FICO Score 10T Gains Momentum in Mortgage Credit Modernization

Executive Summary

Mortgage credit-score modernization is moving from policy discussion to real-world evaluation.

In a July 30, 2026, HousingWire interview, FICO executive Ethan Dornhelm argued that the public release of historical Fannie Mae and Freddie Mac loan data gives lenders, investors and analysts a meaningful opportunity to test FICO Score 10T against older and competing scoring models.

The newly available data covers approximately 46 million to 50 million mortgages originated beginning in 2013. A July 2026 Milliman analysis found that FICO Score 10T produced stronger predictive results than VantageScore 4.0 across every origination year and all three performance measures examined. However, the research was commissioned by FICO, and the historical dataset has limitations that lenders must consider before drawing portfolio-level conclusions.

For mortgage lenders, the practical message is clear: credit-score modernization is no longer only a compliance project. It is becoming a testing, pricing, risk-management and competitive-strategy decision.

What Happened

Ethan Dornhelm, vice president of scores and predictive analytics at FICO, told HousingWire that the mortgage industry has reached a pivotal point in the transition to newer credit-scoring models.

Although the basic question behind mortgage underwriting remains unchanged—how likely a borrower is to repay a loan as agreed—the data, technology and analytical methods used to answer that question have evolved significantly.

Dornhelm described FICO Score 10T as the latest generation of FICO’s consumer credit-risk model and encouraged lenders to begin testing it against Classic FICO within their own portfolios.

The interview was published by HousingWire’s content studio and links directly to FICO resources, so its performance claims should be understood as FICO’s position. Those claims can now be compared with publicly available GSE data and Milliman’s accompanying analysis.

Why the GSE Data Release Matters

On July 1, 2026, Fannie Mae and Freddie Mac expanded their historical loan-level datasets to include FICO Score 10T information. The data spans more than 12 years, from April 2013 through September 2025, and allows market participants to compare:

• Classic FICO scores

• FICO Score 10T

• VantageScore 4.0

• Subsequent mortgage-loan performance

The underlying files include tens of millions of conforming mortgages and provide insight into borrower and loan characteristics such as credit scores, debt-to-income ratios, loan-to-value ratios, purchase and refinance transactions, and first-time homebuyer status.

That level of transparency is important because lenders, investors and researchers can now conduct their own analyses rather than relying exclusively on model-provider claims or private datasets.

Why lenders should pay attention

Public historical data can help mortgage companies examine how a new scoring model could affect:

1. Approval rates

2. Pricing and loan-level price adjustments

3. Default-risk estimates

4. First-time homebuyer outcomes

5. Capital and reserve assumptions

6. Secondary-market execution

7. Fair-lending and model-governance reviews

A score change may also interact with automated underwriting systems, overlays, debt-to-income limits and other eligibility rules. Lenders therefore cannot evaluate a new credit score in isolation.

What Milliman Found

Milliman analyzed approximately 46 million Fannie Mae and Freddie Mac mortgages from origination years 2013 through 2023.

The firm compared FICO Score 10T and VantageScore 4.0 using three established measures of predictive performance:

For origination years outside the primary COVID-19 exposure period, Milliman reported the following aggregate results:

Among first-time homebuyers, Milliman reported an 8.6% improvement in the K-S statistic and approximately 7.6% improvements in both bottom-decile lift and the Gini coefficient for FICO Score 10T relative to VantageScore 4.0.

The analysis concluded that FICO Score 10T achieved higher fit statistics in aggregate and in each individual origination year examined.

Important context

Milliman disclosed that the study was commissioned by FICO. That does not invalidate the findings, but lenders and policymakers should consider the sponsorship when interpreting the results and should independently reproduce the analysis where possible.

What Makes FICO Score 10T Different?

The “T” in FICO Score 10T refers to the use of trended credit-bureau data.

Traditional credit reports often show a borrower’s current balances and payment status. Trended data provides a longer view of how those balances and payment patterns have changed over time.

