Chapter 39: Running the Pipeline: Managing Thirty Files at Once Without Dropping One

"Nothing is ever lost on the file you were worried about." — constructed; the line a branch manager repeats until it stops being a joke

Overview

Chapter 6 taught you what happens to one loan. This chapter is about what happens to your career when there are thirty of them, and it is a completely different problem with a completely different failure mode.

Here is the shape of it. A new loan officer's first genuine crisis is almost never a hard file. Hard files announce themselves. A self-employed borrower with declining income, an appraisal thirty-five thousand dollars under contract, a title commitment with somebody else's lien on it — those are difficult, but they are visible, and difficulty concentrates attention rather than scattering it. The first crisis is twelve easy files at the same time. Every one of them is simple. Every one of them has four or five live dates, a dozen documents in various states of arrival, and at least two people waiting on an answer. Together they exceed the thing a human being is actually able to do, which is hold the state of a complicated situation in their head.

And the file that dies is not the hardest one. It is the quiet one.

This is the single most counterintuitive fact in production lending, and it is worth sitting with before we go any further. Noise is a signal that somebody is paying attention. An agent who calls you every morning is holding that file's calendar for you. A borrower who emails twice a day is auditing your progress for free. An underwriter who returns a stip sheet is telling you exactly what remains. The loud files are the safe files. The file that has said nothing for eleven days is the file where nobody — not the borrower, not the agent, not the processor, not the underwriter, and not you — is holding any state at all. That is the one that expires, or lapses, or goes stale, or gets away.

The Linden Street file is the proof, and this chapter is going to be hard on it. That file took fifty-one days against a contract that named day 45. It was documentation-complete on day 33. Then nothing happened for eleven days, during which the lock expired and the borrowers financed five thousand two hundred dollars of furniture. Two failures that look independent are the same failure, and it is the subject of this chapter.

In this chapter, you will learn to:

• Name the failure mode of a growing pipeline and explain why quiet is the danger signal
• Decompose a file's calendar into stages and say who owned each day
• Run a weekly pipeline review that forces a decision on every file
• Triage by what is about to become irreversible, not by what is shouting
• Build a milestone cadence with a floor, and compute what it saves
• Batch a week so that acquisition and administration both fit in it
• Hand a file to processing without creating an ownership gap
• Escalate on a date rather than on a feeling
• Define pull-through and fallout, and say what each one costs and to whom
• Build a pipeline board that finds the file that is about to die

Learning Paths

🎓 Exam — this is the least directly testable chapter in the book, and you should know that going in. The exception is §39.9, where the Equal Credit Opportunity Act's notification timing turns a pipeline-hygiene question into a compliance question. Read that section and the Compliance Check inside it. 🏠 New LO — all of it, twice, and then build the board in §39.10 before you take your next application. This chapter is the difference between a fifteen-file year and a fifty-file year. 🤝 Partner — §39.5 and §39.7. If you are a real estate agent wondering why some lenders never make you ask, the cadence in §39.5 is the whole answer, and you can ask a lender for it by name. 📊 Operations — §39.2, §39.3, and §39.9. The velocity decomposition is the analytical tool; the weekly review is the control; pull-through is the metric that connects your desk to the secondary marketing desk.

39.1 The failure mode nobody warns you about

Every training program in this business teaches you how to originate a loan. Almost none of them teach you how to originate the ninth loan while the other eight are still moving, and that is the skill that actually determines whether you have a career.

Let us be precise about what breaks.

A single purchase file contains somewhere between forty and sixty discrete facts that matter — two incomes with their components, four or five debts with balances and remaining terms, three credit scores per borrower, an appraised value, a lock rate and a lock expiration, a contract price and a contract date, a list of conditions with owners and due dates, a title commitment with exceptions, an insurance binder, a closing date, a settlement agent, and a wire deadline. It also contains about a dozen live dates, of which perhaps four are genuinely unforgiving.

One file is easy to hold. Three files is easy to hold. At around six or eight, something quietly changes: you stop holding the files and start holding a summary of the files, and the summary is where the errors live. At twelve you are no longer holding even the summary. You are responding.

And when you are responding, you have — without deciding to, without noticing — adopted a work-selection rule. That rule is: work whatever is making noise.

It is a perfectly reasonable rule. It feels like responsiveness. It generates good customer-service reviews. It is also exactly backwards, for a reason that takes a moment to see.

Why noise is safety and silence is risk

A file makes noise when somebody outside your head is tracking it. There are only a few sources of noise in this business, and every one of them is somebody doing part of your job for you:

• The agent who calls every morning. Annoying, and also a free second set of eyes on the contract calendar. That file will not lapse a contingency without somebody mentioning it.
• The anxious borrower. The Harlow Street borrower calls twice a week and has nearly walked away three times. That is exhausting. It also means that file cannot go eleven days without a status check, because the borrower will not permit it.
• The underwriter who returns conditions. A stip sheet is a work order. It arrives, it is specific, it creates its own to-do list.
• A vendor with an automated reminder. The title company's system emails you about the outstanding payoff. Free monitoring.

Now list the ways a file goes quiet. There are exactly three.

1. It is genuinely complete and waiting on a scheduled event — a closing already set, a rate lock running out its term with everything cleared. Rare, benign.
2. It is waiting on a third party who has stopped working it and told nobody. The appraisal order that was never assigned. The verification of employment sitting in an HR inbox. The HOA certificate nobody signed.
3. It is waiting on you, and you have forgotten.

Only the first is safe. And here is the operational problem: from your inbox, all three look identical. Absence of email is absence of email. There is no message that says "nothing is happening on this file and that is a problem." You cannot distinguish safe silence from fatal silence by feel. You can only distinguish it by looking, deliberately, at every file on a schedule — which is why the rest of this chapter is mostly about a board and a calendar rather than about effort.

The two failure modes, and which one is worse

When a pipeline exceeds a loan officer's ability to run it, two things happen. New originators fear the first and are destroyed by the second.

The dropped file. One loan dies. It is dramatic, it is memorable, it costs a commission and a relationship, and it is rare enough that you will remember all of them.

The degraded pipeline. Nothing dies. Every file just gets slower. Conditions that used to clear in a day clear in three. Milestone updates go out late or not at all. The agent who used to send you two deals a quarter starts sending you one, and never tells you why, because there is nothing to tell you — you did not do anything wrong, you were just a little slower than the other guy. This costs vastly more than the dropped file and it produces no incident to learn from.

The degraded pipeline is worse because it is a feedback loop, not a threshold. Watch the mechanism:

THE CAPACITY SPIRAL                                    [constructed teaching example]

   more files
       │
       ↓
   slower responses ──────────────┐
       │                          │
       ↓                          │
   borrowers and agents           │
   start checking in              │
       │                          │
       ↓                          │
   unplanned inbound rises        │
       │                          │
       ↓                          │
   less time for scheduled work ──┘
       │
       ↓
   slower responses  (and around again)

This is the thing to understand about capacity: it is not a number of files you can hold. It is the number of files you can hold at a given service level, and the loop above means you do not discover the limit and stop — you cross it and then accelerate. The symptom is never "I have too many files." The symptom is "I have no time," which sounds like a personal failing and is actually a structural one.

The fix is not to work harder inside the loop. The fix is to break the loop by making the inbound endogenous — by generating the information people are calling to get, before they call. That is §39.5, and the arithmetic is startling.

⚠️ Where Deals Die

The quiet file. Not the hard one. The one that has not generated an email in a week and therefore never reaches the top of anybody's list.

The mechanism on the Linden Street file is textbook. All nine prior-to-doc conditions cleared by day 33. The processor's queue went empty. The underwriter had nothing to return. The borrowers had nothing to send. The agent had a closing date on the calendar for day 45 and no reason to worry about it yet. Every single person involved in that transaction had a defensible reason to think about something else, and every single one of them did, for eleven days.

In those eleven days: the lock expired on day 42 — costing a fifteen-day extension at 0.250 point, \$914.38** — and the borrowers financed **\$5,200 of furniture on day 41 at \$611.00 a month, which took the back-end ratio from 42.66% to 48.48% and nearly ended the transaction on day 44.

Notice that neither event was caused by difficulty. The file was finished. Nobody made a mistake of analysis. The failure was that a completed file with slack in its calendar was treated as a file that needed nothing, when in fact it needed exactly one thing: a closing date moved up into the slack.

The disciplined version: a file that is documentation-complete is not a file that is done. It is a file with an unconverted asset — days — and days expire.

There is one more thing to say before the analysis. The instinct at this point is to conclude that the loan officer should have been more attentive. That is true and it is useless. Attention does not scale; you cannot be twenty percent more attentive across thirty files. What scales is structure — a board, a review, a cadence, and a triage rule that survives a bad week. The rest of this chapter builds them, and it builds them out of a single closed file, because the fastest way to learn to run a pipeline is to perform an autopsy on a loan that funded.

39.2 File velocity and where days actually go

File velocity is the rate at which a file moves through the stages of origination, measured in days per stage rather than days in total. Total days is a vanity metric — it tells you the file was slow and nothing about why. Velocity is diagnostic, because a file is not uniformly slow. It is fast in six places and catastrophically slow in one, and until you decompose it you will "fix" the wrong stage.

So decompose the Linden Street file, and do it without mercy.

The contract was written the night of day 0 and named a day-45 closing. It executed on day 4, which means only 41 days actually remained when the clock started for real — a distinction Chapter 6 draws and one that every "45-day contract" hides. The file funded on day 51.

