65 Out of 100 Loans Now Fall Out. The Real Cost Is Hiding in Your P&L.

In 2020, 65 of every 100 mortgage applications made it to closing. Today, that number is 35.

That's not a typo. According to Equifax data reported by National Mortgage Professional in April, the industry's pull-through rate has inverted in five years. Nearly two-thirds of every application a lender touches now burns staff time, vendor spend, and processing cost — and produces nothing.

Fallout has always been part of lending. What's changed is the scale, and the price of each miss. And most lenders are measuring both wrong.

The Cost You Think You Know

Ask a lending team what it costs to originate a loan, and many will tell you somewhere between $5,000 and $6,000. That's the number that shows up in internal accounting — direct labor, processing fees, the obvious line items.

The Mortgage Bankers Association puts the fully loaded cost to produce a single funded loan at $11,109 (Q3 2025). The gap between those two numbers — the $5,000 a lender books and the $11,000 it actually costs — is the part of the P&L almost nobody manages well. It's the technology overhead siloed in the IT budget. The compliance and exception handling spread across departments. The vendor fees that escalate quietly. None of it shows up next to a specific loan, so it never gets reconciled against production volume.

The cost you can't see is the cost you can't control.

Now layer fallout on top. Every one of those 65 applications that doesn't close still consumed real money before it died — credit, verifications, partial title and appraisal, and hours of processor and LO time. When only 35 of 100 close, the true cost of producing a closing isn't $11,000. It's $11,000 plus everything you spent on the 65 that went nowhere, redistributed across the few that funded.

And the per-miss cost is climbing. A tri-merge credit report that ran about $50 in 2022 now runs roughly $540, according to the Community Home Lenders of America — a 1,500% jump in four years. Credit costs alone rose 40–50% in 2026. The denominator shrank and every miss got more expensive at the same time.

Here's what that actually looks like on a single dead file:

Line Item	Cost
Tri-merge credit pull	$540
Full appraisal	$650
Processor hours	$420
Total spent before the file died	$1,610
Killed at application — bad data caught upfront	$0

One documented lender was bleeding $60,000 to $80,000 a month in fallout costs driven by poorly sequenced credit ordering and unnecessary pulls (Certified Credit case study). That's not a rounding error. That's a hiring decision, a marketing budget, or a technology investment — gone.

I wrote earlier this spring about the 53,000 home-purchase contracts that collapsed in a single month. That was the visible tip. Fallout is the same problem upstream — quieter, larger, and far more expensive than the headline cancellations.

Not All Fallout Is Created Equal

Here's the part that matters for anyone trying to manage it: most fallout sorts into two very different buckets.

The market-driven bucket. Rates move and a borrower can no longer qualify. A lock expires. An appraisal comes in below contract. A buyer gets cold feet, or the seller walks. NAR's Confidence Index shows roughly 7% of contracts delayed and 6% terminated on appraisal issues alone. You can't verify your way out of any of this. It's the cost of doing business in a volatile market, and pretending otherwise wastes everyone's time.

The preventable bucket. This is the one lenders systematically underestimate. It's the loans that die — or worse, close and get repurchased — because the data was wrong from the start:

• Income that doesn't match what the borrower claimed
• Employment that changed, or was never solidly verified
• Identity and SSN mismatches against authoritative sources
• Undisclosed liabilities, judgments, and liens surfacing late
• Hidden mortgage obligations and undisclosed properties
• DTI breaking once every debt is finally on the table — still the single most common denial reason

The defect data shows this bucket is growing fast. ACES Quality Management's latest review found eligibility defects up 292% year over year and credit defects up 166%. Fannie Mae's own post-purchase reviews keep flagging the same culprits: occupancy misrepresentation, miscalculated rental income, undisclosed liabilities, undisclosed mortgages, borrowers no longer employed by closing. And fraud is increasingly the accelerant — the FBI's IC3 logged $275.1 million in mortgage fraud losses, up 59%, explicitly tied to AI-generated documents, while the Federal Reserve Bank of Boston pegs synthetic identity fraud at $35 billion a year. Point Predictive found 70% of early-payment defaults show signs of application fraud.

The honest framing: you can't stop the market-driven bucket. But the preventable bucket is bigger than most lenders think, and it's where the money actually leaks — because those failures happen after you've already paid for credit, verifications, appraisal, and title.

Your Credit Report Is a Rearview Mirror

Here's a truth the industry doesn't talk about enough: a credit report tells you what already went wrong.

