Mortgage Pipeline Stages: Where Files Actually Stall

The file was submitted on the 6th. You are looking at the board on the 14th and it still reads In Processing.

Nobody lied to you. The LO submitted on the 6th and mentally released it. The processor opened it on the 10th — nothing forced an earlier look — found a bank statement missing page three, an unsourced deposit, and no verification of employment ordered, and sent a needs list back that hour. The LO read it between appointments and worked it on the 13th.

Eight days. Maybe ninety minutes of actual work inside them. Every one of those days rendered as In Processing on a report you looked at twice.

Here is what no board shows you: for six of those eight days, that file was not on anybody's list. It had a stage. It did not have an owner.

A stage is a place. A queue is a line.

The stage labels are fine. Application, processing, underwriting, conditional approval, clear to close, docs out, funded. Every branch runs some version, and the labels are rarely the problem.

The problem is that a stage describes a location, and work happens in queues.

Every boundary between two stages is an entry into somebody else's queue, and queues have three properties a stage label cannot show. Depth — how many files are stacked ahead of yours. Position — where yours sits right now. And a re-entry penalty — what it costs to get back in line after you get bounced.

That third one is the mechanism most pipeline conversations miss. Your lender posts a 48-hour underwriting turn time. No underwriter spends 48 hours on your file; a first review is closer to half an hour of real attention. The rest is wait. So when a file comes back over one missing page:

You didn't lose the review. You lost your place in line.

"Fast processing" is not typing speed or heroics. It is how few times a file has to re-enter a queue.

The handoff test: three questions at every boundary

You can audit any boundary in a minute. If you cannot answer all three, files are dying there.

1. Is the packet complete, by a written definition? Not "mostly there." A one-page standard the receiver wrote and the sender agreed to.
2. Did a named person acknowledge it, with a timestamp? A department is not a person. "Submitted to processing" names nobody.
3. Whose clock is running right now? If the honest answer is nobody's, the file is in a dead zone — and dead zones never surface on a report sorted by total days in pipeline.

Run it on your three real boundaries: LO to processor, processor to underwriting, clear-to-close to the closing table. Most branches leak hardest at the first two.

The submission gap, and the kickback that fixes it

Two SLAs cover most of that boundary.

Define "complete submission" in writing. Income docs, asset docs, credit explanations, signed disclosures, third-party orders placed. A submission that misses the standard is not a submission; it is a draft. Have that fight once, in a meeting, not file by file forever.

Acknowledge within one business day. The processor either accepts the file or kicks it back, dated either way. A gap with a clock on it stops being a gap.

Then fix the kickback format, because "needs more docs" just opens a second dead zone. Every item gets three things: what is missing, who owns it — borrower, LO, or third party — and the date it is due back. No prose. An LO reading email between appointments can act on a list, not a paragraph.

The words are worth rehearsing, because they teach the behavior you want:

"I'm not starting the clock on this one yet. Four items, all borrower-side, all listed with dates. Send them together and I'll pick it up the same day."

That names the clock and tells the LO that batching gets rewarded. Both halves matter.

The conditions loop and the loop tax

Conditional approval feels like progress, and borrowers hear it that way. Operators know it is a loop, not a stage.

The underwriter issues conditions. The processor routes them. The LO chases the borrower. The borrower sends the wrong document. Back it goes, and the next review surfaces something new. Every lap is a fresh queue entry at the back.

Which gives you arithmetic worth doing.

Same four conditions, two habits. Trickled back one at a time: four laps, roughly twelve business days. Batched into one complete resubmission: about five. Seven business days bought back with sequencing alone — no faster underwriter, no extra headcount.

Three habits get you there:

• Triage every condition the hour it lands into borrower-owed, internal, and third-party. Third-party has the longest tail — verification of employment, insurance binder, payoff statement, HOA cert, title curative — and you do not control the vendor's clock. Order those the same day.
• Every open condition carries one named owner. Not "the team." A person, with a date.
• Resubmit complete. The one exception: a single prior-to-doc condition that is the sole gate to clear to close. Send that alone and say why.

And know which conditions are prior-to-doc and which are prior-to-funding. PTDs sit directly on the critical path to your CD clock. PTFs do not. A branch that treats them identically sprints on the wrong ones.

Run the runway backward

Nobody decides to blow a close date. It drifts. The submission gap cost four days, a conditions lap three more, and every slip got repriced as "still fine" until the remaining work no longer fit the runway.

The mechanics make drift expensive. The Closing Disclosure has to reach the borrower three business days before consummation, so clear to close carries a deadline upstream of the closing date, not at it. The lock expires on its own schedule, and an extension comes out of somebody's margin — branch net or LO comp, depending on your plan. And a moved closing cascades into a buyer's lease, a seller's own purchase, and an agent who will remember it.

So run the math backward on every file with a contract date.

Then use the flag day. Any file whose remaining work does not fit its runway gets flagged today, not discovered the week of closing. A flag triggers two honest options: surge it, or reset the date early. Resetting eleven days out is a phone call. Resetting two days out is a reputation.

Make aging visible: two numbers, not one

Days in pipeline is the number every branch has and nobody can use. Replace it with two.

• Days in current stage. How long since this file crossed a boundary.
• Days since last real activity. Real means something that changed what the next person has to do. A note typed into the LOS is not activity. An ordered VOE is. A dated kickback is.

A 40-day file that moved yesterday is healthy. A 12-day file that has sat six days in the submission gap is on fire, and a report sorted by total age buries the burning one under the healthy one.

So sort by stage age instead. Every morning, one person scans the top of that list and asks one question per file: what handoff is this waiting on, and who owns it? This is where software earns its keep — MAVYN puts the pipeline behind one login, and MAVIS watches files for stalls so the quiet ones surface on their own. But discipline outranks tooling. A whiteboard with honest stage-age numbers beats a dashboard nobody opens.

Your weekly meeting changes too. Stop reading the board stage by stage — everyone can read. Walk the stalls instead: oldest stage age first, owner named, unblock decided, next file. If you run a Monday pipeline meeting, this is the version that moves files.

What to do Monday

Three things, in order.

1. Sort the pipeline by days in current stage and put the top five on one page. Next to each, write the handoff it is waiting on and the one person who owns getting it through. Not the department. The person.
2. Write the two SLAs you leak on — almost certainly submission-to-processing and conditions. For each: what "done" means, and how many business days it gets. One page, recite-able.
3. Count laps for a month. Add one column: trips through underwriting. One or two laps is fine. Files at four tell you exactly which boundary to fix, in a number nobody can argue with.

By the fourth Friday you won't have a faster underwriter or a bigger team. You'll have a pipeline that flags its own dead zones — the only kind that works on the weeks you're too busy to look.