Fair Isaac

IThe business

Fair Isaac Corporation, known as FICO, runs its main office in Bozeman, Montana.¹³ It has two segments. Scores sells the FICO Score, a number from 300 to 850 that estimates how likely a person is to repay a loan. Software sells programs that help banks, insurers and others make decisions in real time, such as whether to approve a card or flag a payment as fraud.⁶ In fiscal 2026 Scores brought in $1.65B of revenue and software $0.88B. Total revenue was $2.53B, up 27%, and operating income was $1.27B.²

That split hides how different the two are. Scores earned a 90% operating margin in the first nine months of fiscal 2026. Software earned 28%.³ So about 65% of revenue produces more than 80% of segment profit C (calculated).³ Scores is the business that matters to the share price, and one part of it matters most.

How a score makes money

FICO does not keep credit files. The three bureaus, Experian, TransUnion and Equifax, do. FICO sells them a formula. When a lender asks a bureau for a score, the lender pays the bureau, and the bureau pays FICO a fee for each one.⁶ So FICO’s customer in business scores is not the lender but the bureau, and the three bureaus together were 63% of FICO’s revenue in the June quarter, up from 54% a year earlier.³ The same three own VantageScore, which competes with FICO.⁶ I come back to that, because it is the oddest feature of this business.

Scores sold to businesses, called B2B, were $0.95B in fiscal 2025, or 81% of Scores revenue. Scores sold to consumers, B2C, mostly through myFICO subscriptions and bureau partners, were $0.22B.⁶ In fiscal 2026 B2B rose to about $1.42B, on nine months reported and my estimate for the fourth quarter C (calculated).³² B2C was about $0.23B. The lift in Scores is nearly all B2B.

The mortgage, and the price

Mortgage lending is where FICO earns the most per score. Fannie Mae and Freddie Mac, which buy most US home loans, have long required a FICO Score on each loan they buy.⁶ FICO’s own filing names that requirement as a risk: if other models are approved, the effect could be “material” and adverse to revenue, results and the stock price.⁶ The filing does not say how much of B2B revenue is mortgage. I can’t judge that number, and it is the most important one on this page. A company that told me the split would make my work easier.

What the filings do show is that price did the work. In the June quarter, business-score revenue rose $131.6M, and the filing credits “a higher mortgage origination scores unit price.” In the fourth quarter business-score revenue rose 39%, again mostly on price, with fewer mortgages originated.³² Over the year, B2B went from about $0.95B to about $1.42B C (calculated). In the December and March quarters the filings credit both a higher unit price and more mortgage originations. By the June quarter the credit is to price, and in the fourth quarter volume fell.¹⁵¹⁶³² The company raised what it charged, and lenders paid.

Here is the price list as FICO set it in October 2025. A lender could pay $10 a score for 2026. Or it could pay a $4.95 royalty plus a $33 fee when a loan closes, per borrower and per score.¹¹ A mortgage usually pulls a score for each of the three bureaus. At $10 that is $30 for one borrower and $60 for two.¹⁷ Set that against what the other side charges. TransUnion offered its VantageScore at $4 for 2026, Experian said free, Equifax $4.50.¹¹ In March 2026 TransUnion was reported to have cut its price to 99 cents, with the others following.¹² A fee that had been small enough to ignore had become the cost the whole industry fought over.

Reports differ on when the one-grid pricing takes effect. Scotsman Guide, in a preview, says it applies to loans settled or issued from 1 October. HousingWire’s report on the same plan gave no date.²¹⁴ I can’t resolve that. What I can say is that KBW counted VantageScore at 5.5% of Fannie and Freddie originations in September, flat on August, and that two lenders, Rocket Mortgage and UWM, were 95% of it.⁴ So the damage so far is small. A trade piece, citing Fannie Mae data, reports that 37% of September Desktop Underwriter applications included a VantageScore 4.0 score.¹⁷ That counts applications that carried one, not loans scored by it, so I do not read it as a 37% share. It does show lenders running both scores already. The fear is what happens if the other lenders follow.

