Fair Isaac Corporation: Competing on Mortgage Scores
Download case PDF (5 pages)Request teacher’s pack
Setting: San Jose, California
Information cutoff: 10 September 2026, 5:00 p.m. Eastern Time
Management perspective: FICO senior leadership (author-created decision assignment)
Case convention: The decision assignment is hypothetical. Company, regulatory and market facts are historical; this case does not describe a private management meeting. No information after the cutoff is included.
The decision
On 9 September 2026, the Federal Housing Finance Agency (FHFA) changed an important condition in the conforming-mortgage market. All lenders approved to sell loans to Fannie Mae or Freddie Mac could use VantageScore 4.0 on eligible loans without first obtaining written approval. A lender could continue using Classic FICO. On any single loan, however, the lender had to use the same score model for every borrower. Existing credit-reporting requirements were unchanged. FICO Score 10T was not then eligible for delivery to the Enterprises. [S1]
Eligibility created a choice. It did not show which lenders would use the new option, what they would pay, whether borrowers would benefit, or how loan performance would compare. FICO’s latest public results showed strong growth in its Scores segment, while the company’s own filings identified both the mortgage market’s importance and its dependence on the major credit bureaus. FICO also offered a direct mortgage-score license with two fee designs. At the same time, historical FICO 10T data had become available for further analysis, but the model could not yet be used for Enterprise delivery. [S2–S4]
Decision assignment: Recommend how FICO should sequence three connected priorities over the next planning cycle: defend Classic FICO’s position, develop the direct mortgage licensing channel, and build the evidence and operating readiness for FICO 10T. State what FICO should do first, what it should preserve or limit, and which evidence would change the sequence. Distinguish actions FICO controls from decisions reserved to FHFA, the Enterprises, lenders and credit bureaus. Do not assume a model launch date, lender migration, price response or consumer savings.
A market built around common rules
Credit scores convert information in a credit file into a measure lenders may use in underwriting and pricing. Their value depends on more than the score itself: lenders have policies and decision systems, the Enterprises set loan-delivery requirements, and credit bureaus supply reports and workflows. A change in the score option can therefore affect product choice, operational processes and commercial relationships at once.
FHFA had approved both FICO Score 10T and VantageScore 4.0 in 2022. That approval was not the same as permission to deliver loans using either model. In April 2026, the Enterprises began a limited VantageScore 4.0 rollout. FHFA’s 9 September update then removed the prior written-approval requirement for all approved Enterprise lenders using VantageScore 4.0 on eligible loans. Classic FICO remained available. FICO 10T was still not eligible for delivery. [S1]
The September policy applied the same model to all borrowers on a given loan. It did not require a lender to adopt both models. Credit-reporting requirements remained unchanged, including the then-current tri-merge/bi-merge rules. Thus a score-model choice did not itself replace the lender’s existing report arrangements. [S1]
On 1 July 2026, the Enterprises published historical score data: FICO 10T scores for loans acquired from April 2013 through September 2025, and additional VantageScore 4.0 data for acquisitions from April 2023 through September 2025. The periods were not identical. The data offered a basis for additional research, but no comparative re-estimation appears in this case. [S1]
Academic research helps frame what evidence matters, but does not answer which mortgage model will prevail. Chatterjee and coauthors model credit scores as public information about borrowers’ persistent, partly hidden characteristics. Their U.S. consumer-credit model finds that changing credit-history tracking can help some borrowers while weakening repayment incentives and raising credit prices in the model. It is not a study of score vendors or conforming mortgages. Yannelis and Zhang examine competition and screening in U.S. auto lending; their analysis suggests effects can vary across borrower segments. It does not estimate the effect of an additional mortgage-score model. [S5–S6]
A Federal Reserve study based on 2003 credit records is useful as a reminder that comparisons depend on a common sample, outcome and performance horizon. The often-cited 73.0% Kolmogorov–Smirnov statistic was for the Federal Reserve’s estimated base model. It was not a 2026 FICO-versus-VantageScore mortgage result. [S7]
FICO’s commercial position
FICO reported two segments: Scores and Software. For the quarter ended 30 June 2026, Scores revenue was $458.9 million, Software revenue was $215.3 million, and total revenue was $674.2 million. Scores represented 68% of total revenue. These segment figures include more than U.S. mortgage scores. [S2]
Year-over-year Scores revenue rose $134.6 million in the quarter, including $131.6 million in business-to-business scores. Management attributed the B2B increase primarily to a higher unit price for mortgage origination scores. For the first nine months of FY2026, the company attributed B2B Scores growth primarily to both higher unit price and mortgage origination score volume. The filing did not disclose mortgage-only score volumes, price elasticity or the separate contribution of price and volume. [S2]
The bureau relationship was material across the company. Experian, Equifax and TransUnion collectively represented 63% of total revenue in the quarter and 60% in the first nine months; each accounted for more than 10% in both periods. These are company-wide figures, not mortgage-only revenue shares. [S2]
The question was not simply whether a direct sales route could lower a lender’s bill. Direct licensing could change who pays whom, who provides the workflow and how a fee is passed through. A different fee may also alter how revenue is earned when a loan closes. The public disclosures did not show customer-level adoption, total costs by channel, migration expense, or the share of FICO’s mortgage score business exposed to a particular lender decision.