The seven stages

WHERE THE FIFTY-ONE DAYS WENT                              [the Linden Street file]
  Each block = one day. Day 0 (the agent's call) to day 51 (funding).

  lead -> application         d0-d5    #####                        5    9.8%
  application -> orders out   d5-d7    ##                           2    3.9%
  waiting on third parties    d7-d19   ############                12   23.5%
  assembly -> submission      d19-d23  ####                         4    7.8%
  underwriting turn           d23-d28  #####                        5    9.8%
  CONDITIONS                  d28-d47  ###################         19   37.3%
  clear to close -> funding   d47-d51  ####                         4    7.8%
  ------------------------------------------------------------------------------
  TOTAL                                                            51  100.0%
                                             (percentages rounded; they sum to
                                              99.9 and the missing tenth is
                                              rounding, not a lost day)

Read that chart before reading the commentary, because most loan officers guess wrong about which bar is longest. The common guess is underwriting. Underwriting was five days — under ten percent of the file. The common second guess is the appraisal. The appraisal came back on day 16, nine days after it was ordered, and it came back at value with no gap.

The longest bar is conditions: nineteen days, thirty-seven percent of the entire file. The second longest is third-party wait: twelve days.

Take the third-party window first, because it contains a real lesson that is not the chapter's main argument. Orders went out on day 7. Two things were ordered, and they returned on different tracks:

Nine days and twelve days, running in parallel. The file could not move to assembly until both were in, which means the appraisal's nine days were free — they were absorbed entirely by the title track. The binding constraint in that window was title, at twelve days, and the appraisal turn time was irrelevant to the calendar.

This matters because of where loan officers put their anxiety. Everybody chases the appraisal. It is the order with the borrower's money attached, the one the agent asks about, and the one with a value outcome that could kill the deal. Nobody calls the title company on day 10 to ask how the search is going. On this file, the call that would have mattered was the one nobody made.

The general rule: in any stage with parallel orders, only the slowest one is on the critical path, and it is usually not the one generating the noise. Find the binding constraint before you spend a phone call.

Who owned each day

A second decomposition, and the more useful one. Instead of naming stages, name the party who could have made the file move faster on each day.

Five plus six plus twelve plus five plus five plus eleven plus three plus four is fifty-one. Every day is accounted for and every day has an owner except eleven of them.

The two largest blocks on this file are twelve days you could not control and eleven days nobody was controlling at all. One of those is a fact about the world. The other is a decision that was never made.

Inside the conditions block

Nineteen days is the biggest bar on the chart, and it is not one thing. Break it open:

INSIDE THE CONDITIONS BLOCK — 19 days                      [the Linden Street file]

  d28-d33   clearing 9 prior-to-doc conditions   #####       5 days
            (day 29, day 30, day 33 — only THREE business days;
             days 31 and 32 were a weekend)

  d33-d44   NOTHING                              ###########  11 days

  d44-d47   the crisis and its cure              ###           3 days
  -----------------------------------------------------------------------
                                                             19 days

Five plus eleven plus three is nineteen.

Now look at what the first bar actually represents. Nine conditions — six owed by the borrowers, two by third parties, one by the lender — were issued on day 28 and all nine were cleared by day 33. Because days 31 and 32 fell on a weekend, that was three business days. Three business days to produce signed disclosures, thirty days of paystubs for two borrowers, a written verification of employment with commission continuity, two signed 4506-C forms, a letter of explanation and source documentation for a \$4,900 deposit, a gift letter with evidence of transfer, a cleared title commitment after a mechanic's lien was released and re-recorded, an insurance binder, and a mortgage insurance certificate.

That is excellent work. It is genuinely excellent work — nine items, three business days, two of them requiring a third party's cooperation. If this chapter were about condition-clearing we would stop here and applaud, and Chapter 19 in fact does.

And then the file sat for eleven days.

The indictment

Here is the whole argument of this chapter, stated as plainly as it can be stated.

On day 33 the Linden Street file was documentation-complete. Every prior-to-doc condition was satisfied. The only two conditions left were prior-to-funding items — a verbal verification of employment and the pre-closing credit refresh — which by definition are performed immediately before the note date and cannot be "worked" in advance. There was nothing left to do on that file except schedule it.

The lock expired on day 42. That is nine days later. The contract named day 45. That is twelve days later. The file had, sitting unclaimed in its calendar, more than a week of slack.

Nobody converted it.

On day 41, inside that window, the borrowers walked into a furniture store and financed \$5,200 at \$611.00 a month, which is a completely ordinary thing for people who are about to move into a house to do, and which nobody had recently reminded them not to do. On day 42, inside that same window, the rate lock expired on a file that had been ready to close for nine days, and the lender paid \$914.38 for a fifteen-day extension.

Two independent failures, one cause: slack that was never converted into an earlier closing date.

That sentence is the chapter. Everything after this section is machinery for making sure it does not describe your files.

🧮 Run the Numbers

The closing that was available on day 37.

Work forward from day 33 using nothing but this file's own facts and the same TRID count the file actually used.

• Day 33 (Monday): all nine prior-to-doc conditions cleared. The file is documentation-complete.
• Day 34 (Tuesday): clear to close issued; Closing Disclosure prepared, issued, and received.
• The consumer must receive the Closing Disclosure no later than three business days before consummation. On the actual file, the CD was received Tuesday day 48 and the closing was Friday day 51 — Wednesday, Thursday, Friday. Apply the identical count from Tuesday day 34: Wednesday day 35, Thursday day 36, Friday day 37.
• Closing available: day 37.

Now price the difference against what happened.

	Available	Actual	Difference
Closing day	37	51	14 days
Days inside the lock (expires day 42)	5 to spare	9 past	no extension needed
Lock extension cost	\$0.00 | \$914.38	\$914.38
Days before the contract date (day 45)	8 early	6 late	14 days
Furniture financed day 41	after funding	before funding	the entire day-44 crisis

The \$914.38 in perspective.** The extension was 0.250 point on \$365,750 — exactly half the 0.500 point (\$1,828.75) the borrowers paid to buy the rate down to 6.625% in the first place. It was lender-paid as a tolerance cure, so it never touched the borrowers' cash to close, but it is real money and it came out of this loan. Spread across the fifteen days it bought, that is \$914.38 ÷ 15 = \$60.96 per day** — sixty-one dollars a day, for days the file did not need, purchased because eleven earlier days were not used.

And be honest about what closing early would have cost. Prepaid interest runs from the closing date to the end of the month. At this file's per-diem of \$66.3861, closing on day 51 (October 24) meant 8 days of prepaid interest — **\$531.09. Closing on day 37 (October 10) would have meant 22 days — 22 × \$66.3861 = **\$1,460.49. The extra fourteen days are 14 × \$66.3861 = **\$929.41. (Subtracting the two rounded totals gives \$929.40; the penny is rounding, and this book would rather name a penny than hide one.)

That is more, at the closing table, than the \$914.38 the lender paid. It is also not waste. It is interest on money the borrowers actually have, for fourteen days in which they actually own the house, and the first payment is December 1 either way. The extension bought nothing. The prepaid interest bought two weeks of ownership. Compare like with like: one of those is a purchase and one is a penalty.

(Every figure above is a frozen figure from this file except the day-37 date, which is derived from the file's own TRID count.)

One caution before we move on, because the arithmetic above is cleaner than reality. You do not always get the earliest available closing date. The seller has a contract with a date on it, may have a move-out scheduled, and may have their own purchase closing the same week. The settlement agent has a calendar. You may ask for day 37 and be told day 44.

That is fine, and it does not weaken the argument, because the goal was never "close as early as possible." The goal is close inside the lock with margin. On this file the disciplined target is not day 37; it is somewhere around day 40 — two days inside the lock, five days before the contract date, and — as it happens — one day before a furniture store would have mattered. The prize is not speed. The prize is margin, and margin is what absorbs a day-41 surprise without a day-44 crisis.

39.3 The weekly pipeline review

If you take one operational habit out of this book, take this one.

The weekly pipeline review is a fixed appointment, on the same day and at the same time every week, in which every open file in your pipeline is examined in sequence and a decision is made about each one. Not a status check. A decision — the next specific action, its owner, and the day it happens.

Three properties make it work, and all three are non-obvious.

It is exhaustive. Every file, including the ones you are certain are fine. The value of the review is entirely concentrated in the files you would not have thought about, and you cannot know in advance which those are. A review that covers "the ones I'm worried about" is not a review; it is the noise heuristic wearing a suit.

It happens whether or not you have time. This is not motivational language. The week you are too busy for the review is, by construction, the week your pipeline is largest and your inbound is highest — which is precisely the week a file will slip. Cancelling the review in a busy week is like cancelling the smoke detector inspection during a fire.

It reconciles two records. Your board is a view. Your loan origination system is the record. Chapter 36 covers what the LOS does and does not do well; the relevant fact here is that an LOS is excellent at storing state and mediocre at surfacing absence — it will happily show you a milestone that has been sitting unchanged for eleven days without ever telling you that eleven days is strange. The review is where the board and the LOS get re-synchronized, and where you catch the file whose LOS status says "In Processing" and whose actual status is "nobody has touched this since the third."

The agenda

Forty-five minutes for a thirty-file pipeline. Run it with your processor if you have a dedicated one — and if you do, the division of labor below is worth adopting verbatim.

THE WEEKLY PIPELINE REVIEW — a 45-minute agenda        [constructed teaching example]

  0:00  RECONCILE (5 min)
        Board against the LOS, file by file. Fix the board, not the memory.
        Anything in the LOS that is not on the board is the first finding.

  0:05  THE READ-THROUGH (12 min)
        Every file, one sentence, out loud, no discussion:
          "Linden Street, conditions, nine days in stage, four business days
           quiet, lock expires day 42, blocking party: nobody."
        The LO reads the CALENDAR. The processor reads the CONDITIONS.
        Interrupting the read-through is how a review becomes a meeting.

  0:17  THE FLAG PASS (5 min)
        Apply the two filters (see 39.10):
          - quiet 3+ business days
          - next irreversible date inside 5 days
        Circle the intersection. Do not discuss yet.

  0:22  TRIAGE (10 min)
        Rank the flagged files by 39.4's rule. Assign each one to a SPECIFIC
        HOUR on a SPECIFIC DAY next week. "Follow up" is not an assignment.