Software

Software revenue was $882.7M in fiscal 2026, up 7%, and it rose 19% in the fourth quarter, against 3% over the first nine months.²³ The measure to watch is annual recurring revenue, which strips out one-off licence timing. It was $747.3M a year ago. At 30 September, the company reports, it was up 14%, with platform up 72% and everything else down 18%.³² By my arithmetic that is about $850M, of which about $453M is the FICO Platform, a modular system lenders can build on, and about $397M is older products C (calculated).² The net retention rate, a measure of how much existing customers spend year on year, was 112% for software as a whole and 156% for the platform.²

It is a real change. Platform ARR has gone from $264M to roughly $453M in a year. But two things temper it. First, some of the platform growth is existing customers moving across from older products, and the old ones are shrinking at 18%. Second, Software segment operating income fell 8% in the first nine months while revenue rose 3%, with the margin down to 28% from 31%, on hosting costs.³ Software is changing shape and has not yet shown more profit.

Scale and the rest

In fiscal 2025 financial services were 92% of revenue and the Americas 87%.⁶ The company says its clients include three-quarters of the largest 100 US financial institutions and eight of the top ten US property and casualty insurers.⁶ It had 3,876 employees at the end of June, and on 1 October said it would cut about 15% of positions, with about $27.0M of charges, as it simplifies layers and brings AI into product development.³¹³ I take that at face value: a firm whose revenue grew 27% does not cut 15% of staff unless it sees something in its costs or its revenue it wants to be ready for.

IIThe moat

The protection is a standard. A bank that approves a loan wants a number its own investors, regulators and borrowers already understand, and since 1989 that number has been the FICO Score.⁶ A score is useless on its own: it matters because everyone else uses it. Once it is in a lender’s models, in secondary-market rules, in the pricing of mortgage bonds, switching costs time, risk and approval. That is the source of the 90% segment margin.

Evidence it is holding

Three things say the franchise still has force. First, price. FICO took B2B revenue up about 50% in a year, and lenders kept paying.³² A weak franchise cannot do that. Second, FICO’s answer to the grid was not a price cut. It held $10 for a bundle of two scores for lenders in the Preferred Program, who commit to FICO through their resellers.⁵ Analysts at Mizuho read the program as meant to stop “waterfalling,” where a lender orders the cheaper score first and buys FICO only when it must. They had assumed a 5% price rise for 2027 and now model flat.¹⁹ Third, VantageScore’s measured share is small: 5.5% of GSE originations in September, mostly two lenders.⁴ But the market did not take comfort. The shares fell 13% on the Monday after the price hold, when the FHFA director posted that he was “taking more action this week.”²⁰⁹

The Preferred Program has a hook. Enrolled lenders must pull FICO on 100% of applications, including soft pulls, so ordering VantageScore first no longer lowers their cost.¹⁹ A reseller told HousingWire that lenders say they will now pull both, so total costs could rise even at $10.¹⁹ That helps FICO and tells you why regulators and lenders may push back.

Evidence it is narrowing

The case against is shorter and heavier. The regulator that made FICO the standard has now accepted VantageScore and priced the two alike.⁴⁷ The company that sells the rival is the company FICO sells through. All three bureaus have priced VantageScore at about $1 or less for lenders who want it.¹²¹¹ And the bureaus have a reason to prefer their own score, which costs them nothing to sell, over FICO’s, which costs them the royalty. FICO’s own 10-K says its score’s ability to compete in mortgages may be “constrained by the pricing and other business practices” of the bureaus.⁶

The weak point in the case against is timing. The new pricing grid has no clear date. Mortgage bond investors have not shown how they value VantageScore pools. I do not know how fast lenders will move, and neither does the company: on 4 November it said it was “in the middle of this pricing transition” and gave no guidance for fiscal 2027.² That is a sign of honesty. It is also a sign that the one number most investors want does not yet exist.

