Case Study 2 — Twenty-Two Minutes on a Friday: An Application That Fell Apart on Day 40

Type: Clearly labeled composite, constructed from documented industry patterns. This is not one of the book's four anchor files and it is not a real borrower's file. Every figure is illustrative and internally consistent; none of it should be quoted as data. Call it the Fairmont Road file.

Why a composite here. Case Study 1 examined the public record of an industry-wide change. This one examines the ordinary private failure that record cannot show you, because nobody publishes it: an application taken in a hurry by a competent person, which produced four defects that all surfaced in the same week, five weeks later. The pattern is common enough that every experienced processor and underwriter will recognize it. The specifics are constructed so that the arithmetic resolves.

The file

A purchase of a three-bedroom on Fairmont Road at \$298,000, conventional, 10% down (\$29,800)**, loan amount **\$268,200, LTV 90.00%. Two borrowers, married, both on the loan. A 45-day contract.

The quoted housing payment, principal, interest, taxes, insurance, and mortgage insurance together, is \$2,214.00 a month. (Illustrative; the payment build is not the subject here.)

Day 5: the application

It was taken by telephone at 4:40 on a Friday afternoon and it lasted twenty-two minutes.

Nothing about that is unusual, and nobody involved was being lazy. The contract had been executed on Wednesday, the agent wanted the file moving, both borrowers work during the day, and Friday afternoon was the window they had. The loan officer was working from the pre-qualification notes, which were three weeks old, and had the borrowers on speakerphone in a car.

Here is what went into the loan origination system.

FAIRMONT ROAD — the 1003 as keyed, day 5              [composite; illustrative figures]

  INCOME (monthly, as stated on the call)
    Borrower 1 — salaried, 6 years .................... $5,400.00
    Borrower 2 — hourly, "about $3,500 a month" ....... $3,500.00
                                                      -----------
                                        TOTAL          $8,900.00

  MONTHLY DEBTS (as stated)
    Borrower 1 auto ...................................   $389.00
    Revolving minimums ................................   $266.00
                                                      -----------
                                        TOTAL            $655.00

  DECLARATIONS
    All fourteen answered in a single pass, 90 seconds, all "no."

  RATIOS AS KEYED
    Housing   $2,214.00 / $8,900.00 = 24.88%
    Back-end  $2,869.00 / $8,900.00 = 32.24%

A 24.88% housing ratio and a 32.24% back-end ratio is a comfortable file. The automated underwriting system agreed on day 8 and returned an approve recommendation. Everyone relaxed, which is exactly what a comfortable ratio built on unverified data is for.

What was actually wrong

Four defects. Every one of them was created in those twenty-two minutes, and not one of them was a lie.

Defect 1 — income taken as a total rather than as components. Borrower 2 was asked "what do you make?" and answered "about \$3,500 a month," which was true of his deposits. He is paid **\$16.50 an hour** and works overtime most weeks. His documented base is:

$$\$16.50 \times 2{,}080 \div 12 = \$2{,}860.00 \text{ per month}$$

The overtime is real, but he had been with this employer eleven months, and the file did not contain the history the guideline requires to average variable income. The underwriter used base only. Income fell \$640.00 a month and nobody had asked the question that would have exposed it on day 5: what is your hourly rate, and how much of that \$3,500 is overtime?

Defect 2 — a co-signed debt declared "no." Borrower 1 co-signed her brother's auto loan two years ago. The payment is \$412.00 and her brother has never missed one. When the declarations were read as a single sentence at speed, she heard a question about her debts and answered accurately about her own. The obligation is on her credit report, and it counts against her unless twelve months of payments made by the other party can be documented — which was theoretically available and practically not, because her brother pays cash at the credit union counter and has no canceled checks.

Defect 3 — a short sale declared "no." Borrower 1 sold a prior home in a short sale three years ago, in another state. She answered "no" to the short sale declaration for a reason that is entirely comprehensible: the bank agreed to it, so in her mind it was not a foreclosure and not a default — it was a negotiated sale. The event's seasoning had, as it happens, run. The problem was not that the answer was disqualifying. The problem was that it was undocumented, and proving the completion date required obtaining a settlement statement from a three-year-old transaction handled by a title company in another state that had since been acquired.

That took eleven days.

Defect 4 — the address history that did not reconcile. Because Borrower 1 had lived somewhere else three years ago, the two-year address history on the application did not match the addresses on the credit report. Under ordinary circumstances that is a two-line letter of explanation. Arriving in the same week as the other three, it read to the underwriter as one more thing the application had gotten wrong, and the file's credibility was gone.