  0:32  THE FORWARD PASS (8 min)
        Every file scheduled to close in the next 14 days:
          - does the CD timing work backward from that date?
          - is the lock long enough, with margin?
          - is anything outstanding that needs a third party's signature?
        This is where next week's emergency gets prevented.

  0:40  THE DEAD PASS (5 min)
        Anything quiet 10+ business days with no scheduled event:
          revive it today, or make a decision about it today.
        A file with no decision is a file you are storing, not running.
        (Dead files have a compliance tail — see 39.9.)

Two notes on the agenda.

The read-through is deliberately mechanical. The temptation is to stop and solve each problem as it surfaces, and if you do that, the review takes two hours, you never reach the last eight files, and the last eight files are alphabetically unlucky rather than genuinely fine. Separating finding from fixing is what lets you get through thirty files in forty-five minutes. Find everything first. Then decide.

The forward pass is the highest-yield eight minutes in your week. It is the only routine in this chapter that looks at the future rather than the present. Working backward from a scheduled closing date — is the CD timing achievable, is the lock long enough, does anything need somebody else's signature — surfaces problems while they are still cheap. A lock that is three days short is a trivial problem on day 20 and a \$914.38 problem on day 42.

What the review would have caught

Run the Linden Street file through this agenda on day 37, a Friday, four days after the last condition cleared.

• Read-through: "Linden Street, conditions, nine days in stage, four business days quiet, lock expires day 42, blocking party: nobody."
• Flag pass: quiet four business days — flagged. Next irreversible date five days out — flagged. It trips both filters.
• Triage: the file is complete and the lock expires in five days. Next action: call the settlement agent and the listing side today and move the closing up.
• Forward pass: issue the Closing Disclosure today, and the arithmetic changes. Received Friday day 37, the precise count that governs the CD clock runs Saturday day 38 — Saturdays count — skips Sunday day 39, and skips Monday day 40, the second Monday in October and a federal legal public holiday. Tuesday day 41 is the second, Wednesday day 42 the third: closing may occur on day 42, one day inside the lock. Wait until Monday to send it and the identical three days land on Thursday day 43 — outside the lock, and the \$914.38 is owed after all. One afternoon's delay in issuing a disclosure is worth a business day here, and the two business-day definitions and one federal holiday are why.

That review takes about ninety seconds for that one file, and it saves \$914.38 and prevents the day-44 crisis outright, because the pre-closing credit refresh would have run before the borrowers ever walked into the furniture store.

Ninety seconds. That is the entire cost of the discipline this chapter is arguing for, on the file it is arguing about.

🔍 Check Your Understanding

1. Why does the read-through forbid discussion, and what specifically goes wrong when you allow it?
2. The Linden Street file's LOS milestone on day 37 reads "Conditional Approval — Conditions Received." What is true about the file that the LOS milestone does not say, and which column on your board says it instead?
3. Your review is scheduled Friday at 3:00. It is Friday at 3:00 and you have four files closing Monday. Argue both sides in one sentence each, then say what you actually do.

(3 is the real one. The honest answer is that you run the review anyway and shorten the read-through, because a thirty-file pipeline with four Monday closings is the exact condition under which a fifth file is being forgotten.)

39.4 Triage: which file gets your next hour

You have thirty files, four uninterrupted hours, and eleven things that all feel urgent. Triage is the discipline of deciding what to work in what order — and the decision has to be made on a rule, because if it is made on feel it will be made by whatever shouted most recently.

Start by discarding the rule everybody uses.

"Work the urgent file" is not a rule. Urgency is a property of communication, not of risk. A file is urgent when someone has told you it is urgent, which means urgency measures the assertiveness of the people attached to a file. It correlates weakly with what will actually cost you money, and — per §39.1 — it correlates negatively with the thing you most need to find, because the file with no one shouting is the file with no one watching.

The rule that works is different and it is not intuitive.

Triage by time-to-irreversibility

Rank files by how soon something on them becomes irreversible, or irreversibly more expensive. Not by how loud, not by how big the loan, not by how much the borrower likes you. By how soon a door closes.

Irreversibility in mortgage origination has four common shapes:

1. A clock that expires and costs money to restart. - A rate lock hitting expiration (Chapter 30). The lock is the archetype: it expires on a specific day, extension is priced in points, and worst execution or a re-lock at market can be far more expensive than the extension. - A payoff quote with a good-through date. - A credit report, appraisal, verification, or set of AUS findings aging past its permitted window. Every one of these has an age limit measured in days, the specific number is set by the applicable guide and by your lender's overlays, and it changes — verify the current figure in the Selling Guide, HUD Handbook 4000.1, or your own credit policy before you rely on it. A document that goes stale is not merely re-orderable; re-pulling credit means re-running findings, and re-running findings means a new opportunity to discover a \$611.00 furniture payment.

2. A contract right that lapses. - The financing contingency, the appraisal contingency, the inspection contingency (Chapter 20). These do not extend themselves and they do not warn you. The day after a financing contingency lapses, your borrower's earnest money — \$5,000 on the Linden Street file — is exposed. The right to walk away is worth exactly nothing the day after it expires and cannot be repurchased.

3. A regulatory waiting period that has not started. - The Closing Disclosure's three-business-day receipt requirement (Chapter 24) is the one that bites weekly. It is not a task; it is a countdown that must begin early enough. If the CD has not issued, the closing date is not a closing date, it is a wish.

4. A condition that needs somebody else's signature. - This is the one new loan officers systematically misjudge, and Chapter 19 states the rule: when a condition needs somebody else's signature, find that out on a Wednesday, not at four o'clock on a Friday.

That fourth category deserves a moment, because it is a rule about lead time, not about difficulty. An HOA certificate, a corrected title instrument, a subordination agreement, an employer's written verification, a payoff statement, a gift donor's bank records, a trustee's signature — none of these are hard. All of them require a human being outside your building to do something, on their schedule, during their business hours, in an order you do not control. The task takes four minutes and the calendar takes four days, and the four days do not start until you ask.

On the Linden Street file, condition 7 — the mechanic's lien on Schedule B-II — is exactly this shape. The lien was found on day 19, the release was recorded, the update search found the release described the wrong lot, a corrected instrument was executed, and the title cleared on day 30. Eleven days, for a document nobody had to think hard about. It cleared because somebody started it on day 19 instead of day 27.

The rule, stated so you can use it

TRIAGE — the working rule                              [constructed teaching example]

  STEP 1   For every file, write the next date on which something becomes
           irreversible or irreversibly more expensive. Not the closing date.
           The NEXT DOOR THAT CLOSES.

  STEP 2   Sort ascending. Shortest fuse first.

  STEP 3   Break ties with: does the next action belong to somebody OUTSIDE
           your building? If yes, it moves up, because their clock has not
           started and yours has.

  STEP 4   Overlay the silence filter. Any file that is BOTH quiet and inside
           its irreversible window jumps the queue, regardless of loan size,
           borrower charm, or who called this morning.

  THE ANTI-RULE
           Being asked about a file is not evidence that the file needs work.
           It is evidence that somebody is watching it. Answer the question
           (that is 39.5's job, and it takes two minutes), then go back to the
           board.

Step 3 is worth defending because it looks like a technicality and is not. When the next action is yours, the elapsed time is under your control — you can do it at 4:45 tonight. When the next action belongs to an appraisal management company, a county recorder, an HOA management office, a payroll department, or a title examiner, the elapsed time is their turn time plus the delay before you asked. You cannot compress their turn time. You can only compress the delay before you asked, and that is the only lever you have, so pull it first.

Step 4 is the Linden Street lesson encoded as a rule.

📞 On the Phone

Two calls, both of them triage.

The first is the one nobody teaches you to make. It is Tuesday, and an agent you value has called three times about a file that is in underwriting and fine.

What not to say: nothing, and then answer at six o'clock, having felt guilty for six hours.

What works: "It's in underwriting, submitted Friday, and I expect a decision Wednesday or Thursday. Nothing is wrong with it. I'm going to be straight with you — I've got another file whose lock expires Friday and I'm spending today on that one. If anything changes on yours before Wednesday I'll call you inside the hour. Fair?"

That answer costs ninety seconds, gives the agent a real date, and ends the inbound. The version where you say nothing costs you six phone calls and the agent still does not know when to expect a decision.

The second is the Wednesday call. You have a condition that needs an HOA management company to sign a certificate.

The wrong version, Friday at 4:05: "Hi, I need the HOA cert for a closing next Thursday." The office closes at 4:30 on Friday, the person who signs is out Monday, and you have just spent your entire weekend of lead time buying nothing.

The right version, Wednesday at 9:10: "Good morning — I need a certificate for a closing a week from Thursday. What's your turnaround, what form do you use, who signs it, and is that person in this week? If there's a fee, tell me now and I'll get it to you today."

Four questions, ninety seconds, and every one of them is about their calendar rather than yours. That call does not clear the condition. It tells you, on Wednesday, whether the condition is going to clear — which is the only thing you can act on. Chapter 19 works the clearing itself.

What triage is not

Triage is not the same as prioritization by value, and it is worth saying so explicitly. A \$700,000 file and a \$180,000 file with the same lock expiration get worked in the order their doors close, not in the order of their loan amounts. This is partly professional obligation — the borrower on the small file signed the same thirty-year note and is entitled to the same competence — and partly self-interest, because a dropped small file generates exactly as much reputational damage as a dropped large one, in the same market, among the same agents.

It is also, in aggregate, a fair-lending consideration. A triage practice that systematically deprioritizes smaller loans, down-payment-assistance files, or government-program files will produce a service pattern with a demographic shape you did not intend and cannot defend. Chapter 25 covers this properly. The rule here is simple: your triage rule should be about dates, and only about dates, and you should be able to show it.

39.5 Milestone communication and the cadence that prevents calls

Everything so far has been about finding work. This section is about eliminating it, and it contains the highest-leverage arithmetic in the chapter.