IIIThe competitive advantage

Almost no cost sits between a FICO Score and the cash it earns. Scores segment operating expenses were $142M against $1,169M of segment revenue in fiscal 2025, or 12%.⁶ The formula is written once and run on the bureaus’ computers. That is why a price rise lands as profit, and why a price cut would too, in reverse. Incremental margin on price in this business is close to 100%.

Where FICO wins

It wins on trust and track record in lending that is not mortgage. Credit card, auto and personal lending decisions use FICO in most US banks, and the recent rule changes were about mortgages, not these.⁶⁷ It wins on cash. Operating cash flow was $1.07B in fiscal 2026, and free cash flow $1.03B, which is 41% of revenue C (calculated).² It wins on pricing power, shown by the 2026 numbers. And it wins on software momentum: platform ARR up 72% and 156% net retention are strong numbers, however small the base.²

Where it loses

It loses on distribution. FICO sells to three bureaus, and each owns a rival. In a normal business a seller with a rival-owning distributor has weak bargaining power. FICO’s power came from the rule that mortgage lenders needed its score. With the rule softening, the bureaus’ leverage grows. Bureau revenue went from 51% of FICO’s total in fiscal 2025 to 63% in the June quarter, so the dependence is growing.⁶³

It loses on price against a free rival, which has no other way to win. A $0.99 product does not need to be better. It needs to be good enough that a regulator calls it fair, and the 20-point penalty that was applied to VantageScore in the old grid is gone.⁴ And it loses on political risk. The man who runs FHFA announces policy on social media, and the effects are large: a 26.5% fall in one day.¹⁹⁹ FICO cannot negotiate with that.

The two businesses against each other

At 90% margin, the mortgage score business is worth many times what software is, so the shares trade on what the mortgage-score decision does. The software business had $882.7M of revenue and made about 27% at the segment level E (estimate).³² If software were worth 3 times revenue and Scores nothing, FICO would be worth about $2.6B E (estimate). The market caps the company at $15.1B and enterprise value at $20.2B C (calculated). So close to 90% of what an owner is paying for is the score.

IVManagement

Will Lansing has led FICO since January 2012. He is 67, a former McKinsey consultant who ran internet and retail companies before joining.⁶ A CEO of fourteen years has seen a full cycle, and the share price record under him is outstanding: from $117.53 in November 2016 to $2,382.40 at the high in November 2024, a twentyfold rise, before the 70% fall C (calculated).⁹ I give him credit for the first part. I give him no credit yet for the second.

How they are paid, and what that pays them to do

For fiscal 2025 Lansing’s reported pay was $36.0M: $0.75M of salary, $1.1M of cash bonus and $34.1M of stock awards. That was 368 times the median employee, who was paid $98,000.⁸ The stock is split three ways. A third is restricted units that vest with time. A third is performance units, earned on one year of results: half on adjusted revenue and half on adjusted EBITDA. The fiscal 2025 units paid out at 150.2% of target, on revenue of $1,990.9M and EBITDA of $1,109.8M.⁸ A third is units that pay by how FICO’s total return ranks against an index over three years, with the target set at the 55th percentile.⁸ The company says two-thirds of the long-term award is performance-based, against 52% at peers.⁸

What does this pay him to do? It pays him for one-year revenue and profit. In a business where a single price decision can add hundreds of millions of revenue in a year, as it did here, a one-year plan rewards raising the price now. I do not say that is what he did. I say the plan would not have penalized it if the damage showed up in the three-year rank, and it does not distinguish revenue that is price from revenue that is volume. The share-rank award is the useful check. After a 70% fall it is likely to pay little.

One more thing: Lansing is retirement-eligible, which means his equity awards keep vesting on schedule if he leaves.⁸ I find no sign he plans to. It is still a fact an owner has to weigh.