Two ways to license a FICO mortgage score
In October 2025, FICO announced a direct mortgage licensing program. Its performance-based option set a $4.95 fee for each score plus $33 per borrower per score when a FICO-scored mortgage closed. Its alternative was $10 per score without a funded-loan fee. FICO said it offered the same options to bureaus and did not control any bureau’s downstream markup. The announcement described direct licensing as a way to remove bureau markups and improve price transparency. These were FICO’s terms and rationale, not independently measured net savings. [S3]
Experian disputed FICO’s savings framing. It argued that direct licensing could add technology, operating and regulatory complexity and cost. Experian is a major counterparty with an interest in the distribution arrangement; its statement does not independently quantify those costs. The public evidence did not resolve how fees would be passed through to a lender or borrower, or how direct and bureau-mediated arrangements would compare on total cost and service. [S4]
| FICO-announced option | Publicly stated fee design | Unresolved for a lender |
|---|---|---|
| Performance-based | $4.95 per score, plus $33 per borrower per score if the FICO-scored loan closes | Pulls per loan, funded incidence, who bears the fee, pass-through and integration costs |
| Per-score | $10 per score; no funded-loan fee | Same unresolved channel, usage and total-cost questions |
The table reproduces the announced fee structure. It does not estimate total cost per closed loan. [S3]
Three linked priorities
Defend Classic FICO
Classic FICO remained available after the policy change. FICO could emphasize its established use, product performance, operational continuity and customer support. But legacy familiarity alone would not demonstrate value under an expanded-choice regime. A lender could assess model performance, implementation effort, policy compatibility and cost together. At the cutoff, the case record did not provide lender-by-lender intentions or a current common-sample performance comparison.
FICO had a near-term commercial position to protect: management had attributed recent B2B score growth in part to mortgage unit price. A lower price might improve the offer’s appeal, but the available public numbers could not establish the volume response or contribution lost. Retaining current commercial terms might preserve near-term revenue while leaving a newly eligible alternative available in negotiations.
Develop direct licensing
The two fee designs made the payer and funding trigger more visible than an unspecified reseller markup. The performance option also tied part of the fee to a funded loan. But the public terms did not disclose lender score-pull patterns, loan conversion, resellers’ service costs or fees ultimately charged to borrowers. A nominal fee comparison was therefore not a complete channel comparison.
FICO also relied on the bureaus beyond mortgage scoring. The 10-Q’s revenue-concentration disclosure made channel design a broader relationship question, though it did not quantify the effect of this mortgage program on other products. FICO could seek to make direct access workable while retaining bureau participation. It could also favor a more direct route to improve price visibility. Either path would require evidence about counterparties’ roles, lender implementation and total service costs.
Prepare FICO 10T
FICO 10T could become a second FICO option for lenders if it became eligible and institutions were prepared to use it. Historical Enterprise data created an opportunity to test performance more rigorously. The asymmetric data windows and differences in sample eligibility would need attention. A persuasive comparison would define the target outcome, borrower and loan populations, score availability, performance horizon, and the treatment of missing or excluded observations.
