obligations
• Payments to a local retailer or a rent-to-own arrangement

Documentation is either a Non-Traditional Mortgage Credit Report — a product the credit reporting agency assembles by contacting the sources directly — or independent verification: twelve months of canceled checks, bank statements showing the payments, or letters from the source on letterhead.

The requirements are program-specific and they move. Broadly, programs that accept non-traditional credit look for twelve months of history from a small number of independent sources, with rent generally required as one of them when the borrower rents. Conventional and government programs handle no-score borrowers differently, and eligibility can depend on the number of borrowers with scores. Verify the current requirements in the applicable guide — Fannie Mae's Selling Guide, Freddie Mac's Seller/Servicer Guide, or HUD Handbook 4000.1 — before you tell a borrower what will be accepted.

Two developments worth knowing. The agencies have added consideration of positive rent payment history to their automated underwriting, drawing rent payments out of bank statement data; the mechanics belong to Chapter 15. And rent-reporting services now exist that furnish a tenant's rent payments to one or more bureaus, which over time thickens a file legitimately.

Building a file, honestly

When a borrower's file is genuinely too thin and the purchase is not urgent, the right answer is sometimes a calendar rather than a loan. The legitimate builders:

• A secured credit card, used lightly and paid in full monthly. Twelve months of history.
• A credit-builder loan from a credit union.
• Rent reporting, prospectively.
• Being added as an authorized user on a household member's seasoned account (§10.7's limits apply).
• Keeping the first accounts open. Age is the one thing that cannot be bought.

Every one of those has a runway measured in months. Saying so costs you a transaction this quarter.

The conversation, in one line: "I can probably get you approved for something today, and I don't think you should take it. Give me nine months and a secured card and you'll buy this same house at a payment you'll actually like. Put a reminder in your phone for March and put my number in it."

A borrower who is told that closes a loan with you in March, and tells four people about it in the meantime. It is also, independently, the correct thing to do.

10.10 What you may never promise

This is the shortest section in the chapter and the one to read twice.

You may never promise a score outcome. Not "this will get you to 720." Not "that'll be worth about forty points." Not "trust me, once that reports you'll be fine." You do not have the model. Nobody outside the model has the model. Every legitimate description of a score change is conditional and approximate, and the moment you attach a number to a promise you have created an expectation you cannot meet and, on a bad day, a claim you cannot defend.

You may never advise a borrower to dispute accurate information. It does not work — the furnisher verifies the item and it stays — and §10.8 showed what it can do to a transaction in flight. There is a worse version: advising a consumer to make a statement to a credit bureau that is untrue is specifically prohibited by federal law and is not a gray area.

You may never refer a borrower to an operation charging advance fees to dispute accurate items. The Credit Repair Organizations Act governs this industry. Among other things it prohibits a credit repair organization from charging or receiving payment before the promised services are fully performed, requires a written contract and a specified disclosure, gives the consumer three days to cancel, and prohibits advising a consumer to make untrue or misleading statements to a consumer reporting agency. An outfit that takes \$99 a month up front to "fix" a report is at minimum operating outside that statute's fee rule, and what they actually do is mail blanket disputes.

Instead, when a borrower needs help beyond what you can give: refer them to a HUD-approved housing counseling agency. The counseling is free or low-cost, the agencies are legitimate, and the referral costs you nothing and protects everyone.

You may never pull credit without permissible purpose, never charge a borrower for a rapid rescore, and never coach anyone to omit a debt from an application. That last one is not a credit issue. It is fraud, and Chapter 27 explains what it costs.

You may never discourage an application. A borrower with a 580 score, a bankruptcy, and three collections is entitled to apply and entitled to a written decision. You may tell them honestly what you think will happen. You may not tell them not to bother.

