How to Improve Your Credit Score, Ranked by What Actually Works

A credit score is not a report card on your character, and it is not a measure of how much money you have. It is a statistical estimate of one narrow thing: the probability that you will fall 90 or more days behind on a credit obligation in the next couple of years. Everything in the scoring model exists because it helped predict that outcome across millions of past borrowers.

Once you hold that idea, most of the advice sorts itself out. Some behaviors move the estimate a lot and quickly. Some move it a little and slowly. And a surprising amount of popular advice moves it in the wrong direction, or does nothing at all while costing you real money.

This guide ranks the levers by how much they can do and how fast they work, explains the mechanism behind each one, and then takes apart the five myths that cause the most avoidable damage.

#What the score is built from

The two dominant scoring families in the United States are FICO and VantageScore. Both currently run on a 300 to 850 scale, and both read the same raw material: the data in your credit files at Equifax, Experian and TransUnion. They weigh that data differently, which is why your numbers rarely match across sources.

FICO publishes approximate category weights, and those weights are the most useful public map of the system:

• Payment history, about 35%
• Amounts owed, about 30%
• Length of credit history, about 15%
• New credit, about 10%
• Credit mix, about 10%

Those percentages are averages across the whole scored population, not fixed rules for your file. If you have only two accounts, the age of them matters more than it would for someone with fifteen. If you have a recent missed payment, that single item crowds out almost everything else. The model is adaptive, which is why two people with identical utilization can see different responses to the same action.

#Payment history: why it dominates

Payment history is the largest single category because it is the closest thing in the file to a direct answer to the question being asked. A borrower who missed payments before is meaningfully more likely to miss them again.

What actually enters the file is narrower than people expect. Creditors generally report to the bureaus once a month, and most only report a payment as late once it is 30 days past the due date. Being five days late triggers a late fee from the card issuer and possibly a penalty rate, but it usually does not create a 30-day mark on the report. The card company's internal clock and the bureau's clock are different clocks.

Severity is graded. A 30-day late is bad, a 60-day is worse, a 90-day is worse again, and a charge-off or an account sent to collections is the bottom of the scale. Recency matters roughly as much as severity: a 30-day late from two months ago typically weighs more than a 90-day late from five years ago, because the model is trying to describe your behavior now.

A late payment stays on the report for seven years from the date it occurred. Its scoring impact fades well before that, usually beginning to soften after the first year or two and continuing to shrink as the item ages and as you add clean months on top of it. It does not vanish on a schedule you can predict, and nobody outside the model can tell you the exact figure.

Note: The single most protective habit is automating at least the minimum payment on every revolving account. You can always pay more manually. The automated minimum exists purely to make sure a distracted month never becomes a 30-day mark.

#Amounts owed: the fastest lever you have

Roughly 30% of a FICO score comes from amounts owed, and the biggest component of that is revolving utilization -- your reported credit card balances divided by your credit limits.

The critical mechanical detail is that utilization has no memory. The model reads the balance that appeared on your most recent report from each creditor, which for a credit card is usually the statement balance. It does not average your utilization over the year. It does not remember that you were at 80% last spring. Bring the reported figure down this month and the score generally responds on the next update.

That is what makes it the fastest lever. Almost nothing else in the file can change materially in 30 days.

Both overall utilization and individual-card utilization are read. One card sitting at 95% can hold a score down even when the portfolio total is modest. There is much more detail on the timing tricks and the per-card arithmetic in the guide to how credit utilization works.

Worked example: Suppose you carry three cards. Card A has a $2,000 limit and a $1,700 statement balance. Card B has a $5,000 limit and a $400 balance. Card C has a $3,000 limit and a $0 balance. Total balances are $2,100 against $10,000 of limits, so overall utilization is 2,100 / 10,000 = 21%. That looks fine. But Card A alone reports 1,700 / 2,000 = 85%, and that individual figure is also being read. Move $1,200 from Card A to Card B before the statement dates close. Card A now reports 500 / 2,000 = 25%, Card B reports 1,600 / 5,000 = 32%, and the overall figure is unchanged at 21%. You have paid down nothing and reduced no debt, but you have removed the maxed-out card signal.

#Length of credit history: slow, and easy to damage

This category looks at the age of your oldest account, the average age of all accounts, and how long each account has been active. It rewards nothing you can do today. It only rewards not disturbing what you already have.

The arithmetic is unforgiving in one specific way. Average age is a mean, so opening a new account drags it down in proportion to how few accounts you have.

Worked example: You hold three cards, opened 12 years, 8 years and 4 years ago. The average age is (12 + 8 + 4) / 3 = 8 years. Open a fourth card today and the average becomes (12 + 8 + 4 + 0) / 4 = 6 years, a 25% reduction in one afternoon. Now imagine the same new card added to a file with ten accounts averaging 8 years: (80 + 0) / 11 = 7.3 years, a 9% reduction. Same action, very different cost, and the deciding variable is simply how many accounts you already have.

Closing an old card does not remove it from the age calculation immediately. Closed accounts in good standing generally stay on the report for about ten years and keep contributing to history length while they remain. The real and immediate damage from closing a card is different: the limit disappears from the utilization denominator that same cycle.

#New credit and inquiries

Applying for credit creates a hard inquiry, which is visible to lenders and typically shaves a small amount off the score. Hard inquiries stay on the report for two years but generally stop being scored after one.