It's backward-looking by design. By the time it's back in your system, you've already opened the file, assigned a processor, and — if the timing went against you — pulled the appraisal. A credit report can't tell you about the synthetic identity. It won't surface the mortgage the borrower omitted. It won't catch the lien sitting in a civil court database, or the undisclosed self-employment, or the employer that doesn't exist. It's a rearview mirror on a file that's already consumed your team's time.

What lenders need is a windshield — something that reads the whole file the instant it lands and gives you one verdict before you spend a dime: fund it, or kill it.

That's the architecture we built PitchPoint's ADV (Application Data Verification) around. And it runs five checks your credit report will never run:

1. Identity & Synthetic-Fraud Screening. Name, SSN, address, and date of birth authenticated against the SSA Death Master File, OFAC, GSA-EPLS, and HUD-LDP watchlists. Historical SSN usage traced to expose synthetic identities before a processor ever opens the file.

2. Civil Courts & Hidden Liabilities. Civil court records scanned directly for bankruptcies, judgments, and DTI-killing liens that never appear on credit. MERS queried by SSN to surface undisclosed mortgages — the ones that turn a 42% DTI into a 58% DTI at the worst possible moment.

3. Employment & Affiliated-Business Verification. Stated employer validated. Affiliated-business search run to catch undisclosed self-employment and conflicts of interest. Stated income becomes verified fact before a processor burns hours chasing a VOE on a file that was never going to fund.

4. Property & Valuation Defense. Ownership history reviewed for illegal-flipping patterns. FEMA disaster impacts surfaced. AVMs pressure-test value before you commit to a full appraisal order — another $650 you don't have to spend on a file you're going to kill anyway.

5. Total Participant Validation. Every settlement agent and appraiser confirmed licensed and clear of government exclusion lists. An unlicensed participant isn't a paperwork problem. It's a repurchase demand with your name on it.

All five checks run together. One pass. Every check. One verdict. The output isn't a 40-page PDF — it's a single Risk Score with critical alerts surfaced at the top and a go/no-go you can act on in seconds, not hours.

How the Math Works

ADV lives natively inside most leading LOSs as an interactive workbench, not a static read-only document. Every flag can be cleared in place: mark "Cleared by Lender," attach the supporting doc, move on. No swivel-chair. No second system. Report pushes into the LOS eFolder and auto-triggers matching underwriting conditions. Zero rekeying.

The phased verification model matters here. Critical identity and liability checks run upfront for the kill-or-pursue decision. Reconfirm and Enhanced Reconfirm re-verify MERS, participants, and financials right before closing — without triggering a duplicate order. You pay once. You verify when it counts.

The numbers on the other side of that model:

• Up to 30% lower underwriting time and cost on files that proceed to closing
• <1% false positives — your team works real exceptions, not data hygiene
• 3,000+ institutions trust PitchPoint to run these checks across their pipelines
• Fannie Mae LQI-ready — the verification trail is built for post-purchase review before a reviewer ever opens the file
• SOC 2 Type II certified, 99.9% platform uptime, continuous audit — zero exceptions across 13 consecutive months

And every datapoint, note, and status change is logged on a deterministic, rule-based audit trail. If the CFPB asks why you denied a loan, you have the answer — traceable, explainable, documented. If a GSE sends a repurchase demand, you have the receipts. As I've argued before, AI-native doesn't mean black box. Speed you can't defend isn't speed. It's exposure.

What to Do Monday

You don't need a transformation project to start managing fallout. You need to measure it honestly and move the failure point upstream.

1. Recalculate your true cost per funded loan. Include the dead files. Divide total origination spend by funded loans, not applications. The number will be uncomfortable — and it's the real one.
2. Split your fallout into the two buckets. What share is market-driven and genuinely unavoidable, versus preventable data and verification failures? Most lenders have never run this cut.
3. Move verification upstream. Verify income, employment, identity, and undisclosed obligations at application — before you invest in appraisal and title. The $1,610 in wasted per-file costs is avoidable. Most of it requires catching the problem before you've spent it.
4. Insist on auditability. Whatever you automate, make sure every decision has a traceable, explainable path. If a vendor can't show you the chain, that's your answer.
5. Align vendor economics to loan progress. You shouldn't carry the full verification cost of a loan that was never going to close. Phased verification — critical checks up front, reconfirm before closing — means you stop funding dead files.

Fallout isn't going back to 2020 levels. But the lenders who measure it honestly — and stop paying late for problems they could have caught early — will protect margins the rest of the market is quietly bleeding.

---

Ready to pressure-test where your preventable fallout is hiding? We built the tool for exactly this. Stop funding dead files.

— Stephen Schrump, CEO, PitchPoint Solutions