Ask a mortgage operations leader what automation could save, and the answer usually comes quickly. Labor hours. Cycle time. Headcount.
Ask what underwriting costs per loan today, and the room often gets quieter.
That pause matters more than it looks. Every return is measured against a starting point. When the starting point is unclear, so is the return.
The Number Before the Number
When Indecomm polled attendees at its July 2026 executive webinar on ROI, only 35 percent said they have good visibility into their cost per loan by function, meaning processing, underwriting, quality control, and post-close. Another 47 percent said somewhat. Eighteen percent said no.
Put simply, roughly two in three operations leaders in that room could not say with confidence what each function costs them today.
This reflects how mortgage operations are built. Costs sit across several budgets, and few general ledgers are designed to show cost per loan by function. It does mean many automation business cases start one step ahead of where they should.
Business Cases Without a Reference Point
The gap shows up in how technology decisions get made, too. STRATMOR Group, a mortgage advisory firm, ran a live audience survey during a panel at ICE Experience 2026. Only 48 percent of organizations said they conduct a return-on-investment analysis before buying new technology. And 63 percent said they do not conduct one after implementation at all.
The pattern that follows is familiar. A savings estimate arrives with the proposal. The contract gets signed. Months later, results land differently than projected, and the technology takes the blame.
Sometimes the technology deserves it. Just as often, the model does. Without a baseline, there is no way to tell which, and that answer decides what gets fixed next: the platform or the business case.
Industry Averages Give Context
The cost pressure is real. The Mortgage Bankers Association put total loan production expense at $11,898 per loan in the first quarter of 2026. In 2025, lenders in the top 20 percent for net production income averaged $10,074 per loan in production expense, while those in the bottom 20 percent averaged $12,603.
Benchmarks like these are valuable. They show where the market sits and how wide the spread is between strong performers and the rest. Even the industry’s most detailed function-level benchmark, run by the Mortgage Bankers Association and STRATMOR since 1998, sorts lenders into peer groups by volume, institution type, and operating model, because a single average can mislead.
An average tells you where the industry stands. Your own numbers tell you where your gaps are.
Where the Costs Hide
Direct labor is the visible part. Processors, underwriters, closers, and auditors appear on the general ledger, and their fully loaded cost is simple to calculate.
The harder part is everything around them. In Indecomm’s work building ROI models with lenders, four cost buckets regularly go uncounted:
- Technology and platforms: licensing, maintenance, per-seat fees, integration, and upgrades.
- Overhead and allocation: management span, facilities, IT support, and administrative load.
- Risk and quality: rework, cure costs, repurchase exposure, and audit remediation.
- People-adjacent costs: recruiting, training, turnover, and overtime during volume spikes.
They add up. In one illustrative model for a mid-sized operation handling about 500 applications a month, nearly a quarter of total yearly cost sat in a shared-cost pool that many ROI models never see.
Cost is one side of the picture. Business cases also tend to miss integration overruns, the distance between demo accuracy and production accuracy, and the step between time saved and dollars saved. On the other side of the ledger, they often undercount value, because returns ramp over months and keep building well past the first year.
What a Credible Baseline Looks Like
A strong baseline does a few basic things well. It captures direct and indirect costs. It looks beyond labor to rework, exceptions, and risk. And it reflects how the operation runs day to day.
Two habits make it hold up. First, leadership signs off on the figures as the official starting point before go-live. Second, the team re-measures against actual spend and headcount every six months.
That second habit closes the post-implementation gap the STRATMOR survey surfaced. It turns an ROI estimate into an ROI result.
There is a bonus, too. Building the baseline often surfaces manual handoffs, exception leakage, and gaps between policy and practice before any automation goes live. Sometimes the most valuable output of an ROI model is the conversation it starts across operations, technology, and leadership.
Read the Full POV
This blog covers the starting point. Our point of view paper, The Real ROI of Mortgage Operations: Why the Baseline Comes First in Mortgage Automation ROI, goes deeper. It walks through the seven gaps business cases keep missing, the full breakdown of uncounted cost buckets, and an illustrative model showing what changes when AI-enabled document classification and extraction comes into the picture.