A soft inquiry -- checking your own score, a lender pre-screening you for an offer, an existing creditor reviewing your account -- is not visible to other lenders and does not affect the score at all.

Scoring models handle rate shopping deliberately. Multiple inquiries for the same type of installment loan, such as an auto loan or a mortgage, are grouped into a single inquiry if they land inside a defined window. Depending on the model version, that window is commonly 14 to 45 days, and newer FICO versions also ignore any such inquiries from the most recent 30 days entirely. The practical point is that comparison shopping for one loan does not cost you five times what one application costs. Credit card applications are not bundled this way; each is counted separately.

#Credit mix and authorized user accounts

Credit mix, about 10%, looks at whether your file contains both revolving accounts (credit cards) and installment accounts (auto loans, student loans, mortgages, credit-builder loans). Files with both tend to score somewhat higher than files with only one type.

This is the weakest justification for taking on debt in the entire system. Borrowing money you do not need in order to improve a 10% category is almost always a losing trade once interest is counted. If an installment loan is already in your future for other reasons, the mix benefit arrives free.

Being added as an authorized user on someone else's card places that account's history on your file at most bureaus, including its age, limit and payment record. It can help a thin file substantially. It also imports the primary holder's behavior, so a card that goes 60 days late lands on your report too. Some lenders and some model versions discount authorized user tradelines, particularly where they suspect the arrangement was purchased.

#Ranking the levers

The table below sorts the realistic actions by how much they can move a score and how quickly, assuming an ordinary file rather than an extreme one.

The pattern is consistent. Fast and large means utilization and disputes. Everything else is either prevention or patience.

#How long negatives last

Reporting periods are set by the Fair Credit Reporting Act, and the clock generally starts from the date of the original delinquency rather than from the date you paid.

The detail that matters most here is the date of first delinquency. Selling a debt from one collector to another does not restart it. If a collection account shows a start date that is later than the original default, the account may have been improperly re-aged, and that is a disputable error. The mechanics of that process are covered in the guide to disputing a credit report error.

#What the usual advice gets wrong

"Carrying a balance builds credit." It does not. The model reads the balance reported at statement time; it cannot see whether you paid that balance the following week. Paying the statement balance in full before the due date produces the same reported figure as carrying it, minus the interest. This myth is expensive in the most literal way.

"Checking your own score hurts it." Checking your own report or score is a soft inquiry. It is invisible to lenders and carries no scoring weight, no matter how often you do it. Free weekly reports from all three nationwide bureaus are available at AnnualCreditReport.com, and reading them is covered in how to read your credit report.

"Closing unused cards is tidy and helps." Closing a card removes its limit from the utilization denominator immediately, which mechanically raises your reported utilization even though your debt did not change. If a card has no fee and no temptation attached to it, leaving it open and used occasionally usually costs nothing.

"Paying a collection removes it." Payment changes the balance shown to zero and changes the status to paid. It does not delete the tradeline. Some newer scoring model versions ignore paid collections entirely, while older versions still in wide use at lenders do not. Whether payment helps your score therefore depends on which model version a given lender runs, which you cannot control.

"A credit repair company can remove accurate negative information." Nothing lawful can remove accurate, timely, verifiable information before its reporting period ends. What these firms actually do -- filing disputes -- you can do yourself for free, and a nonprofit credit counseling agency can help with the underlying debt without charging for deletions that cannot be delivered.

Warning: Any service that guarantees a specific score increase, offers a "new credit identity", or suggests using an Employer Identification Number in place of a Social Security number is describing something that ranges from useless to federal fraud.

#A realistic timeline

Improvement is not linear, and the starting point decides the shape of the curve.

If the only problem is high utilization on an otherwise clean file, the change can appear within one or two statement cycles. That is the best-case scenario in the entire system, and it is also the most common one.

If the file contains recent missed payments, the first year is mostly about establishing an unbroken run of on-time months. Movement tends to be slow at first and then accelerate as the negative items age and the positive history lengthens.

If the file contains a bankruptcy or foreclosure, the horizon is measured in years rather than months. People in that position often find that access to credit returns well before the score does, because manual underwriting and secured products look at the recent record rather than the number.

Two useful sanity checks along the way. First, the goal is qualification, not a trophy: past roughly the mid-700s, further points rarely change the terms you are offered. Second, a score is one input. Lenders also read income, employment and debt-to-income ratio, and a strong score does not compensate for a payment that does not fit the budget. If debt payoff is the underlying task, the sequencing question is handled in debt snowball versus avalanche and the tactics in how to pay off credit card debt.

#Edge cases worth naming

A thin file behaves differently from a damaged one. Someone with no accounts is not scored badly; they may not be scorable at all, which is a separate problem addressed in building credit from scratch.

An authorized user arrangement that ends badly can be unwound by the primary holder removing you, which typically removes the tradeline from your file, including whatever positive age it contributed.

A limit reduction imposed by an issuer raises your utilization without any action on your part, and it often happens precisely when other parts of your file already look strained.

Finally, deferment or forbearance on a federal student loan is not the same as delinquency. Accounts in an approved deferred status generally do not report as late, but the account still exists and still counts toward debt-to-income when you apply for something else. Understanding the cost side of any of this depends on knowing how APR is calculated, which is a different question from what the score says about you.