How much of their own money is in

The proxy shows Lansing with 420,431 shares, 1.77% of the company, a stake that includes a small number of options and units, and about $295M at $701.03 C (calculated).⁸ The ownership guideline for the CEO is 100,000 shares.⁸ All directors and officers as a group own 719,282 shares, or 3.02%. Vanguard holds 12.78% and BlackRock 8.60%, according to their filings.⁸ Those two are index and fund managers, not activists, and they are named here as holders, not as views.

I read every Form 4 since November 2025. I found no open-market purchase by any officer or director. Sales were small: about $10.6M over the period, of which $4.2M was Lansing’s, from about 2,400 shares sold in November 2025 at around $1,730; $2.6M was Weber’s, and $3.9M was from three directors, all at prices between $1,227 and $1,826 a share C (calculated).²² Most other entries are shares withheld for tax when units vest. That pattern is not a signal. But none of the top team has bought after a 70% fall, while the company itself bought $3.05B.

What they did with capital over a full cycle

Over the ten years to fiscal 2026 FICO earned $3.9B of net income and generated $3.6B of owner cash, after stock pay. In the same years it spent $8.7B on buybacks, or 2.4 times owner cash and 2.2 times net income. The gap was borrowed. Total debt rose by $4.65B, from $0.60B to $5.26B C (calculated).¹⁰² Stockholders’ equity went from +$0.47B to −$3.87B, because buybacks ran far ahead of earnings.

Did it pay? For per-share growth, yes. Net income grew 6.3 times, but earnings per share 8.8 times, because the diluted share count fell from 32.2M to 23.0M on average, or 29% C (calculated).¹⁰² That is the case for buybacks when the price is below value. The question is the price. Here is what the company paid.

By the end of the cycle the pattern is plain: modest spending when the shares were below $500, and the largest sums when they were above $1,200. That is how it works in most companies that buy back stock. A buyback is not a verdict on value, it is a choice, and this one added $2.2B to debt that costs about 6% a year at the margin. At the June 2026 quarter’s average of $1,149 a share, the 1.7M shares bought in that quarter have lost about 39% of their cost C (calculated).³ If the base on this page is right, FICO overpaid.

Governance is conventional. In March 2026 shareholders approved removing a supermajority requirement for amending part of the charter, and allowing officer exculpation as Delaware law permits.²⁸ I do not read much into either.

VFinancials

Ten years of FICO show two eras. Until fiscal 2020, it was a decent software and scores company that grew revenue about 11% a year with operating margins below 25%. Since then the Scores segment’s pricing took over.

Revenue has compounded at 11.7% a year since fiscal 2017, operating income at 24% and diluted EPS at 27% C (calculated). Growth was 1.7% in fiscal 2021 and 4.6% in fiscal 2022, and then 10%, 13%, 16% and 27% as pricing took hold. The most recent growth rate is not the one to project.

The Scores segment is almost the whole of the difference. Between fiscal 2023 and 2026, it went from $0.77B to $1.65B, and operating income at the segment level went from $0.68B to $1.12B in the first nine months alone.⁶³ Software added $0.14B in the same three years.⁶² The cost of getting there did not rise: Scores segment expenses were $93M in fiscal 2023 and $142M in 2025.⁶

What an owner could really take out

The reported numbers need care. Fiscal 2026 GAAP net income was $835.0M, or $36.32 a share. The company’s non-GAAP figure is $988.3M, or $42.99 a share, which adds back $185.5M of stock pay and the one-offs, and subtracts a tax adjustment.² I use the GAAP figure and treat stock pay as a real cost, because FICO pays its people in shares it then buys back with cash: it paid another $112.6M in cash for employees’ tax on vesting shares this year.²

Free cash flow was $1.03B. Take away stock pay of $0.19B and the owner cash is $0.85B C (calculated).² Three things push the number I would use going forward below that. Working capital added $51.2M to cash this year after taking $18.5M the year before.² Interest will be higher: the fourth quarter’s other expense of $29.1M includes a $47.9M gain on selling an equity stake, so underlying it was about $77M, or about $0.31B a year against about $0.21B in fiscal 2026 C (calculated).² And $65M for the settlement and $27M for the restructuring are accrued but not yet paid.²¹³ Taking out the working capital help and the extra interest, net of tax, leaves roughly $0.7B of run-rate owner cash, and the $65M and $27M still to be paid take about $0.07B more out of next year’s cash E (estimate). Against reported earnings of $0.83B the run-rate is about 13% lower; against the non-GAAP $0.99B it is about 26% lower.