Milestone communication is the practice of notifying the borrower and the referral partner at each defined stage of the file, proactively, before anyone asks. Touchpoint cadence is the schedule those notifications follow — including, crucially, a floor: a maximum interval after which a file gets an update even if nothing has happened.

The floor is the part people skip, and it is the part that matters.

Why the cadence is not a customer-service program

Framed as customer service, the cadence is a nice-to-have that gets dropped in a busy week — exactly when it is load-bearing. Frame it correctly and it is a capacity intervention, because unplanned inbound is the single largest consumer of a producing originator's day, and unplanned inbound is endogenous: it is generated by the absence of the very information the cadence supplies.

A borrower calls to ask "where are we?" for one reason: they do not know where they are, and they are thirty days into the largest financial commitment of their lives. An agent calls for one reason: they have a client asking them where they are and no answer. Neither call is a complaint. Both are requests for a status that costs you two minutes to produce and nine minutes to produce badly, because when the call arrives unplanned you must stop what you were doing, open the file, reload forty facts, answer, and then rebuild the context you dropped.

That gap — two minutes scheduled versus nine minutes interrupted — is where all the leverage is.

The cadence

MILESTONE CADENCE — every file, both audiences         [constructed teaching example]
  "Both audiences" = the borrowers AND the referring agent, same message,
  same hour. An agent who learns a milestone from their client instead of
  from you has learned something about you.

  MILESTONE                            SEND                  TYPICAL DAY
  ----------------------------------  --------------------  --------------
  Application taken, disclosures out   same day              d5
  Automated findings returned          same day              d6
  Appraisal and title ordered          same day              d7
  Rate locked (rate, term, EXPIRATION) same day              d12
  Appraisal received at value          same day              d16
  Title commitment received / issues   same day              d19
  Submitted to underwriting            same day              d23
  Conditional approval + condition list next business day    d28
  Each condition cleared (batched)     end of day            d29-d33
  Clear to close                       within the hour       d47
  Closing Disclosure issued            same day              d48
  Funded and recorded                  same day              d51

  ----------------------------------------------------------------------
  THE FLOOR:  no file goes more than FIVE BUSINESS DAYS without a written
              update, even when the update is "nothing has changed, here is
              what we are waiting for, here is who has it, here is when I
              expect it."
  ----------------------------------------------------------------------

Note what the milestone list produces on the Linden Street file: a dense run of updates through day 33, and then nothing — because there were no more milestones until day 47. The milestone list alone does not catch the dead window. The floor catches it.

Count it out. Day 33 was a Monday. Five business days later is Tuesday 34, Wednesday 35, Thursday 36, Friday 37, and then — because days 38 and 39 are a weekend — Monday, day 40. A floor update comes due on day 40, and to send it you must write a sentence explaining what the file is waiting for. There is no such sentence, because it is waiting for nothing. The act of trying to write the update is what surfaces the problem.

Day 40 is two days before the lock expires and one day before the borrowers walk into a furniture store. It is tight, and it works — barely. Which is exactly why the floor is not the only control: §39.3's weekly review, run on Friday day 37, finds the same file three days earlier and with room to move a closing date rather than merely to panic about one. Two controls, deliberately overlapping, because the cost of missing this file once is \$914.38 and a near-loss.

That is the design principle worth remembering: a cadence with only milestones fails silently in exactly the gaps where files die. The floor exists to make silence generate work — and the review exists to catch what the floor catches late.

Three rules about the messages themselves.

Write, don't call, for routine milestones. A written update is a record, it is asynchronous, it can be sent in batch, and it can be forwarded by the agent to their client without a game of telephone. Call for the two that carry emotion — the conditional approval, because the borrower will read "conditions" as "problems," and the clear to close, because they have been waiting for it for six weeks and they should hear a human voice.

Name the next date in every message. "We're waiting on the appraisal" is not an update. "The appraisal was ordered Wednesday, the appraiser has been assigned, inspection is Tuesday, and I expect the report back by the end of next week" is an update, because it tells the reader when to expect the next one and therefore when not to call.

Send the bad news faster than the good news. The appraisal came in low, the underwriter added a condition nobody expected, the title has an exception — those go out the same hour, with the plan attached. Theme two of this book is that a file is approved when it is documented, not when it is promised, and the corollary is that a problem is manageable when it is early and expensive when it is late. Chapter 18 makes this case on the Cypress Court file, where an appraisal \$35,000 under contract opened a \$28,000 gap eleven days before closing and the only thing that mattered was how fast everyone found out.

🧮 Run the Numbers

What the cadence buys, in hours.

(Constructed model. The rates below are assumptions, not industry statistics — this book does not publish a files-per-originator benchmark or an average contact rate, because none of them would be true for your market. Replace every input with your own measurement; the structure is the lesson.)

The setup. A thirty-file pipeline. Average file life seven weeks. A fifty-hour week, of which 40% — twenty hours — goes to acquisition: new inquiries, applications, agent meetings, the work in Chapters 7 and 38 that creates next quarter's pipeline. That leaves 30.0 hours a week for the files you already have.

Without a cadence. Assume each file generates two unplanned status contacts a week — one borrower, one agent. That is 30 × 2 = 60 contacts a week. Assume nine minutes each: four minutes of conversation, three to open the file and reload the facts, two to get back to whatever you were doing.

$$60 \times 9 = 540 \text{ minutes} = \mathbf{9.0 \text{ hours a week}}$$

File work left: 30.0 − 9.0 = 21.0 hours. Across thirty files:

$$21.0 \div 30 = 0.70 \text{ hours} = \mathbf{42 \text{ minutes per file per week}}$$

With a cadence. Assume ten scheduled touches across a file's seven-week life. Per file per week that is 10 ÷ 7 = 1.43 touches; across thirty files, 42.9 touches a week. At three minutes each — written, templated, sent while you work the board top to bottom, with no context reload because you are already in the file:

$$42.9 \times 3 = 128.6 \text{ minutes} = \mathbf{2.1 \text{ hours a week}}$$

The trade. Assume the cadence deflects two-thirds of unplanned inbound. That is conservative: a borrower who received a written update Tuesday morning rarely calls Tuesday afternoon.

	Hours/week
Available for existing files	30.0
Unplanned inbound, no cadence	−9.0
File work, no cadence	21.0
Available for existing files	30.0
Cost of running the cadence	−2.1
Residual inbound (one-third of 9.0)	−3.0
File work, with cadence	24.9
Net gain	+3.9

Per file: 24.9 ÷ 30 = 0.83 hours = 49.8 minutes, against 42.0 without. That is 7.8 more minutes per week on every file in the pipeline — about 19% more attention on all thirty, purchased with nothing but writing things down before people ask for them.

Across forty-eight working weeks, 3.9 × 48 = 187 hours, or roughly 4.7 forty-hour weeks of recovered capacity a year.

Run it at half these rates and it still pays. Halve the deflection to one-third and the net gain is 3.0 − 2.1 = 0.9 hours a week — still positive, still 43 hours a year, and that is the pessimistic case.

Two limits on that arithmetic, stated because this book does not sell you a technique without its failure modes.

The cadence does not deflect emotional inbound, and should not try. A borrower who calls because they are frightened about signing a thirty-year obligation is not asking for a status. They are asking for reassurance from a person, and answering with a templated milestone email is worse than not answering. Know which call you are getting. The Harlow Street borrower — a single income, a 641 score, ratios of 41.48% front and 51.00% back — called twice a week and nearly walked away three times, and no cadence in the world was going to change that, nor should it have.

A cadence that lies is worse than no cadence. "On track for the 24th" sent on day 37 of the Linden Street file would have been technically true and functionally a sedative — it would have told the agent to stop watching a file that was nine days from a lock expiration. If you are going to promise a date in writing every week, you have to look every week. The cadence and the review are one system.

39.6 Batching and the calendar of a producing loan officer

A producing loan officer's day contains two kinds of work that actively damage each other.

Acquisition is talking to people who are not yet files: returning inquiries, taking applications, structuring, meeting agents, following up on pre-approvals that have not found a house. It is unscheduled by nature, requires warmth and full presence, and it is the only activity on your calendar that determines whether you have a pipeline in ninety days.

Administration is moving the files you have: conditions, orders, updates, escalations, the board. It is scheduled by nature, requires precision rather than warmth, and it is the only activity that determines whether the pipeline you have becomes income.

Interleaving them is what most originators do, and it is expensive twice over. Every switch costs the reload — the same three minutes we priced in §39.5 — and, worse, it puts you in the wrong register. You cannot take an application well in the middle of chasing a payoff, and you cannot chase a payoff crisply while half your attention is still on a borrower who cried on the phone.

Batching is the practice of grouping like work into fixed blocks so that each block is entered once and left once. It is not a productivity fashion here. On a thirty-file pipeline with a 40/60 split between acquisition and administration, it is the only arrangement in which both fit inside a week.

A constructed week

THE PRODUCING DAY — a constructed template. Adapt the hours to your market;
keep the STRUCTURE.

  7:30 - 7:50   THE BOARD (20 min)
                Yesterday's movement in. Today's irreversible list out.
                This is the only time all day you look at all thirty files.

  7:50 - 9:00   THIRD-PARTY BLOCK
                Every outbound request that depends on somebody else's
                business hours: AMC, title, HOA, payroll/VOE, servicers for
                payoffs, county recorders, insurance agents.
                WHY FIRST: a request made at 8:00 gets a same-day answer.
                The identical request made at 4:00 gets a next-day answer.
                Every afternoon request costs one calendar day, permanently.

  9:00 - 11:30  LIVE BLOCK — acquisition
                New inquiries, pre-qualification calls, applications,
                structure conversations, lock decisions. Phone on. Door open.

  11:30 - 12:00 CONDITIONS BATCH 1
                Everything the borrowers owe. One pass. Batched requests, not
                a trickle -- five emails asking for one document each is how
                a borrower learns to stop opening your emails.

  12:00 - 1:30  Lunch. Tue/Thu: agent or partner meeting (Chapter 38).