For example, a model may be able to distinguish between:

• A consumer who regularly pays revolving balances in full

• A consumer whose credit-card balances have steadily increased

• A borrower who is reducing debt

• A borrower whose utilization rate is becoming more volatile

Both FICO Score 10T and VantageScore 4.0 incorporate trended credit data. The debate is therefore not simply about which model has access to more data, but how each model weighs and interprets that information.

Current Mortgage Credit-Score Policy

The Federal Housing Finance Agency approved both FICO Score 10T and VantageScore 4.0 for future use by Fannie Mae and Freddie Mac in October 2022.

In April 2026, FHFA announced that the government-sponsored enterprises were moving forward with newer credit models and immediately allowing approved lenders to deliver certain loans using VantageScore 4.0. FHFA’s current credit-score guidance says Classic FICO remains available during the interim period, while implementation work for FICO Score 10T continues.

This means approval of a model and full operational implementation are not the same event.

Before FICO Score 10T becomes broadly embedded in mortgage originations, industry participants may need to address:

• Credit-reporting workflows

• Automated underwriting-system updates

• Pricing and eligibility rules

• Investor guidelines

• Servicing and quality-control systems

• Model validation and governance

• Consumer disclosures

• Staff and vendor training

The transition is likely to occur in stages rather than through a single industrywide switch.

Why Predictive Accuracy Could Expand Credit Access

Credit access and risk management are sometimes portrayed as competing goals. Dornhelm argued that stronger predictive accuracy can support both.

A more precise model may give lenders greater confidence to approve borrowers who appear risky under an older model but whose broader credit behavior indicates an ability to repay. It may also identify borrowers whose headline credit scores understate emerging financial stress.

In theory, better risk separation can help lenders:

• Approve more qualified applicants without increasing expected losses

• Reduce unnecessary overlays

• Price loans more closely to actual risk

• Limit unsustainable approvals

• Improve investor confidence in mortgage-backed securities

However, a model’s ability to produce a score for more consumers does not automatically mean those consumers will qualify for mortgages. Approval still depends on income, assets, debt obligations, property characteristics, product rules and underwriting requirements.

Limitations of the Historical Data

Dornhelm cautioned that the newly released GSE data must be interpreted carefully.

Few loans below a Classic FICO score of 620

For much of the period studied, the dataset contains relatively few mortgages with Classic FICO scores below 620. That creates a truncated sample and limits conclusions about borrowers outside traditional conforming-credit boundaries.

No Great Recession-style stress period

The dataset includes the pandemic and its unusual forbearance environment, but it does not include a credit cycle comparable to the 2007–2009 housing and financial crisis.

Because emergency forbearance policies changed the relationship between delinquency and borrower risk, Milliman excluded 2019 through 2021 vintages from its primary non-COVID analysis and applied special treatment to portions of the 2018 vintage.

Historical results may not match a lender’s portfolio

A lender specializing in government loans, non-QM mortgages, bank-statement borrowers, manufactured housing or lower-credit-score applicants may see different results than the conforming GSE dataset suggests.

That is why portfolio-level testing remains essential.

What Mortgage Lenders Should Do Next

Dornhelm urged lenders to move from general awareness to hands-on evaluation.

A practical testing process could include:

1. Run parallel scores. Compare Classic FICO and FICO Score 10T on the same applicants without immediately changing underwriting decisions.

2. Segment the results. Analyze differences by loan purpose, channel, geography, first-time buyer status, income, debt-to-income ratio and loan-to-value ratio.

3. Measure score migration. Identify how many borrowers move into different score bands and loan-pricing categories.

4. Back-test performance. Compare historical scores with delinquencies, defaults, prepayments and early-payment defaults.

5. Review fair-lending outcomes. Examine whether model changes create or reduce disparities across protected classes and communities.