Quality of earnings

Four points. First, the earnings are mostly price. B2B scores revenue rose about 50% with volume of mortgage originations “below historical norms.”² Price is the least repeatable kind of growth. Second, the tax rate rose to 24.1% from 18.8% as the excess tax benefit from stock awards shrank with the share price.² That is a real cost, not a one-off. Third, receivables rose 14% to $603M on revenue up 27%, which is fine: it says collections are not stretching.² Fourth, Q4 contained a $47.9M gain, a $65M accrual and a $27.3M restructuring charge, which I remove in my underlying operating income of $1.36B, or 53.6% of revenue C (calculated).²

The balance sheet

Total debt was $5.26B at 30 September and cash $0.22B, so net debt is about $5.04B C (calculated).² That is 3.7 times my underlying operating income and about a quarter of enterprise value. Stockholders’ equity is −$3.87B. Return on equity therefore means nothing here, and I do not use it.²

The debt is fixed or priced off short rates at about 6%, and interest cover on my underlying figure is about 4.4 times. That is comfortable today. It is also the first time in this file that the answer to “what if revenue falls?” has a second part. A 20% fall in operating income leaves the debt service fine but takes owner cash down by more than 20%, and with 2028 maturities close together it would mean refinancing in a bad market.

VIValuation

First, how this goes badly wrong

Before the numbers I asked what would most likely ruin this business or this investment, and worked back. Three things could, and none needs a surprise.

• The mortgage score becomes a commodity. The old rule that mortgage lenders needed FICO has been loosened, a rival sells for about a dollar, and the regulator prices both alike.⁴¹² If lenders end up paying FICO a fraction of $10, or only on the loans that need it, B2B revenue falls, and at a 90% margin nearly all of the lost revenue is lost profit. In my bear case business-score revenue ends 2030 at $0.99B against $1.42B today.
• The distributors use their position. The bureaus paid FICO 63% of its revenue in the June quarter and own the rival.³⁶ They can seek a lower royalty without anyone seeing a headline price change. FICO cannot easily go around the bureaus, because it holds no credit data.⁶ The risk is not a dramatic event. It is a slow trade of royalty for volume.
• Debt turns a setback into a loss. Net debt is $5.04B, about a quarter of enterprise value, and roughly $1.76B comes due in mid-2028.²³ A 30% fall in the value of the business is a fall of about 40% in the shares at today’s mix C (calculated). Buying the stock back at $1,205 with borrowed money made it worse.

They are versions of the same question: will mortgage lenders keep paying a toll that the regulator no longer requires? A fourth I leave out of the model because I can’t size it is the antitrust case. FICO accrued $65M for a proposed settlement of class actions in the fourth quarter, and the June 10-Q described a remaining claim under Section 2 of the Sherman Act in discovery. The proposed settlement may end it; I have not seen its terms.²³

What the price implies

At $701.03 the market values FICO at $15.11B on 21.56M diluted shares, the fourth quarter’s diluted average, because the 10-K with the period-end count is not out C (calculated).¹² Add $5.11B for net debt and the unpaid settlement and restructuring, and the enterprise value is $20.23B. That is 19.3 times fiscal 2026 GAAP earnings and 16.3 times the company’s non-GAAP figure, which adds back stock pay C (calculated).² Owner cash was $0.85B in fiscal 2026, a 5.6% yield, or about 4.8% on my run-rate figure, against a ten-year Treasury at 4.90% C (calculated).²²⁹

The close was before the results, which came out after the market shut. By 7:58 PM Eastern Nasdaq showed the shares at $657.66, 6.2% below the close.¹ I use the close, as the date rule requires. The call began at 5:00 PM Eastern and I found no coverage of it; a number on 2027 pricing from that call would move my base. On the after-hours price the gap to my base would be about 7%, not 14% C (calculated).

