The Margin Math of a 6.7% Market

Nobody originating loans in 2026 needs a market summary. Freddie Mac’s weekly survey has the 30-year fixed in the mid-6s, Fannie Mae’s August forecast holds near 6.8 percent through year end, and the shops still standing have stopped waiting for relief. What the rate environment has really changed is the value of a lead. Every prospective borrower now costs more to find, more to win, and more to lose.

Consider the file most shops carried a version of last quarter. A borrower sourced through a referral relationship that took months to build. The tri-merge is pulled, processing hours accumulate, and then the file dies in underwriting a dozen points short of the pricing tier that made the payment work, or stalls while a paid balance takes a full reporting cycle to post. Under 2021 volume, that outcome was absorbed as a cost of doing business. At today’s volume, it can be the difference between a profitable month and a flat one.

That is the process problem this piece is about, and it is a process problem, not a market one. When leads were plentiful, the winning discipline was speed and capacity: get more files in, push them through, let volume forgive the waste. When leads are scarce and expensive, the discipline inverts. The object becomes extracting the maximum from every file that enters the pipeline, and spending as little as possible to learn that a file will not close. The countermoves are specific, and they are the subject of everything below. The shops managing that inversion track three numbers.

The three numbers

Lenders navigating this market well can answer three questions precisely:

1. What is our pull-through rate? Of the files that enter the pipeline, how many actually close?
2. What do we spend on credit data per closed loan? Not per file pulled. Per loan that funds.
3. How many of our borderline borrowers sit within about 20 points of a better pricing tier?

Most shops can answer the first. Far fewer track the second, and almost none track the third. Yet the second and third are where margin is won in a tight market, because both are controllable this quarter, without waiting for a single basis point of relief.

Number one: pull-through sets the multiplier

Pull-through is the multiplier on every acquisition cost, and honest numbers here are humbling. On the broker side, commonly cited pull-through runs between 25 and 30 percent. Relationship-driven retail does better, and the marketing mix behind the leads swings the number wildly: advertised leads pull through at a fraction of the rate referred ones do. Even an optimistic shop converting 60 percent of files is paying for roughly 1.7 files of credit data, processing, and effort per closing. At 30 percent, that becomes more than three full files per closing, and every dollar of per-file cost more than triples on its way to the bottom line.

The instinct in a thin market is to widen the funnel and work more marginal files. The math argues for the opposite: qualify harder, earlier, and cheaper, so the files that consume full-cost resources are the files likely to close.

Number two: spend per closed loan is a design choice

The pain here is the dead file paid for in full. Credit data spend per closed loan is not fixed by the market. It is a function of how a shop sequences its pulls.

A soft-pull pre-qualification such as SoftQualify resolves the qualification question at roughly 70 percent less cost than a full tri-merge report, with no hard inquiry on the borrower. Files that were never going to qualify exit the pipeline before they consume tri-merge pricing. Conditional-cascade tools take the same logic further: pull a single bureau, check it against lender guidelines, and escalate to the full report only when the file supports it. On failing files, that sequencing saves roughly two thirds of the cost.

Pricing structure matters as much as sequencing. Under closed loan pricing, lenders pay modest pre-qualification fees up front and then pay for full credit reports only on loans that close. The result is that credit data cost tracks revenue instead of activity, which is precisely the alignment a low-pull-through market punishes lenders for not having.

Number three: the 20-point borrowers

The pain here is the decline that did not have to be one. Loan-level pricing adjustments step at credit score boundaries, which means a borrower sitting just below a tier line is more expensive to place, at today’s rates, than the same borrower was at 4 percent. Depending on the loan and the pricing grid, a tier move can change the rate a borrower qualifies for, the monthly payment, and occasionally whether the loan is viable at all.

Those borrowers are more common than most pipelines assume, and their scores are often held down by data that is simply outdated or incorrect: a paid balance still reporting, a resolved collection, utilization that dropped after the statement date. Rapid rescoring updates corrected information in days rather than the 30 to 60 a normal reporting cycle takes. Files rescored through Score Express see an average improvement of 23.9 points, typically within 72 hours, with every file reviewed by an FCRA-certified analyst team.

One distinction matters here: rescoring is not credit repair. It does not dispute accurate information. It ensures the bureaus reflect corrected, creditor-confirmed data quickly enough to matter inside a lock period. In a market where 20 points can move a pricing tier, the shops that systematically screen for near-tier borrowers are converting loans their competitors are declining.

Playing offense in a thin purchase market

Margin discipline protects the downside. The upside in a slow purchase market comes from referral flow, and referral relationships deepen when lenders give partners and borrowers something concrete to use. SmartPay puts referral tools in a lender’s hands and lets consumers pay for their own credit reports up front, which keeps prospect flow moving without adding acquisition cost to the lender’s side of the ledger. In a market where every lead is fought for, a referral channel that funds itself is the closest thing to free volume.

Run the math on your numbers

Industry averages make interesting reading and poor decisions. The three numbers above only become actionable when they are your numbers: your pull-through, your spend per closing, your borderline files.

The Profit Simulator runs this arithmetic on your own volume and mix in about two minutes, with no obligation and no hard sell attached. And if numbers on a page raise questions a calculator cannot answer, a conversation with our team can. Credit Technologies has helped lenders manage credit costs and qualify more borrowers since 1990.