Evidence would not by itself complete deployment. Lenders would need to consider policy mappings, system changes, model-risk governance, controls and operational readiness. FHFA’s approval history did not establish that FICO 10T was delivery-eligible at the cutoff, or say when eligibility might change. FICO could prepare the research and implementation pathway, but it could not promise an Enterprise decision or a launch date.
Choices and implementation
FICO did not have to make these priorities mutually exclusive. It could protect Classic FICO while improving direct-license transparency and testing 10T, but resources and partner attention were limited. The leadership team would need to choose a sequence, define evidence gates and explain how it would know whether each action worked.
Possible starting points include:
- Protect the incumbent first: preserve Classic FICO’s current product and access, support lender evaluation, and expand direct licensing only where total costs and partner arrangements are understood. Advance 10T through evidence and readiness work without promising timing.
- Make channel choice the priority: use direct licensing and its published options to clarify commercial terms, while maintaining bureau-mediated access and measuring lender experience. Preserve Classic FICO as the product and build 10T readiness in parallel.
- Lead with future-model readiness: devote more attention to a rigorous 10T evidence package and lender implementation path, while keeping Classic FICO commercially available and limiting channel changes until customer and partner effects become clearer.
These are analytical starting points, not claims about FICO’s plans. Students may combine or modify them. Any proposed spending, adoption target, migration schedule, model-performance threshold or savings estimate must be labeled as an assumption unless supported by evidence.
The team’s decision memo should identify: the first two actions; the responsible parties; what must be learned before a broader commitment; how FICO will preserve customer and channel continuity; and what evidence would cause it to change course.
Discussion questions
- What changed on 9 September 2026, and what remained unchanged? Which parties now had a choice, and what decisions were still outside FICO’s control?
- Which evidence should FICO use to defend Classic FICO? What comparison would be fair and decision-relevant?
- How should FICO assess its two direct-license fee designs against bureau-mediated distribution? What information is required before describing one as less costly?
- What should FICO do now to prepare FICO 10T while it is not eligible for delivery? Which evidence gates should precede a broader investment or rollout commitment?
- Recommend a sequence across the three priorities. What is the strongest argument against your choice, and what evidence would reverse it?
Sources and conventions
S1. FHFA, Credit Scores policy and FAQs, status as of 10 September 2026. The 9 September 2026 update and dated policy history are used; later page updates are excluded. See “Recent Updates” and FAQs on eligible lenders, one model per loan, reporting requirements and FICO 10T delivery eligibility.
S2. Fair Isaac Corporation, Form 10-Q for quarter ended 30 June 2026, filed 29 July 2026. MD&A pp. 20–28, segment tables and Note 11 customer concentration. Dollar amounts are reported in USD millions and rounded to one decimal place in the narrative.
S3. Fair Isaac Corporation, “FICO Launches Cost-Cutting Direct License Program for Mortgage Lending”, 1 October 2025, fee and channel paragraphs. FICO’s characterization of savings is an issuer claim; announced terms do not establish realized system savings.
S4. Experian, “Experian Statement on New FICO Pricing Schemes”, 3 October 2025. Interested counterparty statement, not an independent cost study.
S5. Chatterjee, Satyajit, Dean Corbae, Kyle P. Dempsey, and José-Víctor Ríos-Rull, “A Quantitative Theory of the Credit Score”, Econometrica 91(5), 2023, pp. 1803–1840. Structural model of consumer credit; not a mortgage-score competition study.
S6. Yannelis, Constantine, and Anthony Lee Zhang, “Competition and Selection in Credit Markets”, Becker Friedman Institute Working Paper 2021-99 (version read), August 2021; later published in Journal of Financial Economics 150(2), 2023. Auto-credit setting; paper-version and transfer limits apply.
S7. Federal Reserve Board, Report to the Congress on Credit Scoring and Its Effects on the Availability and Affordability of Credit, 2007, data/model section; score-performance section. Based on 2003 data; the 73.0% KS value is the Federal Reserve’s estimated base-model statistic, not a current FICO–VantageScore mortgage comparison.
All company, regulatory and market facts are frozen at the stated cutoff. Academic findings are described within their study settings; no estimates are transferred to FICO’s mortgage market. No lender adoption, consumer savings, price elasticity, comparative model performance, margin or FICO 10T delivery eligibility is assumed.