  1:30 - 3:30   LIVE BLOCK — acquisition and structure

  3:30 - 4:00   CONDITIONS BATCH 2
                What arrived since 11:30. Submit it. Do not let documents
                sleep overnight in your inbox: a condition received at 3:00
                and submitted at 3:45 is cleared tomorrow; the same document
                submitted at 9:00 tomorrow is cleared the day after.

  4:00 - 4:45   THE CADENCE
                Milestone updates and floor updates. Written. Batched.
                Board open beside you, working top to bottom.

  4:45 - 5:00   TOMORROW'S IRREVERSIBLE LIST
                Write it tonight. Three or four lines. It is what lets you
                start at 7:30 tomorrow instead of at 9:00.

THE WEEK'S FIXED POINTS

  MON   Heaviest third-party block of the week. Their queues are longest and
        their week has not filled up yet. Anything needing a signature from
        outside your building gets asked for TODAY, not Thursday.

  TUE   Agent/partner meeting. Acquisition-weighted.

  WED   THE SIGNATURE DAY. Chapter 19's rule: anything requiring somebody
        else's signature is identified and requested by Wednesday, which
        leaves Thursday and Friday as recovery days inside the same week.

  THU   Agent/partner meeting. Forward pass on next week's closings.

  FRI   3:00 - 3:45  WEEKLY PIPELINE REVIEW (39.3), with the processor.
        3:45 - 4:30  Floor updates on every quiet file, so the weekend does
                     not start with somebody wondering.
        Friday afternoon is the WORST time to discover a problem and the
        BEST time to look for one.

The rule that makes it survivable

A batched calendar collapses the first time something genuinely cannot wait, and things genuinely cannot wait several times a week. So the rule is not "never interrupt the block." The rule is:

Batch everything except the irreversible.

Which is to say: the triage rule from §39.4 is also the interrupt rule. A lock decision when the market has moved, an agent calling about a contingency that expires today, a wire that has to go before the settlement agent's cutoff, an underwriter with a question that is holding a decision — those break the block, because the door is closing. A borrower asking how it is going, an agent asking for a status, a request for a pre-approval letter revision, a condition that arrived and can be submitted at 3:30 — those go in the queue.

Inbound is a queue, not an interrupt. Say it to yourself until it is automatic, and then tell your borrowers and agents what the queue's service level is, in writing, at application: "I answer everything the same business day. If it is genuinely time-sensitive, text me and say so." People are remarkably reasonable about a stated service level and remarkably unreasonable about an unstated one, and this is the whole reason the sentence works.

The honest limits

This calendar assumes you have a processor. If you do not — and many originators do not, especially early — the administration blocks roughly double and the acquisition blocks shrink accordingly, and your realistic capacity is a materially smaller number of files. That is not a failure. It is arithmetic, and pretending otherwise is how a fifteen-file originator ends up running a thirty-file board badly.

Your market sets the hours, not this page. If your referral partners work evenings and weekends, if you serve borrowers on night shifts, if your metro is split across time zones, the blocks move. The Linden Street borrower is a registered nurse; a template that only offers 9-to-5 application appointments is a template that does not serve half the households in your market. Move the blocks. Keep the batching.

Nothing here survives an unprotected calendar. Blocks that exist only in your intentions are not blocks. Put them in the calendar as appointments, with the same status as an appointment with a person, because that is what they are — appointments with thirty files.

39.7 The handoff to processing, done well

Files are dropped at boundaries. Not in the middle of a stage, where one person is clearly working, but at the seam between two people — where each of them has a defensible belief that the other one has it.

The handoff is the transfer of a file's day-to-day execution from the loan officer to the processor. Chapter 1 named the roles; this section is about the moment the file crosses between them, which is the single most common place a pipeline leaks.

What actually goes wrong

The failure is not laziness. It is ambiguity of ownership, and it takes two forms.

The gap. You submit the file to processing and mentally close the tab. The processor receives forty files a week and works them in the order they became actionable. Your file arrives incomplete in a way that is not obvious — a missing page of a bank statement, an unexplained address, an employment gap you knew about and did not flag — and it goes into a "waiting on LO" state that nobody is watching, because you think it is being processed and the processor thinks it is being completed. Four days evaporate. Nobody did anything wrong.

The overlap. You both work the same condition. The borrower gets two emails asking for the same paystub, decides one of you is disorganized, and starts screening. Worse, each of you now assumes the other has the outstanding item, and the next one falls through.

Both come from the same root: at any given moment, an ambiguous number of people own the file. The fix is a rule that sounds bureaucratic and is not:

At every moment, exactly one person owns the next action on a file. The handoff is a transfer, not a share.

And the standing exception that makes the rule workable: the loan officer never stops owning two things — the borrower relationship and the calendar. The processor owns the conditions, the orders, and the document flow. You own the dates and the humans. Those never transfer, which is why §39.3's read-through has the LO reading the calendar and the processor reading the conditions. The division is not ceremonial; it is the ownership rule made audible.

The handoff memo

Ten minutes of writing at submission buys days later. Write it once per file, attach it to the file, and send it to the processor.

HANDOFF TO PROCESSING — the memo                       [constructed teaching example]
Loan L-2214 · 4412 Linden Street, Ridgeview · Conventional 30-yr fixed

1. THE DEAL IN THREE LINES
   $385,000 purchase, 5% down, loan $365,750, LTV 95%. Two borrowers,
   married, first-time buyers, primary residence. AUS: Approve/Eligible.

2. THE CALENDAR — the part I own and you need
   Contract executed        day 4
   Contract closing date    day 45
   Rate locked              day 12, 6.625% + 0.500 point
   LOCK EXPIRES             day 42     <-- three days SHORT of the contract
                                           date. Flagged at lock. We close
                                           by day 40 or we pay for it.
   Appraisal ordered        day 7
   Title ordered            day 7

3. WHAT I HAVE ALREADY PROMISED, AND TO WHOM
   Borrowers: weekly written update; a call the day the approval comes back.
   Agent: same updates, same hour. She has closed four files with us.
   Nobody has been promised a rate change, a fee waiver, or a closing date
   earlier than the contract date.

4. THE SOFT SPOTS — what I would flag before the underwriter finds it
   a. B2's commission is QUARTERLY and lumpy. Expect a large-deposit
      condition on the $4,900. It is the net of a $6,900 gross commission
      after $2,000 withholding. It is ALREADY inside the 24-month average --
      it is an asset-sourcing item, NOT additional income. Do not let anyone
      count it twice.
   b. $10,000 of the $38,000 in assets is a gift from B1's parents. Gift
      letter plus evidence of transfer, both sides.
   c. Back-end is 42.66%. There is no room. Any new debt breaks this file.
   d. B2's second auto has 19 payments left -- the ten-month rule does NOT
      reach it. Do not omit the $429.00.

5. HOW TO REACH THE BORROWERS
   B1 works twelve-hour hospital shifts; text first, call after 7 p.m.
   B2 is in the field weekdays; email lands, phone does not.

6. THE HANDOFF ITSELF
   As of today the conditions and orders are YOURS. The calendar and the
   borrowers are MINE. If a condition needs somebody outside this building
   to sign something, tell me the same day -- that one comes back to me.

Section 4 is the section that earns the memo. You know things about this file that the underwriter will discover in four days and ask about in six. Writing them down converts a future condition into a present disclosure and frequently prevents the condition entirely. It also protects you: a soft spot you flagged in writing on day 23 is diligence, and the same soft spot discovered by an underwriter on day 27 is a surprise.

Section 2 earns it a second time, and it is worth being blunt about what that entry says. The lock on this file was three days shorter than the contract's closing date on the day it was taken. Chapter 30 makes that critique properly. The handoff memo is where a competent originator writes it down so that at least one other person in the building knows the file has a structural deadline problem before it becomes a \$914.38 problem on day 42.

⚠️ Where Deals Die

"I thought you had it."

The condition arrives from underwriting on a Friday: a corrected instrument for the mechanic's lien. You see it and think the processor is on the title work. The processor sees it and thinks that came in addressed to the LO, and it needs the title company, which the LO has the relationship with. It is nobody's, and being nobody's it does not appear on either person's list — an item that exists in two inboxes and zero task lists.

Four days later the underwriter asks for a status and you both answer at once.

The tell: any condition that arrives with two names on it. Third-party curative items are the most common, because they are simultaneously a document task (processor) and a relationship task (loan officer).

The discipline: the acknowledgment reply names the owner and the date, in writing, every time. "I've got 7 — calling the title company this morning, expect an update by Tuesday." Nine words. It converts an ambiguous item into an owned one, and it timestamps the ownership so that the weekly review can find it if Tuesday comes and goes.

Handoffs you did not schedule

The submission handoff is the one you plan for. Several others happen without an announcement and deserve the same discipline:

• Processor to underwriter (submission) — you do not control it, but you can control whether the file arrives complete, because an incomplete submission returns as conditions and costs a full turn time.
• Underwriter back to you (the stip sheet) — Chapter 19's territory, and the moment where an unassigned condition list becomes a queue with no owner.
• You to the closer (clear to close) — the highest-stakes seam in the file, because the CD timing starts here and a day lost is a day the whole closing moves.
• Coverage during absence — vacations, illness, licensing exams. A pipeline handed to a colleague without the memo above is a pipeline handed to nobody. If you are out for a week, every file needs its calendar written down, because the thing your colleague cannot reconstruct is the thing you have been holding in your head.
• You to the servicer (after funding) — the borrower's last handoff, and the one they are least prepared for. One sentence at closing about the transfer of servicing prevents a frightened phone call in February. Chapter 23 covers the notices.

39.8 Escalation paths

An escalation path is the predetermined sequence of people you contact, in order, when a file stops moving — with a trigger date attached to each step, decided in advance.

Two things about that definition do the work. Predetermined, so you are not inventing a response under pressure. With a trigger date attached, because the most common escalation failure is not escalating to the wrong person; it is escalating three days too late, having spent those three days hoping.

