Every time you check your credit score you are reading the output of a company founded in 1956 by an engineer, Bill Fair, and a mathematician, Earl Isaac, who met at the Stanford Research Institute. The system they started has plenty of critics. It is also a vast improvement on what came before it, and 2026 has brought the biggest changes to how it is used in a generation. Here is the story so far.

Before the score

For most of banking history, a loan was a judgment call. A borrower sat across a desk from a loan officer, who looked at savings, past borrowing and income, and then at everything else: reputation around town, family, whether the applicant looked like the officer’s idea of a reliable person. That “everything else” is where the system failed. Two people with identical finances could get different answers depending on who they were and who they knew, and borrowers who did not fit the banker’s picture of a good risk — single women, minorities, newcomers — were routinely turned away regardless of the numbers. Decisions were slow, inconsistent from branch to branch, and impossible to audit.

1956: two researchers and a formula

Fair and Isaac’s idea was that repayment behavior could be predicted statistically from data a lender already had, and that a formula would do it more accurately and more consistently than a person. They founded Fair, Isaac and Company in 1956 and sold their first credit-scoring systems to lenders beginning in 1958. Adoption was gradual. Lenders had to be convinced that a model built from historical accounts could beat the experienced officer down the hall, and for decades the business was selling scoring systems directly to lenders rather than to the public.

1987 to 1989: going public and the general-purpose score

Two things happened in the late 1980s that turned a niche analytics firm into a household name. Fair, Isaac went public in 1987. And after a trial with the three national credit bureaus that ran from 1986 to 1988, the first general-purpose FICO score launched in 1989. The phrase “general-purpose” is the important part. Instead of a custom model scoring one bank’s applicants, this was a single score, calculated from bureau data, that any lender could pull on any consumer. Credit decisions got faster and more consistent, and a borrower’s standing became portable from one lender to the next.

2003: Fair Isaac Corporation

The company took the name Fair Isaac Corporation in 2003, and FICO went from being the name of a score to the name of the business. By then the score had the shape you know: a range of 300 to 850, driven by five factors — payment history (35%), amounts owed (30%), length of credit history (15%), new credit (10%) and credit mix (10%). Those weights are still how FICO describes the score today. The model has been re-estimated many times, but what it rewards has not changed: pay on time, keep card balances well below their limits, and let accounts age.

Versions and rivals

Success bred competition and a family tree. FICO has released new versions over the years and lenders are slow to change; FICO 8 is still the most widely used version in 2026, even though newer ones exist. The chief rival is VantageScore, which has its own formula. The practical result is that you do not have one credit score, you have many, and the number a lender sees depends on which model and which bureau it pulls. The free score in your banking app is a real score, but it may not be the one that prices your next loan.

2026: the mortgage market finally moves

For years, mortgage lending ran on older FICO versions. That changed on April 22, 2026, when Fannie Mae and Freddie Mac began accepting FICO Score 10T and VantageScore 4.0 from all lenders. The Federal Housing Finance Agency had approved VantageScore 4.0 in July 2025, and HUD is adopting both models for FHA loans. The switch is optional for lenders and the traditional tri-merge report from all three bureaus is still required, so the transition will be gradual. But both new models can read rent, utility and phone payments when they are reported to the bureaus, which opens the mortgage market a crack wider for borrowers with thin files.

A month later, on May 20, 2026, FICO made its next-generation UltraFICO Score generally available. Built with Plaid, it lets a consumer choose to share bank-account cash-flow data — inflows, balances, spending patterns — and layers that evidence onto the standard FICO scale. For someone with little borrowing history but a steady, well-managed checking account, it lets the score see proof that a traditional file cannot show.

What this history means for you

Three things. First, the score is a model of behavior, not a verdict on character, and it responds to the same handful of inputs it always has; the weights above are the whole playbook. Second, because there are many versions in circulation, mind the report that feeds all of them: it is free every week from each bureau at AnnualCreditReport.com, and an error there drags down every score built on top of it. Third, if you have a thin file, the 2026 changes widen the path for the bills you already pay; alternative credit scoring explains how to get rent and utility history in front of a lender. And if your score is healthy but stalled, the usual reasons a score will not budge are more mundane than the model’s history suggests. The mechanics of each factor are covered in how credit scores work.

Fair and Isaac set out to replace a handshake and a hunch with arithmetic. Seventy years on, the arithmetic is still being argued over, refined and expanded. The critics have a point. The alternative had a worse one.