What does the price pay for? Holding my base for software and costs, it needs business-score revenue to grow 3.5% a year for ten years from today’s $1.42B, with no step-down first. If software does not grow at all, it needs 5.0% a year. Either is faster than my base, where B2B revenue is lower in 2028 than today.² The market is paying for the franchise to come through the mortgage fight intact and then grow.

What I can’t judge

Two inputs decide this valuation and I can’t know either. The first is how much of FICO’s revenue is mortgage origination, and how much of each $10 FICO keeps. FICO does not disclose either, and the $10 is a wholesale price to lenders through resellers.⁵¹⁷ If mortgage is a third of business-score revenue, a halving of it takes about 17% off B2B; if it is half, about 25% E (estimate). I don’t have a better basis than that arithmetic. The second is how lenders, mortgage bond investors and the bureaus act. Nobody has seen a year of one-grid pricing, and the company itself gave no guidance. So here is a range, not one number. Read my $617 with the grid beside it.

The discount rate

I discount the cash of the whole business at a weighted cost of capital of 9.0%. The ten-year Treasury yields 4.90% and Damodaran’s implied equity risk premium was 4.20% on 1 October.²⁹³⁰ My ten-year beta on month-end returns against the S&P 500 is 1.29, the five-year 1.35 and the three-year 1.70, which give costs of equity of 10.3%, 10.6% and 12.0% C (calculated).⁹³¹ I use 10.5% for equity. Debt costs about 6.1% before tax, 4.6% after, and is 26% of capital at market prices. That gives about 9.0% E (estimate). It is the rate on my Visa and Mastercard pages too, as a hurdle for large, established companies. I can defend 9% and I can defend 10%; the table shows what the difference does.

Three cases

The model has three revenue lines. Business scores start at about $1.42B and consumer scores at about $0.23B, both earning a 90.5% segment margin, in line with the 90% reported for nine months.³ Software starts at $0.88B and earns 27%, rising in the base to 33%. Unallocated corporate costs start at $190M, which is the figure that reconciles my starting point to fiscal 2026 operating income before the $65M and $27M one-offs, and rise 3% a year, less $60M a year from the workforce cut, which is my estimate because the company has not given one E (estimate).²¹³ Stock pay is a cost, at 6.8% of revenue in 2027 and 6.5% later in the base. Tax is 24%, net capital spending 0.8% of revenue and working capital 5% of added revenue. What is left is owner cash before interest, which I discount at 9.0% from a start 0.9 years out, with 3.0% growth after 2036. Then I subtract net debt.

My base fiscal 2027 operating income after stock pay is $1.42B against $1.36B of underlying fiscal 2026 E (estimate). That is up 5%, helped by lower costs and software, and with B2B revenue down 3%. There is no company guidance to check it against, and there is no consensus that I opened, so the near year is not tested.

Where does $2.9B of fiscal 2030 revenue come from? About $1.43B from business scores, which is about where it stands today, $1.18B from software, which would be 8% growth for four years, and the rest from consumer scores E (estimate). That is not a heroic path. It asks for software to do what its platform ARR already says, and for business scores to survive the grid.

The price sits 14% above my base, and that is the plain result. Why might the market be right? Because it can see what I can’t: the company holds the price at $10 and the lenders signing on, a program that analysts read as protecting volume, and it may know the shape of 2027 pricing that I do not.⁵¹⁹ The company said it has “several deals” announced under its Preferred Program, and that other programs are in progress for 2027.² Mizuho still expects limited near-term loss of FICO volume, though it now models flat pricing for 2027.¹⁹ If the market is right, my base is too pessimistic about mortgage pricing, and the bull of $971 is nearer the truth. I could be wrong in that direction as easily as the other.