Chapter 19 covers escalating a condition — what to do when a specific stip will not clear. This section is about escalating a file, which is a different problem: something is not moving, possibly several somethings, and the question is who can make it move.

The ladder

THE ESCALATION LADDER                                  [constructed teaching example]

  LEVEL 0   THE ASSIGNED PERSON, IN WRITING
            A specific ask, a specific date, and a specific consequence.
            Not "checking in." "I need the VOE by Thursday to submit
            Friday; the lock expires the 15th."
            SET THE TRIGGER NOW: if no response by <date>, go to Level 1.

  LEVEL 1   THE SAME PERSON, BY PHONE
            The writing created the record. The call creates the movement.
            Say the date out loud. Ask one question: "Can you do it by
            Thursday -- yes or no?" A no on Wednesday is worth more than a
            maybe on Friday, because a no lets you go to Level 2 today.

  LEVEL 2   THEIR SUPERVISOR — OR YOUR OWN
            Processing team lead, underwriting manager, AMC escalation
            desk, title branch manager. Bring ARITHMETIC, not adjectives:
            "Lock expires Wednesday. An extension is 0.250 point --
            $914.38 on this loan -- and somebody in this building pays it."

  LEVEL 3   YOUR BRANCH OR OPERATIONS MANAGER
            You are now asking your own company to spend authority. Use it
            for files, not for feelings, and expect to be asked what you
            did at Levels 0 through 2. Have the dates.

  LEVEL 4   THE STRUCTURAL LEVER
            - Broker/correspondent: your account executive at the wholesale
              lender (Chapter 31). Their job is exactly this.
            - Lock and pricing: the secondary desk, through your manager.
            - Third-party vendor: the ordering party, not the vendor.
              You often cannot escalate at an AMC; the lender who assigned
              it can.

  LEVEL X   THE BORROWER — a sideways move, not an upward one
            Some doors only open from the borrower's side. An employer's HR
            department answers its own employee faster than it answers a
            vendor. A servicer talks to its own customer. An HOA responds to
            an owner. When you are stuck outside a wall, check whether your
            borrower is already inside it.

Level X is not a last resort; it is frequently the fastest move available and it is systematically underused. The reason it is underused is that it feels like admitting you cannot do your job. It is not. It is recognizing that some organizations are structurally more responsive to their own customer than to a third party, and routing the request accordingly. Ask precisely, give the borrower the exact wording, and take the task back the moment the door is open.

Five rules

Escalate on a date, not on a feeling. Set the trigger when you make the request, not when you get annoyed. Write it on the board in the same motion. This single habit converts escalation from an emotional act into a scheduled one, and scheduled escalations happen on time.

Never escalate something you have not completed on your side. If the condition is unclear, un-indexed, or missing a page you were supposed to supply, escalating makes you the problem, permanently, with a person you will need again next week.

Escalate early and small rather than late and loud. The first escalation should be boring — a short note, a named date, no adjectives. Loud escalations work once. Boring escalations work indefinitely, which matters because you will submit files to the same underwriting department for years.

Bring the arithmetic. "This is urgent" is an opinion. "The lock expires Wednesday, an extension is 0.250 point — \$914.38 — and the contract's financing contingency lapses Friday" is a set of facts with a cost attached, and it converts your request from a preference into a decision somebody has to make.

Spend the relationship budget deliberately. You have finite standing with a processing manager, an underwriting manager, and an account executive. Every escalation spends some of it. The originator who escalates everything has no lever left on the file that actually needs one — and everyone in operations knows exactly who that originator is.

The escalation the Linden Street file needed

It is worth noticing, at the end of a section on escalation, that the Linden Street file did not need an escalation at any point in its eleven dead days. Nobody was slow. No third party was unresponsive. No underwriter sat on anything. There was nothing to escalate to, because there was nothing stuck.

That is precisely why the file is dangerous, and why an escalation habit alone will not save a pipeline. Escalation is the tool for a file that is blocked. The Linden Street file was not blocked. It was finished and parked — and no escalation ladder in the world has a rung for that. Only the board has a column for it, which is §39.10.

39.9 Pull-through and fallout

Two numbers connect your desk to the rest of the company, and most originators can define neither.

Pull-through rate is the share of loans that reach funding out of some earlier population — most usefully, out of locked loans. Fallout rate is its complement: the share that never funds.

$$\text{pull-through} = \frac{\text{loans funded}}{\text{loans locked}} \qquad \text{fallout} = 1 - \text{pull-through}$$

Insist on the denominator before you discuss the number. Some shops measure funded over applications, some over locks, some over "submissions," and the same pipeline produces wildly different percentages under the three. A pull-through figure without a stated denominator is not a statistic; it is a mood. When your manager quotes you one, the first question is out of what?

This book is not going to give you an industry pull-through benchmark, and you should be suspicious of anyone who does casually. The figure moves with the rate cycle, with the purchase/refinance mix, with lock policy, and with how a given company defines its denominator. Measure your own, over at least a couple of quarters, and compare it to your own past, not to a number in a book.

Why the lender cares: fallout costs money on a loan that never existed

This is the part that is genuinely counterintuitive, and Chapter 29 built the machinery for it.

When a lender locks a borrower's rate, it has made a promise: that on a future date it will deliver a loan at an agreed rate and price. To protect itself against the market moving in between, the lender takes an offsetting position in the secondary market — in effect, forward-selling loans it does not yet have. That is what hedging a pipeline means, and the size of the hedge depends on an estimate of how much of the locked pipeline will actually fund. Which is to say: the hedge is sized on pull-through.

Now watch what happens when the estimate is wrong.

WHY FALLOUT IS EXPENSIVE                               [constructed illustration only;
                                                        round numbers, no real pricing]

  The lender has $100,000,000 of locked loans.
  It expects 80% pull-through, so it forward-sells about $80,000,000.

  CASE A -- rates RISE.
    Borrowers hurry. Pull-through comes in ABOVE expectation, say 90%
    ($90,000,000 funds). The lender has more loans than it sold forward.
    It sells the extra $10,000,000 into a market where prices have FALLEN.
    Unpleasant, and roughly the risk it signed up for.

  CASE B -- rates FALL.
    Borrowers with a lock at yesterday's rate discover today's is better.
    Some renegotiate. Some walk to another lender. Pull-through comes in
    BELOW expectation, say 65% ($65,000,000 funds).
    The lender has forward-sold $80,000,000 and can deliver $65,000,000.
    It must BUY BACK $15,000,000 of forward sales -- in a market where
    prices have RISEN, because prices rise when rates fall.

  The asymmetry is the point: fallout is worst exactly when covering it is
  most expensive. The lender loses money on loans that never existed.

This is why lock policy exists, why renegotiation policies exist, why some lenders price extensions the way they do, and why your secondary desk is interested in your personal fallout pattern in a way that can feel intrusive and is not. Chapter 30 covers what this means for how you advise a borrower about locking.

The originator's version is simpler and hits closer to home: you are paid on funded volume, and every hour spent on a file that never funds is unpaid.

🧮 Run the Numbers

What a ten-point improvement in pull-through is worth.

(Constructed model. These are not industry figures — substitute your own.)

Today. You take 20 applications a month; 16 fund.

$$\text{pull-through} = \frac{16}{20} = \mathbf{80.0\%} \qquad \text{fallout} = \mathbf{20.0\%} \text{ (4 files)}$$

What the fallout costs you in time. Assume each dead file consumed six hours before it died — the application, the disclosures, the credit pull, the structure conversation, the orders, the follow-up, the eventual withdrawal.

$$4 \times 6 = 24 \text{ hours a month} \qquad 24 \times 12 = 288 \text{ hours a year}$$

At eight hours a day, 288 ÷ 8 = 36 full working days a year on loans that never existed. Note that this is not waste in a moral sense — some of those borrowers will come back, and taking marginal applications is part of the job. It is simply capacity, and it is invisible unless you count it.

What improvement is worth. Suppose the board, the cadence, and the handoff move you from 80% to 90% — 18 of 20 fund instead of 16.

$$+2 \text{ funded loans per month} = +24 \text{ per year}$$

At the Linden Street loan amount of \$365,750, that is 24 × \$365,750 = \$8,778,000 of additional funded volume a year — with zero additional leads, zero additional marketing spend, and zero additional applications taken. Whatever your compensation structure (Chapter 26), it applies to that volume the same as to any other.

The point is not the percentage. It is where the volume came from. Chapter 38 is about generating more applications, which is hard, competitive, and expensive. This chapter is about funding more of the applications you already have, which is free. Most originators spend all of their improvement effort on the first and none on the second.

The three kinds of fallout, and what you can actually do

Not all fallout is the same, and treating it as one number hides the only useful information in it. Sort your dead files into three buckets:

Credit fallout — the file could not be approved. Income that did not verify, a ratio that would not come down, an appraisal that came in short with no way to bridge the gap, title that could not be cleared. Partly preventable, and the prevention happens at the front of the file, not the back: a rigorous pre-qualification conversation (Chapter 8), an early honest read on value (Chapter 18), and verification before commitment (theme two). The Cypress Court file is the canonical case — a \$35,000 short appraisal opening a \$28,000 gap — and whether that file lives depends almost entirely on how early everyone found out.

Competitive fallout — the borrower left for a better price. The Linden Street borrowers were shopping an online lender and had a lower quoted rate in hand, which is completely ordinary and is the market working. The response is not to match every quote; it is to have been honest and useful from the first call, and to have been present — a borrower who has heard from you every week is a borrower who tells you they got another quote instead of simply vanishing. Chapter 40 works that comparison in full.

Transaction fallout — the deal died and the loan was fine. The inspection found something. The seller took a better offer. The borrower's own house did not sell. Their employment changed. This is largely outside your control, and the useful discipline is to label it correctly so that it does not contaminate your read on the other two buckets.