6. Model financial effects. Estimate changes to approvals, pull-through rates, pricing, reserves and expected losses.

7. Update governance. Document model assumptions, limitations, testing methods and approval processes.

FICO says it offers programs that allow lenders to evaluate FICO Score 10T alongside Classic FICO at no added score cost, including live-origination and historical archive testing options.

What This Could Mean for Borrowers

Most consumers are unlikely to see an immediate, universal change in mortgage qualification.

The score used in a mortgage transaction may depend on:

• The loan program

• The lender

• The investor

• The automated underwriting system

• The credit-report provider

• The stage of industry implementation

Over time, newer scoring models could help some borrowers whose responsible financial behavior is not fully captured by older models. This may include consumers with reported rental-payment histories or credit patterns that become clearer when balances are viewed over time.

Other consumers may receive lower scores if trended data reveals rising utilization, growing balances or deteriorating payment behavior.

Borrowers should continue focusing on the fundamentals that generally support mortgage readiness: making payments on time, limiting revolving balances, avoiding unnecessary new debt and reviewing credit reports for errors.

Key Takeaways

Frequently Asked Questions

What does the “T” in FICO Score 10T mean?

The “T” refers to trended credit data, which shows how a consumer’s balances and payment behavior have changed over time rather than presenting only a current snapshot.

Is FICO Score 10T replacing Classic FICO immediately?

No. Classic FICO remains in use, and the transition to newer models is occurring through a phased regulatory and operational process.

Is FICO Score 10T the same as VantageScore 4.0?

No. Both are modern credit-scoring models and both use trended credit data, but they use different model designs and may generate different scores for the same borrower.

Did FICO Score 10T outperform VantageScore 4.0?

In Milliman’s July 2026 analysis of approximately 46 million GSE mortgages, FICO Score 10T produced higher predictive-performance measures across all origination years examined. The study was commissioned by FICO, so independent replication and lender-specific testing remain important.

Could a borrower’s mortgage score increase under FICO Score 10T?

Possibly. Some borrowers may receive higher scores, while others may receive lower scores. The result will depend on their credit history and how the model evaluates changes in balances, utilization and payment behavior.

Does a higher credit score guarantee mortgage approval?

No. Mortgage approval also depends on income, employment, assets, debts, property eligibility, loan-program rules and underwriting findings.

Why does predictive accuracy matter to investors?

More accurate risk estimates can improve loan pricing, portfolio management and confidence in mortgage-backed securities. That confidence supports liquidity in the secondary mortgage market.

What should lenders do now?

Lenders should begin parallel testing, analyze score migration, review fair-lending effects and measure how FICO Score 10T performs within their own loan populations.

The Bigger Picture

Mortgage credit modernization is entering a more consequential stage.

For years, the industry debated which models should be approved and how the transition should be managed. With historical GSE scores now publicly available, the conversation can increasingly focus on observable loan performance.

FICO Score 10T may prove to be a meaningful improvement over legacy scoring, but responsible adoption requires more than accepting a headline performance claim. Lenders must examine the data, test the model against their own borrowers and understand how a scoring change affects the entire mortgage decision process.

The institutions that begin that work early will be better positioned to manage risk, explain changes to borrowers and compete as the mortgage market adopts a more modern credit infrastructure.

Sources

• HousingWire interview with Ethan Dornhelm, published July 30, 2026

• Federal Housing Finance Agency credit-score guidance and April 22, 2026 announcement

• Milliman, “Analysis of FICO Score 10T and VantageScore 4.0 Predictive Power Using Freddie Mac and Fannie Mae Loan-Level Performance Data,” July 2026

• FICO analysis and mortgage-credit modernization resources

Stay Informed

Follow Mortgage.News for continuing coverage of FICO Score 10T, VantageScore 4.0, FHFA policy, mortgage underwriting and housing-finance modernization.

Homebuyers who want to understand how credit-score changes could affect their financing should speak with a qualified mortgage lender and ask which score model and loan program will be used for their application.