The calendar

FAIRMONT ROAD — what the errors cost                  [composite; illustrative]

  day  0   contract executed, 45-day close
  day  5   application taken by phone, 22 minutes, 4:40 Friday
  day  6   initial disclosures delivered
  day  8   automated underwriting: approve, on the stated data
  day 11   appraisal ordered
  day 19   appraisal returns at contract value
  day 26   file submitted to underwriting (documents trickled in over 3 weeks)
  day 33   conditional approval, 14 conditions
  ──────────────────────────────────────────────────────────────────────────
  day 40   THE WEEK. Verification of employment returns base-only income.
           Credit review flags the co-signed auto. Title work on the prior
           property surfaces the short sale. LOE requested on the address gap.
  day 44   re-underwritten on documented figures; re-disclosed
  day 51   short sale completion date documented at last
  day 52   rate lock expires; 15-day extension purchased
  day 58   closing — 13 days past the contract date, on the second extension

The arithmetic of the correction

The back-end ratio moved 7.48 percentage points — from 32.24% to 39.72% — on a file where nobody lied, in a twenty-two-minute conversation, on the strength of two questions that were not asked and one that was asked too fast.

And here is the part that makes this case study worth reading: the file still qualified. At 39.72%, on these documented figures, the loan was approvable. The ratios survived.

The calendar did not.

What it cost

The loan closed thirteen days late, on the second contract extension addendum, after a seller who had already scheduled movers came within a day of relisting.

The rate lock expired on day 52 and had to be extended fifteen days. Priced at a quarter of a point — illustrative; extension pricing varies daily and by lender — that is:

$$\$268{,}200 \times 0.00250 = \$670.50$$

Six hundred seventy dollars and fifty cents, paid by somebody, for four questions.

The uncounted costs were larger. The borrowers spent five weeks believing they were comfortably approved and then two weeks believing they might lose the house, which is not a neutral experience for people who have never done this before. The underwriter, having found four defects in one file, verified everything else in it that she would otherwise have accepted, which added conditions that had nothing to do with the original errors. And the referring agent, who had sent this loan officer four files, sent the fifth one somewhere else.

What it shows

Every one of the four defects was reachable in three additional minutes on day 5. Not three hours. Three minutes: "What's your hourly rate, and how much of that is overtime?" "Have you ever co-signed for anybody — a car, a student loan, an apartment?" "These next four are the history questions, and anything here is workable, I just need to know today." And then reading the completed application against the credit report before running findings, which would have caught the address gap and the co-signed auto without asking the borrower anything at all.

Speed at application is borrowed, not saved. The twenty-two-minute application produced an eleven-day document hunt, a fifteen-day lock extension, a re-underwrite, and thirteen days past the contract date. The loan officer did not gain twenty-two minutes; they moved forty hours of work from a Friday when it was cheap to a Thursday five weeks later when it was expensive and public.

A "no" given at speed is not an answer, it is an absence of one. Both of the declarations that failed here failed for the same reason: the borrower was answering a question they had construed reasonably and differently. Nobody in this file did anything dishonest. That is precisely why reading the declarations one at a time works — not because it catches liars, but because it catches people who understood the question differently than the form meant it.

Credibility is a file-level property. After day 40 the underwriter did not evaluate four errors independently; she evaluated the application as unreliable and re-verified things she had already accepted. Accuracy at intake buys the benefit of the doubt everywhere else, and there is no way to buy it back once it is spent.

Discussion Questions

1. Rank the four defects by what they actually cost this file — not by how serious they sound. Which one drove the thirteen-day delay, and why is it not the one that moved the ratio most?

2. The case study says the file "still qualified." If the ratios survived, what exactly went wrong? Write the answer as a single sentence you could say to a new loan officer.

3. Borrower 1's "no" to the short sale declaration came from a defensible reading of the question: the lender agreed to the sale, so she did not experience it as a default. Rewrite that question the way you would ask it out loud, so that her answer would have been yes.

4. Recompute the day-40 back-end ratio on the assumption that the co-signed auto payment could have been excluded with twelve months of documented payments by the brother. Does the exclusion change any decision in this file? What does your answer suggest about where the loan officer's attention should have gone?

5. The loan officer took this application at 4:40 on a Friday from a car, working from three-week-old pre-qualification notes. Design a personal rule that would have prevented this file without making you unavailable to borrowers who can only talk on Friday afternoons. Be specific about what you would refuse to do and what you would do instead.

6. The referring agent sent the next file to a competitor. Was that fair? Argue it from the agent's side, then from the loan officer's, and then say what the loan officer should have done on day 40 — the day the problems surfaced — to have had any chance of keeping that relationship.