Here is what you may do, and it is a great deal:

• Read the report completely and explain it in plain language, on the first call
• Compute utilization and show the arithmetic
• Identify errors and get them corrected with a supplement or by directing the borrower to the dispute process
• Order a rapid rescore where there is documentation and the arithmetic supports it
• Model scenarios honestly: "if the balances report at 25% instead of 85%, the largest component of the model improves; I cannot tell you by how much"
• Time a paydown to the reporting date
• Tell a borrower the truth about a nine-month runway
• Price the file at what it actually is, today, out loud

⚖️ Compliance Check

The sentence that ends careers is optimistic, not dishonest. Almost nobody sets out to mislead a borrower about credit. What happens is that a loan officer, wanting to be encouraging on a hard call, says "pay those two cards down and we'll get you to 740 easy." The borrower liquidates a retirement account to do it, the score comes back 718, the pricing does not change, and now there is a family with less money, the same rate, and a recording of you promising something.

The rules in play:

• FCRA — permissible purpose for every pull, the score disclosure to the applicant on a residential mortgage, and the consumer's right to dispute inaccurate information through the statutory reinvestigation process. Disputes belong to the consumer.
• The Credit Repair Organizations Act — no advance fees, written contract, three-day cancellation right, and a flat prohibition on advising a consumer to make untrue statements to a credit bureau. Know it well enough to recognize an operation that is violating it, because your borrowers will be marketed to by several.
• ECOA / Regulation B — no discouraging an applicant, and an adverse action notice with specific reasons when a file is declined.
• UDAAP — unfair, deceptive, or abusive acts or practices. A promised score outcome that does not materialize is exactly the shape of a deception claim, whether or not you meant it.
• GLBA — the report is nonpublic personal information. Protect it accordingly.

The safe formulation, which you should say in these words: "I can't promise you a number — nobody honest can, and anyone who does is selling you something. What I can promise is that I'll tell you exactly what this file prices at today, I'll tell you what would have to change, and I'll tell you the day it changes."

Requirements change and state law varies. This is not legal advice. Verify current requirements with your compliance department and your regulator.

🗂️ The Loan File

Chapter 10 contribution: the credit report, and the number that prices the loan.

Day 1. The report is back, and this file now has its first verified fact.

FIGURE 10.4 — "The credit piece of the file"                 [the Linden Street file]
  THE DOCUMENT     Merged tri-merge residential credit report, both borrowers,
                   pulled day 1, before the pre-approval letter is issued.
  THE CONTEXT      A $385,000 purchase at 4412 Linden Street. Two borrowers, married,
                   both on the loan, first-time buyers. Conventional, 5% down.
                   Nothing else in the file is verified yet.
  WHAT IT SHOWS    SCORES
                     Borrower 1     742 / 738 / 751   ->  middle 742
                     Borrower 2     706 / 712 / 698   ->  middle 706
                     REPRESENTATIVE SCORE FOR THE FILE   =  706
                   REVOLVING -- four accounts
                     bank card          $3,850 / $4,500 limit   85.56%   $97
                     retail store card  $2,100 / $2,500 limit   84.00%   $63
                     credit union card  $1,900 / $9,000 limit   21.11%   $38
                     department store   $  550 / $1,200 limit   45.83%   $14
                     ────────────────────────────────────────────────────────
                     totals             $8,400 / $17,200        48.84%  $212
                   INSTALLMENT
                     auto, Borrower 1   $487.00/mo   31 payments remaining
                     auto, Borrower 2   $429.00/mo   19 payments remaining
                     student loans, B1  $318.00/mo   income-driven plan, documented
                   DEROGATORY
                     no lates in 24 months.  no collections.  no public records.
                   TOTAL MONTHLY DEBTS  $487 + $429 + $318 + $212  =  $1,446.00
  WHAT IT DOESN'T  It does not show the furniture financing account -- $5,200 balance,
                   $611.00 a month -- that these borrowers will open on DAY 41.
                   That account does not exist yet. Nothing in this document, and
                   nothing any loan officer could do on day 1, reveals it.
                   It also does not show income, assets, the property, or any judgment
                   or tax lien (those come from title -- Chapter 21).
  THE DECISION     1. Price at 706. Say it on today's call, before quoting anything.
                   2. Count $1,446.00 of monthly debts. Nineteen payments is not ten,
                      so Chapter 4's ten-month rule does NOT apply to either auto.
                   3. Model the revolving paydown -- and run it against cash to close
                      before recommending it (§10.7).
                   4. Deliver the no-new-credit instruction, out loud and in writing.
  THE LESSON       The most important number on this document is not the score. It is
                   the 19, because it is the one an underwriter will make you prove.

