Case Study 39.2 — The Meriden Row File: A Loan Lost to Nobody

This is a labeled composite. The Meriden Row file is not a real loan. It is assembled from a documented and thoroughly ordinary industry pattern — a signature-dependent condition, requested once, that nobody followed — and every figure in it is constructed for teaching. It exists because the failure it describes almost never appears in a case study: it produces no enforcement action, no lawsuit, no denial letter, and no incident report. It produces a withdrawn application and a percentage point of fallout, and then it disappears.

Nothing in this file was declined. That is the point of studying it.

Background

A rate-and-term refinance. Owner-occupied, single-family, conventional, thirty-year fixed. The borrowers had a first mortgage they wanted to replace and a home equity line of credit they wanted to keep open, which is the ordinary preference of people who have a line and no current balance on it.

The constructed facts:

All figures constructed. Payments computed at the stated rates on \$268,000 over 360 months.

A twenty-nine-month break-even on a loan the borrowers intended to keep for many years. A clean file, an easy approval, six of seven conditions cleared in five days. There was nothing wrong with it.

Condition 7 is the whole case. When a first mortgage is refinanced, the existing HELOC — recorded second — would automatically move into first position when the old first is paid off and released, because lien priority follows recording order (Chapter 21). No investor will buy a first mortgage sitting behind a home equity line. So the HELOC lender must execute a subordination agreement consenting to stay in second position behind the new loan.

This is not hard. It is also not fast, and it is not yours. It requires a specific department at another institution to receive a request on its own form, review it against its own criteria, charge its own fee, and produce a signature. It is precisely the shape of condition §39.4 puts in its fourth category and Chapter 19 puts behind a Wednesday.

The issue: what actually happened, day by day

THE MERIDEN ROW CALENDAR                                [constructed composite]

  DAY   EVENT
    0   Application taken
    8   Rate locked, 6.500%, 45 days, EXPIRES DAY 53
   26   Conditional approval — 7 conditions
   27   Subordination request emailed to the HELOC lender's general
        servicing address. No confirmation requested. No trigger date set.
   31   Conditions 1 through 6 cleared. File is complete except for #7.
        -- LAST EVENT ON THIS FILE --
   50   Loan officer notices the file while pulling a report for something
        else. Calls the HELOC servicer. Learns: the request was never routed
        to the subordination department, that department uses its own form,
        the form requires a fee, and the published turnaround runs weeks.
   51   Correct form and fee submitted.
   53   LOCK EXPIRES. Market has moved; today's comparable rate is 6.875%.
   54   Borrowers are presented with the options below.
   57   Borrowers withdraw the application.

Nineteen calendar days between day 31 and day 50. During that window the file was documentation- complete except for one item, the item had been requested exactly once, and nobody knew the request had gone into a void.

Note the structural resemblance to the Linden Street file's eleven dead days — and note the one difference that makes this file worse. Linden Street was waiting on nobody, which the board's empty blocking cell can detect. Meriden Row was waiting on a named party, which looks entirely normal on a board. The blocking-party column said "HELOC servicer," which was true, and which concealed the fact that the HELOC servicer had never heard of this loan.

The options on day 54, and the arithmetic that decided it

Look at what the delay did.

Option A adds \$1,005.00 to a \$3,400.00 cost structure — a 29.6% increase in the cost of the transaction — and pushes the break-even from 28.8 months to 37.3 months, an increase of 8.5 months, purchased with nothing.

Option B is worse. A 0.375% higher rate cuts the monthly saving from \$118.06 to \$51.43 — a 56.4% reduction in the entire benefit of the loan — and pushes the break-even from under two and a half years to five and a half years.

The borrowers chose C, and they were right to. A refinance is a purchase of a payment reduction, and the price of that purchase had roughly doubled while the product got smaller. No amount of relationship, service quality, or apology changes that arithmetic, which is the humbling part: by day 54 there was no conversation available that could save this loan, because the loan was no longer worth doing.

What it shows

First: the file was never in trouble in any way an underwriter would recognize. Income verified. Credit was fine. Value supported. Six of seven conditions cleared in five days. If you read the file's underwriting record you would call it clean. It died in the gap between two organizations' inboxes.

Second: a named blocking party is not evidence that anybody is working. This is the sharpest and least comfortable lesson here, and it is a direct limitation of the board built in §39.10. The blocking-party column answers who owes the next action. It does not answer do they know they owe it. A request that was sent and never received produces a board that reads exactly like a request that was sent and is being worked.

The missing element is the one §39.8 puts in Level 0: set the trigger date at the moment you make the request. "Subordination requested day 27; if no acknowledgment by day 30, call." Three days. That single line, written on day 27, ends this case study on day 30 with a phone call and a corrected form, twenty-three days before the lock expires.

Third: the two filters would not have caught it, and the read-through would have. Run the board on day 38, a plausible weekly-review day:

   #  FILE          STAGE   DIS   Q  NEXT IRREVERSIBLE       BLOCKING
   -- MERIDEN ROW   COND     12   5  lock expires d+15       HELOC servicer

• Filter 1 (quiet ≥ 3 business days): five business days quiet. Trips.
• Filter 2 (irreversible ≤ 5 days): the lock is fifteen days out. Does not trip.
• Intersection: not flagged.

