By Brian Vieaux, CMB | President of MISMO
I encountered mortgage manufacturing cost claims ranging from $125 to $12,500 this week. Before we celebrate a hundredfold difference, we need to know whether the numbers measure the same work.
That is why I have been researching these claims and talking with industry leaders. We compete for borrowers and recruit loan officers, often doing both at once. Both audiences deserve clarity about the numbers lenders use to win their business or their careers.
Start with the industry benchmarks. MBA reported average production expenses of $10,936 per loan for independent mortgage banks and bank mortgage subsidiaries in the second quarter of 2026, down from $11,898 in the first quarter. For retail-only lenders, including consumer direct, the figures were $11,754 and $12,674.1 Freddie Mac’s November 2025 study put retail-only lenders at approximately $11,800 for the second quarter of 2025.2
Company and vendor claims belong in a separate category. AngelAi founder Pavan Agarwal reported Sun West’s manufacturing cost “below $125”3; National Mortgage Professional identified exclusions for “sales, marketing, and third-party expenses.”4 Better’s $3,000 claim and ElevenLabs’ reported 41% reduction raise a different problem: insufficient disclosure to reconcile their scope and baselines.5
Sales and customer acquisition are central to these comparisons. Some figures explicitly exclude them; others leave us unable to determine what is included.
Karen Postiglioni makes that distinction in The Mortgage Scoop’s “Show Your Math: What’s Inside Better’s ‘$3,000’ Loan?” She traces the $3,000 figure to processing with Betsy, Better’s AI voice agent, compared with an industry average—not a demonstrated company-wide reduction from Better’s own baseline. “That doesn’t mean the number is wrong. It means we still don’t know what $3,000 measures.”6 Better’s 10-Q separately reports compensation, technology, marketing and loan origination expenses; the last category alone is not fully loaded production cost.7
Standards matter.
Across its various benchmarking studies, MBA deserves credit for measuring and reporting cost to originate with defined categories and meaningful peer comparisons. Its production expense measure includes commissions, compensation, occupancy, equipment, other production expenses and corporate allocations.8
The Quarterly Mortgage Bankers Performance Report targets independent mortgage companies and bank subsidiaries, while the separate MBA and STRATMOR Peer Group Roundtables program distinguishes operating models more explicitly.9 Its cohorts include large banks, community banks and credit unions, hybrid banks, and large and mid-size independents, recognizing that depositories and IMBs need appropriate peer groups.10
Princeton Mortgage CEO Rich Weidel offers a defined comparison. He reports Q2 2026 operations and corporate wages of $1,706 per funded loan against $2,762 for a Richey May Select peer group of 67 retail IMBs, using identical line items. He says those figures exclude benefits, technology, vendor costs and originator compensation.
Weidel could advertise $716 by counting only underwriting, closing, post-closing and operations management direct labor. “Obviously that is misleading,” he says. He recommends separating sales costs, direct fulfillment labor and corporate overhead. A narrower metric is useful when its limits are clear; presenting it as the whole cost is where the trouble starts.
Melissa Langdale makes the business-model point directly:
“We’ve been treating cost of origination like one number when there are really five different business models producing it. A TPO lender, a distributed retail IMB, a depository, a broker, and a consumer direct lender each build cost differently, and until we segment the benchmark by model, we’re going to keep comparing things that are really not comparable.”
Langdale separates cost per funded loan, manufacturing, sales and vendor costs for her clients, with vendor costs also included within her manufacturing view. Those are analytical views that need reconciliation, not four independent totals to add together.
She also notes that a third-party origination lender may need fulfillment capacity for volume swings without direct visibility into borrower leads, affecting the balance between sales and manufacturing expenses.
When I hear cost to manufacture, I think about everything required to acquire the borrower and deliver a funded, saleable loan. That includes processing, underwriting, closing, capital markets, compliance, quality control and management, with benefits, payroll taxes and overhead, plus originator compensation, marketing and customer acquisition.
It includes point-of-sale and loan origination systems, integrations, data and cybersecurity, alongside credit reports, appraisals, title services, lender’s title insurance and document preparation. Lender expenses and borrower-paid charges represent different perspectives, so required external services should identify who pays and be counted once; recurring insurance, escrows, taxes and discount points need separate treatment.
Matt VanFossen asks: “what did the technology cost to develop, and what does it cost to maintain? A consistent metric needs expensed development, ongoing maintenance and appropriate amortization of capitalized investments, with investment spending disclosed separately to avoid double counting. If technology supports both the lender’s own originations and outside platform clients, the allocation between those activities needs to be disclosed before either per-loan figure becomes a benchmark.”
Wholesale efficiency must account for both the brokerage and the lender, adjusting for payments between them. Moving sales outside the lender does not make the expense disappear. Channel labels cannot establish which model is cheaper.
UWM’s first-quarter 2026 investor presentation says its business-to-business model “minimizes customer acquisition cost.” Its 2025 Form 10-K explains that compensation paid to independent mortgage brokers enters the purchase price of originated loans used to calculate loan production income. That means operating expenses alone would overlook compensation captured elsewhere in the financial statements.