What this settles. The representative score is 706. Total monthly debts are \$1,446.00. There are no derogatories to clear, no collections to negotiate, no public records to explain. This is a clean file with one soft spot: 48.84% aggregate revolving utilization concentrated on Borrower 2, who happens to be the borrower whose score prices the loan.

What it does not settle. Everything else. Chapter 11 documents the income that turns \$1,446.00 into a ratio. Chapter 12 verifies the \$38,000.00 that has to cover \$25,376.34 of cash to close and still leave reserves. And this report is a photograph of day 1 — it says nothing about day 41.

One number to carry forward. The ten-month rule does not apply to this file. But run the counterfactual once, because it is the clearest demonstration in the book of why the remaining term is worth reading: if Borrower 2's auto had nine payments remaining instead of nineteen, excluding that \$429.00 would take the back-end ratio from 42.66% to 38.58% — four full points, from a number on a tradeline that the report never prints.

$$\frac{\$4{,}479.72 - \$429.00}{\$10{,}500.00} = \frac{\$4{,}050.72}{\$10{,}500.00} = 38.58\%$$

Open questions carried forward:

• Q10.1. Should they pay revolving balances down, given that the money competes with cash to close? (Chapters 12 and 13)
• Q10.2. What does the 706 actually cost, in basis points and in dollars? (Chapters 29 and 30)
• Q10.3. What happens when a credit refresh runs before closing? (Chapter 19 — and it will)

Your task. In Appendix C's workbook, record the credit page of the file. Enter both borrowers' three scores, circle each middle, and write the representative score. Enter the four revolving accounts with their limits and compute utilization per account and in aggregate; check that your total matches \$8,400 and 48.84%. Enter both auto loans with their remaining terms. Then write one sentence you would actually say on the phone today, in the borrower's language, explaining why the loan prices at 706 and not 742. Say it out loud before you write it down. If it takes more than twenty seconds, it is not finished.

Conclusion

The report came back in under a minute and it decided more about this transaction than anything else that will happen for the next fifty days.

Three bureaus hold three separately assembled files, so three scores come back for each borrower. The middle of each borrower's three is that borrower's score; the lowest of the borrowers' scores is the file's. On Linden Street that is 706, and the pricing engine has no opinion about which of the two people at the table earned it. Agency treatment of multiple-borrower selection has moved before and will move again — teach the structure, verify the current rule, and quote conservatively at the lower number.

Below the score sits the document that actually matters. A tradeline carries the payment that goes into the ratio, the balance and limit that produce utilization, and the two figures the report never prints: the remaining term, which you get by subtraction, and the utilization, which you get by division. On this file the remaining term is nineteen, so Chapter 4's ten-month rule stays on the shelf — and the counterfactual, four full ratio points, shows what it would have been worth.

Improvement is real and bounded. Amounts owed is roughly thirty percent of the model, it has almost no memory, and it can change in one reporting cycle — which is why timing a paydown to the statement date and ordering a rapid rescore is genuine, valuable work. Age, payment history, and mix are not for sale at any price, and neither is a promise. You may model, compute, document, correct, and tell the truth. You may not guarantee a point.

There is one more thing the report does not show, and it is the reason Chapter 19 exists. On day 41 these borrowers are going to buy furniture, on nine-month promotional financing, at \$611.00 a month. The day-44 refresh will find it, and the back-end ratio will be 48.48%.

Next: Chapter 11 documents the income. Credit told you what they owe; income tells you what an underwriter is permitted to count — which, as the Fulton Avenue file is about to demonstrate, is not the same thing as what they earn.