Other uses of the same dollar

The ten-year Treasury pays 4.90%. At $701.03 my cash flows return 8.5% a year in the base, 6.6% in the bear and 10.6% in the bull, on the business before debt E (estimate).²⁹ The best comparison on this site is a toll business with fewer regulatory unknowns. Mastercard has a price 13% above my base and a return of 8.4% a year in the base case, and Visa a price 18% above its base. FICO’s premium to its base is 14%, so on my numbers the three look similar on price.¹ The difference is the question mark over the main source of profit. Visa and Mastercard have regulatory questions too, of another kind. Fannie Mae and Freddie Mac, whose rules created the FICO franchise, have their own pages here.

What past lows looked like

FICO has fallen more than 30% from a record close three times in ten years: 51% in February and March 2020, 38% from July 2021 to May 2022, and 75% from November 2024 to a close of $592.47 on 30 September 2026 C (calculated).⁹ The 2022 low was $341.44, about 24 times that year’s earnings of $14.18. The 30 September 2026 close of $592.47 was 16.3 times fiscal 2026 earnings. The same 24 times on this year’s GAAP earnings would be $875, which is above my base and shows what the market paid for a franchise without a regulatory question C (calculated). That is not a forecast. It tells me that the price of the stock has been a multiple of earnings that depends on belief about the toll, not on the earnings themselves.

The margin of safety

• Price$701.03Close, Q4 2026 [1]
• Base value$617My estimate E (estimate)
• Margin of safety below$442set from my bear case C (calculated)

I set the margin of safety from the bear case: at $442, my $332 bear value is at least three quarters of the price paid, and the price is 28% under my base value. I use the bear because the danger here is a change in the shape of the business, and a discount to the base value does not protect against that. At $442 the bear returns 8.0% and the base 10.5% a year E (estimate). The shares last closed below $442 on 7 November 2022, when operating income was $0.54B in fiscal 2022 against $1.27B this year.⁹²⁴² At $701.03 the price is 59% above that, so there is no margin of safety on my numbers. It does not cover the stress case either: business scores down 30%, 20% and 10% in each of three years and no software growth, which gives $127, a number that shows how much debt amplifies a franchise loss.

Two different questions

How good is the business, and how good is the price? They are separate. As a business FICO is extraordinary on most tests: 90% segment margins, 41% of revenue turned into free cash flow, a standard used by most US lenders, and a software arm that is finally growing. Its one real flaw is the one that sets the price: the largest part of its profit rests on a rule that has just been loosened. As for the price, at $701.03 it is 14% above my base value, which already allows for some mortgage erosion, and 59% above my margin-of-safety price. So this is a very good business at a price a little above my base value, with one question still open. I can’t tell you the answer to that question. Neither can the company.

VIIThe case for the price

What most people miss

Three things cut in FICO’s favour and get little attention. First, the software platform is no longer small. Its recurring revenue was $263.6M a year ago and is about $453M now, which is 53% of the software total and growing 72%, with net retention of 156% from existing customers.³² Second, the Preferred Program does more than keep a price. Enrolled lenders must pull FICO on every application, so ordering the cheap score first no longer saves them anything, and one reseller says lenders now plan to pull both.¹⁹ Third, the share count is falling fast: from 23.76M on 30 September 2025 to 21.60M on 30 June 2026, down 9.1% in nine months C (calculated).³

The bull case

In plain words: the toll is still being paid, the price is held, costs are shrinking, and software has turned. The company’s own version comes from Will Lansing, the chief executive. In the 4 November release he said, “I am proud of our performance in FY26,” and described the Preferred Program as having had “several deals already announced,” with more programs for 2027 in progress.²

Two sell-side houses have said things I can check. Mizuho wrote that the Preferred Program “is designed to neutralize” lenders ordering a cheaper score first, and that it expects limited near-term loss of volume while modelling flat 2027 pricing instead of 5% growth.¹⁹ UBS cut its price target to $972 from $1,100 and noted that the economics of the program do not need FHFA or GSE approval.¹⁹ Neither is a call that the stock will rise.