MBA’s Marina Walsh explains that PGR treats fees paid to brokers and correspondents as contra revenue, ultimately netted into gain on sale, to avoid double counting among participants. Account-executive commissions remain expenses in TPO channels, and PGR separately identifies loans a retail lender brokers out.
Walsh also clarifies the limit: PGR does not track the complete operating costs of pure brokerages, beyond the payments captured by lenders. That is where a combined channel measure would require additional brokerage data.
Broker compensation paid by a lender is revenue to the brokerage, supporting expenses and potentially profit; it is not interchangeable with an employee’s salary, benefits and payroll taxes. We should combine underlying production expenses, eliminate duplicated payments and distinguish brokerage operating costs from retained profit, including consistent treatment of a producing owner’s labor. Otherwise, adding the lender’s broker payment and the brokerage’s expenses counts the same work twice.
Coby Hakalir connects the measurement question to the pressure lenders face: “Our industry is living through tightening margins, rising commissions, and a rising cost of doing business, and AI is entering the picture in a way that will genuinely change how fast and how accurately we can process a loan. We won’t know whether that technology is moving the needle, or just shifting the same costs to a different line item, unless we all measure ‘cost to manufacture’ the same way. That includes both sides of a wholesale loan: the wholesaler’s cost and the brokerage’s, not just whichever number is convenient to publish. Our industry’s survival as a core pillar of the housing market, in a world that’s changing faster than we are, depends on the quality of our data, and that starts with the numbers. All of them.”
Building on MBA’s work, I would ask lenders, brokers, wholesalers, technology companies and accounting experts to establish a common reconciliation for public cost claims.
Use funded loans as the denominator, include fallout expenses, and disclose the period, volume, loan mix and average balance, distinguishing loans originated from loans processed for platform clients. Report dollars per loan and basis points; explain overhead, technology, warehouse financing and secondary-market treatment; distinguish average cost from marginal cost and actual results from projections.
Show fulfillment, sales, technology and external services within a reconciled total, then compare the relevant business models. Include loan quality, rework and borrower experience so we can judge whether savings produce dependable loans and reach homebuyers.
For loan officers, I would make this an early question on the next recruiter call: “Can you show me your fully loaded cost to manufacture per funded loan, including sales compensation, and explain what is excluded?” Ask for the period and relevant channel, then compare the same categories across prospective employers. Use that information to question promises about pricing, compensation and operational support.
Consumers use the standardized Loan Estimate to compare loan offers. Loan officers should bring that same discipline to comparing employers. A common cost presentation would make recruiting claims easier to evaluate and efficiency claims to borrowers more accountable. A lender’s production cost is not the borrower’s price, however; consumers still need to compare actual Loan Estimates.
If a lender uses its cost advantage to win a customer or recruit a loan officer, it should be prepared to explain the number.
Sources
1. MBA second quarter 2026 production expenses. August 18, 2026. Quarterly report tables B1 and E1, printed pages 2 and 17; all-channel and retail/consumer-direct figures. MBA release
2. Freddie Mac retail lender benchmark. November 18, 2025. Retail-only lender cost study, Q2 2025. 2025 Updates to the Cost to Originate Study
3. Pavan Agarwal public statement. September 9, 2026. Public statement of manufacturing cost below $125. Pavan Agarwal LinkedIn post
4. AngelAi claim scope and exclusions. September 9, 2026. Claim scope, exclusions and comparison basis. AngelAi Lands 100 Million To Scale Mortgage Automation
5. ElevenLabs case study about Better. February 9, 2026. Vendor case study reporting a 41% reduction. Better case study
6. Karen Postiglioni on the Better cost claim. Karen Postiglioni, “Show Your Math: What’s Inside Better’s ‘$3,000’ Loan?” The Mortgage Scoop, August 27, 2026. Read Karen’s article in The Mortgage Scoop
7. Better financial statement expense categories. Q2 2026 Form 10-Q, Note 7 and expense definitions; software capitalization and amortization. Also cited by The Mortgage Scoop. Better Q2 2026 Form 10-Q
8. MBA production expense definition. April 27, 2026. Categories included in total production expense. MBA Chart of the Week
9. MBA Mortgage Bankers Performance Reports population. Target population for quarterly and annual performance reports. MBA report overview
10. MBA and STRATMOR distinct peer groups. Peer-group information, supplemented by Marina Walsh’s editorial clarification. MBA peer-group participation | MBA peer-group FAQs
11. UWM on wholesale economics and customer acquisition. May 8, 2026, slides 18–19. Qualitative business-model claims, not a reconciled channel-wide cost measure. UWM first-quarter 2026 investor presentation
12. UWM treatment of broker compensation. 2025 Form 10-K, printed page 43 and Note 1. Broker compensation treatment; comparison implications are the author’s analysis. UWM 2025 Form 10-K on SEC EDGAR
80,000+ mortgage professionals get this every weekday morning.
By submitting this form, you are consenting to receive marketing emails from: . You can revoke your consent to receive emails at any time by using the SafeUnsubscribe® link, found at the bottom of every email. Emails are serviced by Constant Contact