Key Terms

Tri-merge credit report — a merged residential mortgage credit report produced by a credit reporting agency that pulls all three nationwide bureaus, reconciles duplicate accounts into one line each, and returns each bureau's score alongside the merged data. (Ch.10)

FICO score — the family of credit scoring models built by the Fair Isaac Corporation, whose base scores run 300–850; mortgage lending has long used older "classic" versions delivered through the three bureaus. (Ch.10)

Representative score — the single score used to qualify and price a loan: for each borrower, the middle of three scores (or the lower of two); for the loan, the lowest of the borrowers' scores. Program and agency treatment has varied — verify current requirements. (Ch.10)

Tradeline — one account as it appears on a credit report, with its creditor, responsibility code, date opened, terms, high credit, limit, balance, payment, status, and payment history grid. (Ch.10)

Revolving credit — an account with a credit limit the borrower may draw against repeatedly, with a minimum payment that floats with the balance and no scheduled payoff date. (Ch.10)

Installment credit — an account with a fixed amount borrowed and a fixed number of scheduled payments, and therefore a remaining term. (Ch.10)

Credit utilization — reported revolving balance divided by credit limit, measured both per account and in aggregate; the largest component of a score that can change within one reporting cycle. (Ch.10)

Derogatory — any credit report item reflecting a failure to pay as agreed, from a single 30-day late to a bankruptcy. (Ch.10)

Charge-off — a creditor's accounting write-off of a balance after prolonged nonpayment; the debt is not forgiven and remains collectible. (Ch.10)

Collection — a debt placed with or sold to a third-party collection agency, which reports it as its own tradeline; the original account may appear separately. (Ch.10)

Public record — court-derived information on a credit report. Bankruptcies remain; most civil judgments and tax liens were removed from consumer credit reports beginning in 2017 and are now found through the title search instead. (Ch.10)

Inquiry — a record that a party accessed the credit report. A hard pull follows a consumer-initiated application and can affect the score; a soft pull does not. (Ch.10)

Soft pull / hard pull — an access that does not affect the score and is not shown to other lenders, versus one that does and is. (Ch.10)

Rapid rescore — a lender-ordered, documented, expedited update of tradeline data to the bureaus producing a fresh score in days rather than a cycle; not a dispute, not a guarantee, and never charged to the borrower. (Ch.10)

Credit supplement — a written verification or update of a specific credit item obtained by the credit reporting agency directly from the source at the lender's request; updates the file, not the score. (Ch.10)

Dispute — a consumer's formal challenge to inaccurate or incomplete credit report information, triggering an FCRA reinvestigation generally within 30 days; it belongs to the consumer, and a disputed tradeline can stall a loan in process. (Ch.10)

Authorized user — a person permitted to use an account without contractual liability for the debt; the tradeline appears on their report, and both scoring models and underwriters may discount it. (Ch.10)

Thin file — a credit report with too few tradelines or too little history to support a conventional assessment, sometimes too little to generate a score at all. No score is not the same as bad credit. (Ch.10)

Non-traditional credit — documented payment history from sources that do not furnish to the bureaus — rent, utilities, insurance, tuition — used to establish willingness to repay when the credit report cannot. (Ch.10)

Spaced Review

1. (Ch.10) Two borrowers apply together. Borrower A's scores are 704, 688, and 731. Borrower B's are two scores only: 745 and 739. What is the representative score for the loan, and what are the two most likely wrong answers a new loan officer would give?

2. (Ch.10 + Ch.4) A tradeline shows: opened 62 months before the pull, terms 72 months at \$511, and a balance of \$4,980. State the remaining term, show the check using balance ÷ payment, and say which Chapter 4 rule this tradeline puts in play.

3. (Ch.9 + Ch.10) The URLA declarations ask the borrower whether there are any outstanding judgments against them. Given what §10.5 established about public records, explain in two sentences why that question exists on the application even though you have a credit report in front of you — and name the document that will settle it.

4. (Ch.4 + Ch.10) On the Linden Street file the back-end ratio is 42.66%. If Borrower 2's \$429.00 auto payment had nine payments remaining instead of nineteen, what would the ratio become? Show the numerator.

5. (Ch.10) A borrower asks you to file disputes on four collections for them — three accurate, one genuinely not theirs. Write the three sentences you actually say, in order.

6. (Ch.9 + Ch.10) Chapter 9 established the six items that constitute an application. Explain why a loan officer who has taken all six but has not yet pulled credit still cannot issue a pre-approval letter — and name the specific number they do not have.