Then there is the fourth cause, the one this chapter exists for: the file that fell out because nobody was watching it. Not declined, not outbid, not sold — expired. It never appears as its own line in anyone's reporting, because it gets recorded under whichever of the three buckets it eventually landed in. A lock that expires and re-locks at a worse market becomes "competitive fallout" when the borrower leaves. A contingency that lapses and costs the borrower their leverage becomes "transaction fallout." The pipeline-management failure is systematically miscategorized as something else, which is exactly why it persists.

And notice the sharpest version of the point: the Linden Street file funded. It is in the numerator. It cost \$914.38, it blew past its contract date by six days, it nearly died on day 44 — and in every report anyone runs, it is a success. Pull-through is a lagging metric that says nothing at all about a file that nearly died and didn't. Your board is the only instrument that would have seen it.

⚖️ Compliance Check

A dead file is not a file you may simply forget.

Pipeline hygiene is not only a business practice. Once an application exists, obligations attach to it, and they do not disappear because the loan is not going to close.

• Notification of action taken. Under the Equal Credit Opportunity Act and Regulation B, a creditor must notify an applicant of action taken on an application within a defined period — 30 days after receiving a completed application — and an adverse action notice must state specific principal reasons. Where an application is incomplete, Regulation B provides for a notice of incompleteness with its own requirements and timing. A borrower who "went quiet" is not, by that fact alone, a borrower who withdrew, and treating an abandoned file as self-cancelling is how a shop accumulates missing notices.
• Record retention. Regulation B imposes a retention requirement on applications and related records (25 months for consumer credit at this writing). Other rules impose their own. Verify the current text and follow your company's retention schedule.
• HMDA coding. Under the Home Mortgage Disclosure Act and Regulation C, applications that are withdrawn, closed for incompleteness, approved-but-not-accepted, or denied are all reportable with distinct action-taken codes. Coding a dead file wrongly — particularly recording a denial as a withdrawal — corrupts data that examiners and the public use to evaluate lending patterns. Chapter 25 covers why that matters.
• The pattern question. If files with certain characteristics disproportionately go quiet and die on your desk, that is a fair-lending exposure regardless of intent (Chapter 25). This is a genuine reason to triage on dates only and to be able to demonstrate it.

None of this is a reason to avoid taking applications from borrowers who might not qualify. You never manage fallout by declining to take applications — that is both bad business and, done by pattern, a serious fair-lending problem. You manage it by working the files you took.

Requirements change and state law varies. Verify current requirements with your compliance department and your regulator.

39.10 Building the board

Everything in this chapter converges on one artifact. Build it, and most of the rest becomes mechanical.

A pipeline board is a single view of every open file with the fields that let you find the one that is dying. It can live in your loan origination system, in a spreadsheet, or on a whiteboard — the medium is genuinely irrelevant and arguing about it is a way of not building one. What is not irrelevant is which columns are on it, because the wrong columns produce a board that looks impressive and finds nothing.

The columns that matter, and the ones that don't

Most originators, asked to build a pipeline report, produce: borrower name, loan amount, rate, closing date, status. That report answers "what is my month going to look like," which is a legitimate question and a completely different one. It cannot find a dying file, because every column on it is a fact about the loan and none of them is a fact about the file's motion.

The board needs six fields, and four of them are about motion:

That last point is the chapter's payoff and it deserves to be stated on its own. A file with no blocking party is not fine. It is unowned. Every other file on your board is waiting on somebody identifiable whom you can call. The file waiting on nobody is waiting on you to make a decision you have not noticed you need to make.

The two filters

You do not read thirty rows every morning. You run two filters and read the intersection.

THE TWO FILTERS

  FILTER 1 -- SILENCE      days quiet >= 3 business days
  FILTER 2 -- IRREVERSIBLE next irreversible date <= 5 days

  Each filter alone is noisy:
    - Silence alone catches every pre-approved borrower still house-hunting,
      which is most of the top of any board and almost never a problem.
    - Irreversibility alone catches today's closing and tomorrow's wire,
      which are already on your list because they are shouting.

  THE INTERSECTION is the working set: files where something is closing in
  and NOTHING IS HAPPENING. That is the definition of a file about to die.

  Within the intersection, sort by BLOCKING PARTY:
    - a named party  -> you have somebody to push. Push them.
    - EMPTY          -> nobody is coming. This one is yours, and it is first.

