Genius AI

Mortgage cost per loan hit $11,615 in Q3. Here is where lenders are winning it back.

Mortgage lenders spent an average of $11,615 to produce each loan in the third quarter of 2026, according to the Mortgage Bankers Association (MBA), up $679 from the second quarter while loans per company fell 5.6%. The fastest recoveries come from matching capacity to volume and automating document, underwriting prep and QC work.

What did MBA's Q3 2026 performance report show?

MBA’s Quarterly Mortgage Bankers Performance Report, released November 17, 2026, covered 335 companies. Lenders averaged a pre-tax production loss of $30 per loan, a $1,003 swing from the $973 profit they posted in the second quarter. Production revenue slipped to $11,585 per loan while expenses climbed to $11,615.
Marina Walsh, CMB, MBA’s vice president of industry analysis, pointed out that Q3 has “historically been the strongest quarter of the year.” She also expects staff reductions and other cost measures to pick up through the first quarter of 2027.

Quarter

Production expense per loan

Source

Q1 2026
$11,898
MBA, via HousingWire
Q2 2026
$10,936
MBA, via HousingWire
Q3 2026
$11,615
MBA, via National Mortgage Professional
Q2 2008 to Q2 2026 average
$7,945
MBA, via Scotsman Guide

Production expense per loan

Q1 2026
$11,898
Source: MBA, via HousingWire
Q2 2026
$10,936
Source: MBA, via HousingWire
Q3 2026
$11,615
Source: MBA, via National Mortgage Professional
Q2 2008 to Q2 2026 average
$7,945
Source: MBA, via Scotsman Guide

Why does cost per loan stay high when volume drops?

Most production expense is people and the space and systems around them. MBA counts commissions, compensation, occupancy, equipment and corporate allocations in the total. When volume falls faster than a team can be resized, fixed cost spreads across fewer loans. That is exactly what Q3 showed: loans per company dropped from 1,958 to 1,848 while production staffing held roughly flat. [1] 2027 gives lenders little volume to grow into. MBA forecasts originations falling 5% to $1.99 trillion, with refinance volume down 17% to $571 billion and loan count at 5.0 million, compared with 5.3 million in 2026. On rates, MBA chief economist Mike Fratantoni told the Annual Convention audience: “I don’t think we’re going to come down that far.” MBA expects the 30-year rate to stay above 7% through 2027. So the opportunity sits inside the operation. Here is where lenders are finding it.

1. Capacity that moves with volume

A fully staffed in-house team is a fixed cost. A tech-enabled fulfillment partner turns part of that into a variable cost: you pay for the loans you close. Many lenders keep a core team and route overflow, specialty products or night and weekend work to a partner. Indecomm’s fulfillment and contract underwriting teams work inside your loan origination system (LOS), with the Genius AI Suite doing the prep.

2. Document intake and indexing

Every file starts as a stack of documents to classify, index and read, and this is where lenders have moved fastest. A joint survey from the American Association of Residential Mortgage Regulators (AARMR), MBA and Boston Consulting Group found data extraction from documents in production at 61% of the 31 lenders and servicers surveyed, with 42% running it at scale. [7] NFM Lending president LaTasha Waddy told HousingWire’s The Gathering that AI can reach “every section of the mortgage process.” [8] IDXGenius | ai classifies 1,400+ pre-mapped mortgage document types and extracts 17,000 data points per loan file, so processors start with an organized file.

3. Underwriting prep

Underwriters spend much of a file preparing to decide: calculating income, organizing conditions, checking guidelines. IncomeGenius® analyzes W2, self-employed and cash flow income in seconds. DecisionGenius assesses credit, assets, income and collateral and returns a clear-to-close recommendation with a pre-organized conditions list to the LOS. The underwriter’s time goes to judgment.

4. QC that agrees with itself

Quality control protects every dollar saved upstream, and a repurchase erases the savings on many loans. AI-assisted QC lowers review cost when auditors trust the output. Indecomm CEO Rajan Nair framed the test this way: “Can you get an automation solution that the auditors can trust, embrace and adopt?” [9] AuditGenius | ai is built for repeatable results, with every finding traceable to the document, page and field.

How do lenders move AI from pilot to production?

Interest is broad and production use is still early. In The Mortgage Collaborative’s June 2026 Pulse of the Network survey, more than 80% of lenders said they were evaluating AI tools, 17% had them in live production workflows, and trust was the most-cited barrier at 25%. [10]

Lenders who make the jump tend to follow the same sequence:

1

Pick one workflow with a clear cost: income calculation, document indexing or post-close QC.

2

Baseline it: cost per file, touches per file, cycle time.

3

Run the new process side by side with the current one for 30 days.

4

Require traceability: every output should point back to its source document.

5

Scale on measured results, then pick the next workflow.

Common ways to measure savings:

Metric

What it tells you

Where to get it

Cost per loan

The number MBA benchmarks

Finance, MBA Performance Report peer data

Touches per file

How much rework is in the process

LOS audit trail

Cycle time by stage

Where files wait

LOS milestones

Loans per FTE

Capacity gained

HR and LOS

QC defect rate

Whether quality held as speed rose

QC platform

Frequently Asked
Questions

What does it cost to originate a mortgage in 2026?
MBA’s performance reports put total production expense at $11,898 per loan in Q1 2026, $10,936 in Q2 and $11,615 in Q3. These figures cover independent mortgage banks and mortgage subsidiaries of chartered banks.
What is included in loan production expenses?
MBA’s total includes commissions, compensation, occupancy, equipment, other production expenses and corporate allocations. Compensation is the largest share for most lenders.
How much can automation reduce cost per loan?
It depends on which workflows you automate and how much of the cost is fixed staffing. Measure it on your own files: baseline one workflow, run a side-by-side test, and compare cost and touches per file before you scale.
What is variable-cost mortgage fulfillment?
It is an arrangement where a partner handles processing, underwriting or closing work and you pay per loan or per task. Cost rises and falls with volume, which protects margins when originations drop.
Where should a lender start with AI?
Start with a high-volume, rules-heavy workflow where results are easy to check, such as document indexing or income calculation. Prove the numbers, then expand.

Request an ROI Review and get your cost-per-loan baseline against
MBA benchmarks.

We use cookies to offer you a better browsing experience, analyze site traffic, personalize content and serve targeted ads. We also share information about your use of our site with our social media, advertising and analytics partners who may combine it with other information that you’ve provided to them or they have collected from your use of their services. Read how we use cookies and how you can control them in our Cookie Disclosure Policy. By using our site, you consent to our use of cookies.