What is the difference between FICO and VantageScore?

At a glance#

Why two score families exist at all#

A credit score is a product, not an official figure. A model vendor builds a formula that turns the data in a credit report into a risk estimate, and sells that formula to lenders. Anyone who believes a better formula is possible can build one and sell it too.

The US market developed along those lines. The older family, FICO, comes from a specialist analytics firm and became the standard for lending decisions over several decades. The younger family, VantageScore, was developed jointly by the three large credit bureaus, partly so as not to depend on a single supplier forever.

For an applicant that leaves an awkward truth. There is no single score, only a handful of numbers that exist at the same time, and which of them counts in a particular application is decided by the lender alone. Newcomers from Germany, Austria, Switzerland or elsewhere often expect one national score, as with a single credit agency at home, and the US system does not work that way.

How the two families differ#

Why lenders mostly decide on FICO versions#

FICO models are widespread in lending decisions and exist in several versions side by side. On top of those come industry versions tailored to particular products, such as car loans or credit cards. The variety of versions is part of the business model. A lender buys the version that fits its risk policy and does not switch with every release. One lender can therefore work with a version that is several years old while another uses the newest. Even within the FICO family your number is not unique. The number depends on the version, and the version is not named on any application form.

Why free apps usually show a VantageScore#

VantageScore is common wherever consumers see their score for free: in banking apps, card portals and comparison services. The reason is commercial, because licensing for display purposes works differently from licensing for lending decisions. The number you check every week in an app is therefore often not the number your next lender will look at. The app number can be higher or lower and can react differently to the same event. A free VantageScore display is still useful for one thing: it reliably shows the direction. When utilization rises or a payment is missed, the app number falls too. As an early warning signal the free display works; as a forecast of an approval decision it does not.

Both families calculate from the same report data#

However different the formulas are, FICO and VantageScore use identical raw material: the entries in the credit report. Both families assess payment history, amounts owed, age of accounts, new accounts together with inquiries, and the mix of account types. The shared data source shows where change is possible. The score itself cannot be negotiated; the report can. Improving the report improves every number calculated from it, regardless of family and version. How that report comes into existence is covered in How do I build a US credit report from scratch?.

Why one person has several scores#

Three causes produce differences, and each works independently of the others.

The bureau. Each of the three large bureaus holds its own data. An account reported to only two of them is missing from the third file, and a model can only assess what is in the file it receives.

The model family. Two vendors, two formulas. The same data leads to different results, especially with short histories and with special cases such as collections or old negative entries.

The version. Within one family, different versions treat the same situation differently. A newer version may, for example, ignore paid collection entries that an older one still counts.

Of the four, only the bureau files and the reporting date are in your hands. The choice of family and version is always made by the lender, and no preparation changes it. How the reporting date works is covered in Statement date versus due date.

Worked example: one newcomer, three bureau files#

Somebody moved to the US about a year ago and now holds two credit accounts, both run without a late payment. Account A is a credit card reported to all three bureaus, with a low balance. Account B is a second card from a smaller lender that reports to a single bureau only, and it was reported with a high balance in its last cycle because the statement date fell before the payment.

A lender pulling Bureau 1 sees two accounts and a high reported balance. A lender pulling Bureau 2 or Bureau 3 sees a single account with low utilization, which is a thinner file but a calmer one. Every one of these numbers is calculated correctly, and they differ without anybody having made a mistake.

The gap between the three results comes from the data, and the model has nothing to do with it. Faced with gaps like these, look first for the account that is missing or reported with an unusual balance, and only then at which model was used.

What this means for an application#

Never plan around a particular number. An application timed on the app showing a particular value is planned on a number the lender may never see.

Thinking in states works better. A report without late payments, with low credit utilization and with several months of reported history is a good report in every family and every version. A report with an open late payment is good in none of them.

Timing matters more than the display. The reported balance depends on the statement date, so the number shifts with the date on which the card was paid. Keeping the balance low in the cycles before an application improves the reported state without manipulating anything.