The chapter's priority filter misses this file. That is a real limitation and it should be stated plainly rather than explained away.

What does not miss it is §39.3's read-through, which covers every file whether or not it is flagged: "Meriden Row, conditions, twelve days in stage, five business days quiet, blocking party the HELOC servicer, lock expires day 53." Say that sentence out loud and the next question asks itself: what have we heard from them since day 27? The answer is nothing, and it takes one phone call to find out that nothing means nothing.

The intersection is the priority set. The read-through is the safety net. A loan officer who runs only the filters has built half a system.

Fourth: the metric did not notice. That quarter the loan officer locked 24 loans and funded 22.

$$\text{pull-through} = \frac{22}{24} = 91.7\% \qquad \text{fallout} = \frac{2}{24} = 8.3\%$$

A pull-through in the low nineties looks fine on any report. One of those two fallouts was a genuine credit decline the file could not survive. The other was Meriden Row, and in the branch's reporting it was coded as a borrower withdrawal — which it literally was.

Nothing in any report said "a request was sent to a void and nobody followed it." So over the next two quarters, the same failure happened four more times.

The pattern, once somebody looked

The four additional files had nothing in common by program, loan amount, borrower profile, or processor. What they shared was one structural feature:

Every one of them died on a condition requiring a signature from an organization the borrower belonged to and the loan officer did not — a homeowners association's management company, an employer's human resources department, a HELOC servicer, a payroll vendor.

And in every case, the request had been made exactly once, by email, to a general address, with no acknowledgment requested and no trigger date set.

That is a process defect, not a diligence defect, and it is worth being precise about the difference. A diligence defect is fixed by trying harder, which does not scale and does not survive a busy week. A process defect is fixed by changing one line in a workflow — in this case, by adopting a rule that sounds trivial and is not:

A request that requires an outside signature is not complete when it is sent. It is complete when it is acknowledged. Set the trigger date for the acknowledgment, not for the document.

Outcome, including the part that went wrong twice

The loan officer found the pattern in the fourth quarter, and then made a second mistake that is at least as instructive as the first.

Having discovered that files were dying quietly, they began escalating everything. Every third-party request went to a supervisor within a day. Every underwriting queue position generated a note to the underwriting manager. The processing team lead heard from them daily.

Within about six weeks, two things were true. Third-party turn times on their files had not measurably improved, because most of those parties were outside their company and unaffected by internal escalation. And their standing with their own operations department had degraded to the point that when a genuinely urgent file needed a Level 3 — a real exception, on a real deadline — the request arrived as one more item from the originator who escalates everything.

This is §39.8's fifth rule, learned expensively: the relationship budget is finite, and escalation spends it. The originator who escalates everything has no lever left for the file that actually needs one, and everyone in operations knows exactly who that originator is.

The eventual fix was not more escalation. It was three lines in a workflow:

1. Every outside-signature request gets an acknowledgment trigger date at the moment it is sent.
2. Every file gets read aloud in the weekly read-through, flagged or not.
3. The blocking-party column gets a second field — last confirmed contact — so that "HELOC servicer" and "HELOC servicer, confirmed Tuesday" are visibly different states.

The lesson

The board finds files that are quiet. It does not find files that are quiet for a reason that looks like a good reason. A named blocking party is the most convincing disguise a stalled file can wear, because it converts "nothing is happening" into "we are waiting on them," and the second sentence sounds like management.

Three things close that gap, and none of them is effort:

• A trigger date on every outside request, set when the request is made, keyed to acknowledgment rather than delivery.
• A read-through that covers every file, so that the priority filters are a fast path and not the whole review.
• A distinction on the board between "assigned" and "confirmed," because those are different states and only one of them means somebody is working.

And a fourth thing, which is a matter of judgment rather than process: know what a delay costs before you decide how hard to push. On this file the cost was computable from day 27 — a lock expiring day 53, an extension priced at 0.375 point (\$1,005.00), and a break-even that would move from 28.8 months to 37.3 months. Anyone who had done that arithmetic on day 27 would have called the HELOC servicer on day 30.

Discussion questions

1. The two-filter system in §39.10 does not flag this file on day 38. Propose a third filter that would, then argue against your own proposal by describing the false positives it would generate on a thirty-file board. Would you actually adopt it?

2. The blocking-party column said "HELOC servicer," which was true and misleading. Design the smallest possible change to that column that would have distinguished a sent request from a confirmed one — and say what it costs to maintain across thirty files.

3. Compute what the delay cost the borrowers under each option, and then answer a harder question: what did it cost the loan officer? Include the commission, the hours already spent, and something that does not have a dollar figure.

4. The loan officer's pull-through for the quarter was 91.7%. Write the two sentences you would want to see attached to that number so that it would have told somebody the truth. Then say who in a mortgage company should be responsible for producing them.

5. The over-correction — escalating everything — failed for two distinct reasons, one structural and one relational. Name both. Then describe a policy that would have raised escalation on outside-signature conditions without spending internal standing.

6. This composite and the Linden Street file fail in the same way and are detected differently. State the difference in one sentence, and say which of the two is more dangerous on a large pipeline and why.

7. A colleague reads this case and concludes that the real lesson is "call everybody more often." Using §39.5 and §39.6, explain why that conclusion would make their pipeline worse, and give the correction in a single sentence.