Against reality

Some of that case is already in the numbers: business-score revenue is up about 50%, free cash flow is 41% of revenue, and the shares are 70% off their high C (calculated).²⁹ The market has also shown how fast that can change: the shares fell 13% on 26 October after the regulator’s director posted that he was “taking more action this week.”²⁰⁹ My base has business-score revenue 3% lower in 2027 and 2% lower in 2028 than in 2026, and 2% to 4% growth after that, and gives $617. For the price of $701.03 to be right on my model, B2B revenue must grow about 3.5% a year from today with software on my base path C (calculated). That means FICO does the following: keeps $10 on nearly all mortgage volume through 2028, sees mortgage originations recover from below historical norms, and lets neither the bureaus nor VantageScore take a bite.²⁵ If FICO does that and software hits its targets, the price is possible, and the bull of $971 is the picture where it goes further.

It helps to multiply it out. My base needs four things: price held on most volume through 2028 (70%), mortgage volume at least flat by 2029 (65%), software growth of 7% or more for three years (70%), and cost cuts that stick (90%). Those are my guesses, and multiplied they give 29% E (estimate). The price also needs business-score growth above my base after 2028: say 60%, which leaves about 17% E (estimate). The items overlap, so true odds are somewhat higher. Even so, the price asks more of the future than my base does.

What a 10x would take. Ten times $701.03 is $7,010.30, about $151B of equity value on today’s diluted shares, or $156B with debt. At 25 times earnings, above the 19.3 the shares trade on, that needs about $6.0B of net income, 7.2 times the $0.83B of fiscal 2026 C (calculated).² At the 33% net margin FICO earned this year, that is about $18.4B of revenue, 7.2 times fiscal 2026, or growth of 22% a year for ten years against the 11.7% a year of the last nine. If buybacks retired 3% of the shares a year, the sums fall by about a quarter, to $13.5B of revenue. FICO’s own file shows a twentyfold rise, from $117.53 in November 2016 to $2,382.40 in November 2024.⁹ But that came from a company worth about $4B C (calculated), on mortgage price rises that the regulator has now opened to competition, and from here the same move would make FICO a $150B company with no new profit pool I can name. Tenfold from here is close to impossible, and I would not build a case on it.

Where I come down

The Preferred Program moved me: holding $10 and tying it to a pull on every application is a smarter answer than a cut, and it is why my base is not lower. The regulator’s public reply, and the 13% fall that followed, tell me he can undo it. The rebound of 19% from the 30 September low did not move me, because nothing changed in the rules. A year of business-score revenue holding at or above $1.42B, with the one-grid pricing in force, would raise my values most. It would not change the debt.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• Politics. The change came from one official. The April release says it advances the Credit Score Competition Act of 2018, so the direction has legal backing, but the timing and the grid came by announcement.⁷⁴ A different official, or a court, could move it either way, and FICO has no vote.
• Mortgage volume. The company says originations are “below historical norms” and rates are weighing on the market.² If volume stays low, price cannot make up the gap forever.
• The bureaus. They are 63% of revenue and they own VantageScore.³⁶ A fight over the royalty would not make headlines until it showed in results.
• Debt. $5.26B of debt, $1.76B of it due in mid-2028, at rates near 6%.²³ Covenants were met at June, and I have no sign of a problem, but the equity is small next to the debt.
• Litigation. A $65M accrual for a proposed settlement, and a claim that was in discovery as of June.²³
• Software and AI. The company is cutting 15% of positions partly to integrate AI into product development.¹³ I can’t tell how fast AI changes decision software, and it could cut either way.
• People. Lansing is retirement-eligible and has run the company for fourteen years.⁸ A change at the top in the middle of a pricing transition would add uncertainty.
• Numbers not yet audited. This page uses the fourth-quarter release. The 10-K can change figures, and I will say so if it does.²