July 2026 Mortgage Pull-Through: What Lenders Can Infer

Optimal Blue’s August 11 report puts July 2026 mortgage pull-through at 78.9% for purchases and 69.9% for refinances, both lower than in June. For lenders, the figures warrant a review of conversion assumptions alongside lower lock activity. They cannot, by themselves, establish how many July locks will fund—or when.

Purchase and refinance conversion both fell

June is calculated from the rounded July figures and changes.

Total lock volume also fell 11% month over month. Lock volume describes activity at the commitment stage, while pull-through concerns conversion to funding. Declines in both could compound pressure on eventual output, assuming comparable populations and timing.

A rate lock is a conditional agreement about pricing. The CFPB explains that the borrower must close within the agreed period and avoid relevant application changes to retain the locked rate. Counting that agreement measures a different event from disbursing a completed loan.

Which loans are behind the measures?

The report draws on two systems. Optimal Blue says its product, pricing and eligibility engine supplies lock data and covers more than one-third of U.S. mortgages in its Market Advantage overview. Its hedging and loan trading system supports approximately 40% of loans hedged and sold into the secondary market, the August release says.

Those coverage shares have different denominators and cannot be added. They also do not establish that every report table represents the same lenders or loans, or that a platform average matches an individual lender’s business mix.

Optimal Blue’s December 2025 pull-through documentation defines the measure as the share of locked loans that close and fund, and identifies CompassEdge as its source. That establishes the earlier series’ source and definition; it does not settle July’s detailed cohort rules.

Match the cohort before forecasting fundings

For mortgage lock conversion, the essential unit is a cohort: a defined group of loans followed through time. The public July release does not specify enough cohort and timing detail to reproduce its conversion rates. Before applying them to a funding forecast, resolve five questions:

• Entry date: Does a loan enter the denominator at its first lock, latest lock or another milestone?
• Outcome and window: Does success mean closing, funding or investor purchase, and how long does each loan have to reach that outcome?
• Unresolved files: Are still-active loans excluded, treated as failures or estimated? Can subsequent outcomes revise earlier months?
• Counting rules: Are extensions, cancellations and relocks linked to one loan? Is the rate weighted by loan count or balance?
• Population: Which lenders, channels and products contribute? Are purchase, rate-and-term refinance and cash-out refinance measured separately?

These definitions belong in the mortgage workflow and data technology specification. An application-to-funding ratio and a first-lock-to-funding ratio answer different questions, even when a dashboard labels both “pull-through.”

Consider a hypothetical cohort of 1,000 July locks. Assume 800 eventually fund: 600 in August and 200 in September, with none funding in July. The cohort’s eventual conversion is 80%. If another 100 loans locked before July fund in August, August production in this example is 700 loans. Multiplying July locks by 80% gives eventual cohort output, not August production.

A monthly funding forecast therefore needs both the probability of funding and the distribution of time until funding. It must account for older active cohorts as well as newly locked loans expected to fund within the forecast month.

Separate workload, funded balances and revenue

For lender operations, weaker conversion warrants a review of where files leave the pipeline. A loan that fails after processing and underwriting has already consumed staff time. Lower eventual output does not necessarily mean an immediate, proportional reduction in workload.

Build staffing assumptions around active files, their stages and the work remaining. Build funded-volume assumptions around matched cohorts and expected funding dates. For a dollar forecast, incorporate loan balances; a conversion rate measured by loan count cannot automatically be applied to aggregate locked dollars.

Revenue planning adds the margins and costs associated with the loans expected to complete. Track extensions, renegotiations and fallout costs separately instead of assuming that every funded loan preserves its original economics.

Keep model forecasts separate from observations

July’s report introduced rate and spread forecasts, including a 12-month conforming 30-year rate projection of 6.58%. The forecasting tool, Virtual Economist, combines machine learning, public economic information and Optimal Blue’s mortgage market data.

Use the projected rate path as a scenario input, with an explicit assumption about how borrowers might respond. It is not a measured future pull-through rate or evidence that today’s pipeline will convert at a particular level.

In a dashboard drawing on housing data and research, give observed results and forecasts separate labels, dates and definitions. Before changing next month’s staffing or revenue plan, reconcile the lender’s active loans with the cohorts, balances and funding dates the plan assumes.