So does the order of applications. Inquiries and new accounts share the smallest factor group, which the largest vendor weights at ten percent for the general population, so a single hard pull moves little. What counts is the pattern of several applications in a short period, and every lender sees that pattern regardless of the model. Some card lenders add their own limits on new accounts, the best known of which is covered in What the 5/24 rule means.

What neither model can see#

A score assesses only what is in the credit report. Four things that a person would consider obviously relevant are missing from it entirely, and their absence explains many rejections that otherwise look incomprehensible.

Income. Salary and wealth appear in no credit report and feed into no score. A lender therefore asks about income separately in the application and decides on both together.

Savings. A large deposit balance at the same bank does not improve the score. The deposit balance can influence the lender’s decision, because the lender sees the relationship, but the number itself stays where it is.

Rent. Rent payments appear only if somebody reports them explicitly, and that is the exception. Years of paying rent on time leave no trace in the file unless someone reports them.

The reason behind a missed payment. A report does not distinguish between a forgotten payment and a disputed bill. Both show up as a late payment, and clearing it up runs through the dispute procedure, not through an explanation to the model.

How to handle the numbers in practice#

One source, regularly. Pick one free display and look at it at fixed intervals, for example monthly. Compare its course with itself, not with other displays. A jump without an obvious cause is a reason to check the reports.

All reports, less often. The reports themselves are what matters. Reading them through once a quarter is enough, and pulling them costs nothing. Look for missing accounts, wrong limits, unknown inquiries and old entries that should have dropped off long ago. How limits are set and changed is covered in What a credit limit is and how it changes.

Before an application, deliberately. A month or so before a planned application a closer look pays off: keep balances low, open no further accounts, dispute open errors. The order of applications belongs in the same session, because the order has more effect than any fine tuning of the number.

With these three habits the question of the model family almost never needs answering. The question only arises when a specific rejection needs explaining, and in that case the lender sends the reasons and the source it used in writing anyway.

When comparing the models is not worth it#

Four situations show when it is not worth comparing the models at all:

• An empty file. Without enough reported data, neither model returns a number. Comparing the two families is pointless, because both say the same thing, which is nothing.
• No application coming up. Without a concrete plan, the score is a number without a use. Watching the number is pleasant and changes nothing.
• Open arrears or expensive balances. Anyone paying interest or owing a payment has a money problem, not a model problem. Every minute spent on comparing models goes to the wrong place.
• No US account yet. This applies to newcomers from abroad in particular. Without a single reported US account, what is needed is an account, and knowledge of scoring models does not help. Which identification number opens that first account is covered in ITIN or SSN for US credit cards.

The most common mistakes#

• Treating the app number as the application number. The app shows a direction. The lender decides on a different number, often from a different family.

• Comparing two sources with each other. Two models at one moment say nothing about each other; one source followed over time says a good deal.

• Paying for a score. The reports are free, and they contain the data that can be changed.

• Checking only one bureau. Three files mean three possible errors, and a missing account only shows up in the file where it is missing.

• Optimising towards a round number. Lenders work with their own thresholds and their own rules, so no particular value guarantees approval.

• Putting the model question before the account question. Accounts that are reported correctly come first. The finer points of the calculation can wait.

A note on the shared score range#

Both families use the same number range in their current versions, and that is precisely what invites the wrong conclusion. The same scale does not mean the same meaning. A value near the top signals low risk in both families, but the threshold at which a given lender approves is not defined by the scale.

Lenders do not publish their thresholds, and they move them with their appetite for risk. What was enough for approval one year can fall short the next without any change in your own file, so there is no magic number to search for. Two families, several versions and three bureaus produce the contradictory figures that puzzle applicants, yet every model reads the same raw data. Paying on time, keeping utilization low and giving the report time puts a file in good shape in every version, whichever family a lender buys.

Related topics#

• What the 5/24 rule means
• What a credit limit is and how it changes
• Business credit versus personal credit
• ITIN or SSN for US credit cards
• Hard pull versus soft pull
• How US credit reports work