📄 Read the File

```text FIGURE 39.1 — "Thirty files on a Friday, and one of them is dying" [constructed pipeline board; row 21 is the Linden Street file] THE DOCUMENT A loan officer's pipeline board, printed at 7:30 a.m. on day 37, a Friday, immediately before the weekly review. Six columns. DIS = calendar days in current stage. Q = BUSINESS days quiet. Dates are expressed as d+N, days from today. THE CONTEXT Thirty open files. Two close within the week. One has a lock expiring in five days. Row 21 is the Linden Street file, whose last event was the ninth condition clearing on day 33 -- Monday -- which is four business days ago. THIS REVIEW DID NOT HAPPEN. The board below is what it would have shown if it had. WHAT IT SHOWS

# FILE STAGE DIS Q NEXT IRREVERSIBLE BLOCKING

---

1 Ashcroft Rd PREAPP 31 12 credit report stale d+9 borrower 2 Bellhaven Ct PREAPP 9 2 -- borrower 3 Cranmer Way PREAPP 44 26 credit report STALE d-2 -- 4 Dunmore Pl PREAPP 5 1 -- borrower 5 Eastgate Row PREAPP 18 18 credit report stale d+22 -- 6 Fenwick Ln PREAPP 2 0 -- borrower 7 Gladstone Ave PREAPP 22 3 credit report stale d+15 borrower 8 Halloway Dr PREAPP 12 12 credit report stale d+28 -- 9 Ivyridge Ct APP 3 0 LE delivered d-3, ok LO 10 Jessup St APP 1 0 intent to proceed d+6 borrower 11 Kestrel Ln APP 6 4 INSPECTION CONTINGENCY d+2 borrower 12 Larkspur Ave SETUP 8 5 APPRAISAL CONTINGENCY d+4 AMC 13 Marlow Ct SETUP 4 1 lock expires d+19 title 14 Norwood Pl SETUP 11 9 FINANCING CONTINGENCY d+3 AMC 15 Orchard Row SETUP 6 2 lock expires d+14 employer 16 Pemberton Way SETUP 2 0 lock expires d+26 -- 17 Quarry Ridge Rd SUBMIT 3 3 closing d+16 underwriting 18 Ravenswood Ct SUBMIT 6 6 LOCK EXPIRES d+4 underwriting 19 Sedgwick Ln SUBMIT 1 1 closing d+24 underwriting 20 Trenton Ave SUBMIT 2 2 financing contingency d+6 underwriting 21 LINDEN ST COND 9 4 LOCK EXPIRES d+5 (day 42) -- NOBODY 22 Underwood Pl COND 2 0 closing d+12 borrower 23 Vandermeer Ct COND 7 1 HOA cert signature; d+8 HOA mgr 24 Westbrook Rd COND 4 2 paystubs stale d+8 borrower 25 Yardley Ave COND 12 6 PAYOFF EXPIRES d+1 servicer 26 Zeller Ct COND 3 1 closing d+9 borrower 27 Ambrose Ln CTC 2 0 CD MUST ISSUE TODAY (d+4) LO 28 Briarcliff Pl CTC 1 0 closing d+7 closer 29 Colgrove St CLOSE 1 0 WIRE BY 10 A.M. d+1 borrower 30 Danforth Rd CLOSE 0 0 CLOSING TODAY 2 P.M. closing agt

STAGE COUNTS PREAPP 8 · APP 3 · SETUP 5 · SUBMIT 4 · COND 6 · CTC 2 · CLOSE 2 = 30 files. The board foots.

FILTER 1 quiet >= 3 business days ...... 1, 3, 5, 7, 8, 11, 12, 14, 17, 18, 21, 25 (12 files) FILTER 2 irreversible <= 5 days ........ 11, 12, 14, 18, 21, 25, 27, 29, 30 (9 files) INTERSECTION ........................... 11, 12, 14, 18, 21, 25 (6 files) OF THOSE, BLOCKING PARTY EMPTY ......... 21 (1 file)

WHAT IT DOESN'T It does not show loan amounts, rates, borrower names, or projected commission, and it should not -- those answer a different question and they bias triage toward big files. It does not show WHY a file is quiet: rows 5 and 8 have been silent for weeks and are almost certainly pre-approved buyers who stopped looking, while row 14 has been silent for nine days with a financing contingency three days out, which is an emergency. The board finds them; only a human sorts them. And it does not show row 21's real problem, because there is no column for "this file is finished and nobody scheduled it." The empty BLOCKING cell is the closest the board can get, and it turns out to be enough. THE DECISION Sorted by pure urgency, LINDEN ST is NINTH. Eight files have something irreversible closer than its lock: a closing at 2 p.m. today, a wire due tomorrow morning, a payoff expiring tomorrow, an inspection contingency in two days. Every one of those eight will get worked today WITHOUT the board, because every one of them is generating email. Row 21 generates nothing. Nothing is wrong with it. It is complete, and it is the only file in the intersection with no blocking party at all. TODAY: do the two zero-day items (27, 30) -- they are short and they are today. Make the four calls that push rows 25, 29, 11, and 14, which take two minutes each because somebody is there to push. Then give the next HOUR to row 21: call the settlement agent and the listing side, move the closing inside the lock, and issue the CD Monday. THE LESSON Urgency ranks this file ninth. Irreversibility plus silence ranks it first. The board exists to disagree with your inbox, and on the one day it matters, it will. ```

Constructed. Row 21 uses the frozen Linden Street facts; the other twenty-nine files are invented to populate a realistic board.

Building yours on Monday

Six columns, thirty rows, one filter pair. Concretely:

1. List every open file, including pre-approvals with no property. If it has an application or a letter with your name on it, it is on the board. Files you leave off the board are files you have decided not to manage.
2. Add the four motion columns. Stage and days-in-stage usually come out of the LOS. Days quiet does not, and you will probably have to maintain it by hand at first — which is fine, and the act of updating it is itself the review. Chapter 36 covers what to ask your LOS administrator for; "last activity date, in business days, on the pipeline view" is the single highest-value request you will ever make of them.
3. Fill in the next irreversible date by hand, file by file, once. This takes about an hour the first time and about five minutes a week afterward. It is the most valuable hour in this chapter, because the act of finding each file's next closing door is the audit.
4. Fill in the blocking party — and force yourself to leave it empty when nobody owes anything, rather than writing "me" or "waiting." An honest empty cell is the entire alarm system.
5. Run the two filters at the weekly review, and every morning at 7:30 in twenty seconds.
6. Sort the intersection by blocking party, empty cells first.

Then the discipline that makes the board real: when a file changes, the board changes the same day. A board that is accurate on Friday and fictional by Tuesday is worse than no board, because it will tell you a file is fine on the morning it is not.

One honest limitation. The board is a detection instrument, not a decision-maker. It found row 21 by noticing an unusual combination of two numbers and an empty cell. It cannot tell you that rows 5 and 8 are harmless while row 14 is an emergency — those files look identical in the "quiet" column and are nothing alike. The board's job is to guarantee that a human being looks at the right dozen files. Judgment is still yours, and it always will be.

🗂️ The Loan File

Chapter 39 contribution: this file plus twenty-nine others — build the board, and find the one about to die.

Until now, every checkpoint has looked into the Linden Street file. This one looks at it from outside, as one row among thirty, because that is how it actually existed: not as the subject of your attention, but as line 21 on a board you did not have.

Add to the file:

1. The velocity decomposition. Seven stages, fifty-one days:

2. The conditions block, opened. Five days clearing nine conditions in three business days (day 29, day 30, day 33 — days 31 and 32 were a weekend). Eleven days of nothing. Three days of crisis and cure.

3. The counterfactual. Documentation-complete day 33; clear to close available day 34; Closing Disclosure received Tuesday day 34; three business days — Wednesday 35, Thursday 36 — closing available Friday day 37. Fourteen days earlier than what happened, five days inside the lock, eight days before the contract date, four days before a furniture store.

4. The row. As it would have read on day 37:

   #  FILE          STAGE   DIS   Q  NEXT IRREVERSIBLE           BLOCKING
   21 LINDEN ST     COND      9   4  LOCK EXPIRES d+5 (day 42)   -- NOBODY

What this settles. The file's calendar is now measurable rather than anecdotal, and the two failures — a \$914.38** lock extension on day 42 and a **\$5,200 furniture purchase on day 41 that took the back-end ratio from 42.66% to 48.48% — resolve into one cause. Slack was never converted into an earlier closing date. Both events happened inside the same eleven-day window, and a single decision on day 33 would have prevented both.

What it does not settle. Nothing about whether this was a good loan. The structure was sound, the borrowers were qualified, the value supported, the title cleared, the ratios worked. A well-underwritten file and a well-run file are different achievements, and this one was only the first. It also does not settle what the loan officer earned on it, what it cost the borrowers against the alternatives they were shopping, or what the file says about the year ahead — those are Chapter 40's.

Open questions carried forward:

• Q1. What did the eleven days actually cost, once you count the extension, the near-loss, six days past the contract date, and the borrowers' experience of it? (Chapter 40)
• Q2. The borrowers had a lower quote in hand from an online lender the whole time. Did the way this file was run make it more or less likely they stayed? (Chapter 40)
• Q3. If you had run this pipeline properly all year, how many more files would you have funded — and is that a bigger number than the one you would get from more marketing? (Chapters 38 and 40)

Your task. In Appendix C's workbook, build the board. List every file you are carrying — or, if you are not yet producing, construct twenty-nine plausible ones around the Linden Street row. Fill in all six columns. Run both filters. Then write one sentence naming the file that is about to die and one sentence saying what you will do about it, with a day and an hour attached.

If the sentence does not have a day and an hour in it, you have not triaged. You have worried.

Conclusion

The failure mode of a growing pipeline is not difficulty. It is volume of easy things, and the file that dies is the one nobody was worried about, because worry is a form of monitoring and the quiet file has none.

The Linden Street file proves it in its own numbers. Fifty-one days against a contract that named day 45 and, once executed, allowed only 41. Underwriting took five of those days. The appraisal was never the problem. Third parties took twelve days and title was the binding constraint the whole time, while everyone chased the appraiser. Conditions took nineteen days — thirty-seven percent of the file — and nine of the eleven conditions cleared in three business days, which was excellent work.

And then eleven days of nothing, inside which a lock expired for \$914.38 and two borrowers financed furniture that took their back-end ratio from 42.66% to 48.48% and nearly ended the transaction. A closing was available on day 37. Nobody took it.

The machinery that prevents that is not effort and it is not attentiveness, because neither of those scales past a dozen files. It is structure: a board with a column for silence and a column for the next irreversible date; a weekly review that reads every file whether or not you are busy; a triage rule that ranks by closing doors instead of by volume; a cadence with a floor, so that silence is forced to generate work; a batched calendar in which acquisition and administration stop damaging each other; a handoff that leaves exactly one owner; and an escalation path with dates attached in advance.

What that structure buys, in the constructed models of this chapter, is roughly four hours a week and about nineteen percent more attention on every file — and if it moves your pull-through even modestly, it produces funded volume you did not have to go find. Chapter 38 is about getting more applications, which is hard. This chapter is about funding more of the ones you already have, which is free, and which almost nobody works on.

One last thing, and it is the reason the chapter was hard on a loan that funded. Pull-through would have recorded the Linden Street file as a success. It closed. It is in the numerator. Every metric anyone runs says that file was fine, and it was nine days and one furniture store away from being nothing at all. The board is the only instrument in this chapter that would have seen it — which is why you build the board, and why you look at it before your inbox.

Next: Chapter 40 closes the book and the file. It assembles the complete Linden Street loan from first call to funding, prices what the borrowers actually got against what they were being offered elsewhere, and turns the whole thing into the first year of a career — production goals, the business plan, and the question of what kind of originator you intend to be.

Key Terms

Pipeline management — the discipline of running many loan files simultaneously at a defined service level: detecting which files are stalled, deciding which gets worked next, and ensuring none is dropped. Distinct from the origination process itself (Chapter 6), which describes one file. (Ch.39)

File velocity — the rate at which a file moves through the stages of origination, measured in days per stage rather than days in total, so that a slow file's slowness can be located rather than merely observed. (Ch.39)

Weekly pipeline review — a fixed, recurring appointment at which every open file is examined in sequence, the board is reconciled against the loan origination system, and a specific next action with an owner and a date is set for each flagged file. (Ch.39)

Triage — deciding the order in which files are worked, using time-to-irreversibility rather than expressed urgency as the ranking key. (Ch.39)

Escalation path — the predetermined sequence of people contacted, in order, when a file stops moving, with a trigger date attached to each step at the moment the original request is made. (Ch.39)

Milestone communication — proactive notification of the borrower and referral partner at each defined stage of a file, sent before either of them asks. (Ch.39)

Touchpoint cadence — the schedule that milestone communication follows, including a floor: a maximum interval after which a file receives an update even when nothing has happened. (Ch.39)

Batching — grouping like work into fixed calendar blocks so each type of work is entered and left once, rather than interleaving acquisition and administration and paying the context-switch cost repeatedly. (Ch.39)

Capacity — the number of files an originator can carry at a defined service level, as opposed to the number that can technically be open. Exceeded gradually and detected late, because the symptom is a feedback loop rather than a threshold. (Ch.39)

Per-file time budget — available working hours in a period divided by files in the pipeline; the honest measure of how much attention each file can actually receive, and the number unplanned inbound consumes. (Ch.39)

Handoff — the transfer of a file's day-to-day execution from one party to another, most commonly loan officer to processor. Sound handoffs leave exactly one owner of the next action; the loan officer retains the calendar and the borrower relationship in all cases. (Ch.39)

Pull-through rate — the share of loans that reach funding out of a stated earlier population, most usefully out of locked loans. Meaningless without its denominator stated. (Ch.39)

Fallout rate — the complement of pull-through: the share of locked loans that never fund. Costs the lender money because the pipeline hedge was sized on an expected pull-through, and covering a shortfall is most expensive precisely when fallout is highest. (Ch.39)

Spaced Review

1. (This chapter.) Your board shows two files that have been silent for nine business days. One is a pre-approved borrower with no contract; the other is in setup with a financing contingency expiring in three days. Both look identical in the "days quiet" column. Explain what the board did for you here and what it could not do, and name the column that separates the two files.

2. (Chapter 6 + this chapter.) Chapter 6 corrected "45-day contract" to the 41 days that actually remained after the contract executed on day 4. Using the velocity decomposition in §39.2, identify the two stages you would attack first to fit this file inside 41 days, and state how many days each is worth. Then say why attacking underwriting turn time would have been a waste of effort.

3. (Chapter 19 + this chapter.) Chapter 19's rule is that a condition needing somebody else's signature must be identified on a Wednesday, not at four o'clock on a Friday. Restate that rule in the language of §39.4's triage hierarchy, and explain why it is a rule about lead time rather than about difficulty. Use the mechanic's lien — found day 19, cleared day 30 — as your example.

4. (Chapters 19 and 30 + this chapter.) All nine prior-to-doc conditions on the Linden Street file cleared by day 33; the lock expired day 42; the contract named day 45; the file closed day 51. Compute the number of days of unconverted slack that existed on day 33 measured against the lock, and against the contract date. Then state, in one sentence, the decision that should have been made on day 33 and who should have made it.

5. (This chapter.) A colleague tells you their pull-through is 92% and yours is 79%, and concludes that you are worse at this. Ask the one question that has to be answered before the comparison means anything, then give two ways a higher pull-through number could indicate a worse business.