A credit score is a probability estimate: how likely you are to be 90 days late on an obligation in the next two years. Once you see it as a risk model rather than a grade, the levers become obvious — and the myths become easy to discard.
What each factor is actually worth
| Factor | Approx. weight | Speed of change | Your leverage |
|---|---|---|---|
| Payment history | ~35% | Slow to repair | High going forward |
| Amounts owed / utilisation | ~30% | One billing cycle | Very high |
| Length of credit history | ~15% | Years | Low — protect it |
| Credit mix | ~10% | Months | Moderate |
| New credit / inquiries | ~10% | 3–12 months | High — just wait |
The utilisation lever, done correctly
Utilisation is your reported balance divided by your limit. Critically, the balance reported to bureaus is usually the one on your statement closing date — not what you owe after paying in full. A cardholder who pays every bill on time and in full can still report 70% utilisation and take a 60-point hit.
The mid-cycle paymentMake a payment three to five days before your statement closes, then pay any remainder by the due date. Same money, same month — but the reported balance drops immediately. Single-digit utilisation optimises the score; zero across every card is very slightly worse than a small reported balance.
The 90-day plan
- Days 1–7: pull all three reports and auditGet the full file from each bureau. List every account, balance, limit and negative mark. Errors appear in a meaningful share of reports.
- Days 7–14: dispute what is wrongFile disputes in writing with supporting documents. The bureau must investigate and remove anything it cannot verify. Keep copies of everything.
- Days 14–30: request limit increasesAsk each issuer for an increase, ideally via soft pull. Higher limits cut utilisation without repaying a cent.
- Days 30–45: fix statement timingNote each card's closing date and schedule mid-cycle payments. This alone often produces the largest single jump.
- Days 45–60: address negativesBring any past-due account current immediately — ageing delinquency does more damage each month. For a first-time late payment on an otherwise clean account, request a goodwill adjustment.
- Days 60–90: hold steadyOpen nothing new, close nothing old, keep utilisation single-digit, and let the file re-report. Bureau data updates monthly, so patience is part of the method.
Myths that cost people points
"Carry a small balance to build credit"False, and expensive. Interest does not improve your score. Use the card, then pay it in full — reported utilisation is what matters, not whether you paid interest.
"Close cards you don't use"Usually harmful. Closing removes available credit, raising utilisation immediately, and eventually shortens average account age. If a fee is the issue, ask to downgrade to a no-fee version of the same account.
"Checking my score hurts it"No. Checking your own score is a soft inquiry. Only applications for new credit create hard inquiries, and each costs a handful of points for a few months.
Rate shopping without damage
Scoring models treat multiple inquiries for the same loan type inside a short window — commonly 14 to 45 days depending on the model — as a single event. This exists precisely so consumers can compare offers. Do all your mortgage or auto shopping inside two weeks and the impact is one inquiry, not seven. That single behaviour can be worth more than the score points, because a better rate compounds over the whole loan. Our refinance decision tree shows how much.
Thin files and rebuilding
- Secured card. A refundable deposit becomes your limit. Use it for one small recurring charge and autopay it in full.
- Authorised user status. Being added to a long-standing, low-utilisation account can import positive history. Confirm the issuer reports authorised users.
- Credit-builder loan. Payments are held in an account and released at the end. Builds instalment history without real borrowing.
- Utility and rent reporting. Several services report on-time payments you are already making.
Limits, errors and borders of the standard advice
Critical reviewCredit-score advice is dominated by folklore and by companies selling monitoring, cards and repair services. Much of what circulates is a rule of thumb that was true for one scoring model in one decade, repeated as though the algorithms had never changed.
"Keep utilisation under 30%."
The claimThirty per cent is the threshold below which utilisation stops hurting your score.
Where it breaksThere is no cliff at thirty. Utilisation is scored continuously, so lower is better all the way down, and the widely repeated threshold seems to have originated as an informal simplification rather than a model parameter. The rule also obscures two mechanics that matter more: utilisation is usually measured on the statement date, not the due date, so someone who pays in full monthly can still report high utilisation; and per-card utilisation is assessed alongside the aggregate, so one maxed card can hurt while your overall figure looks comfortable. Utilisation is also almost entirely a snapshot with no memory, which cuts both ways — it recovers fast and it can be gamed just before an application.
The borderUnder thirty is a useful floor for someone currently well above it. If you are optimising for a mortgage application, aim far lower, check the statement-date timing, and fix the worst individual card rather than the average.
"Carry a small balance to build credit."
The claimPaying interest on a small revolving balance demonstrates creditworthiness.
Where it breaksStraightforwardly false, and unusually persistent. Scoring models reward on-time payments and low utilisation; nothing rewards paying interest. What is true — and probably the origin of the myth — is that a card reporting zero activity across the board contributes less than one showing regular use and full repayment. That distinction became "carry a balance", which benefits only the lender. The myth is also expensive at compounding rates, and it discourages exactly the full-payment habit that produces the best outcomes.
The borderUse the card regularly and pay the statement balance in full. If you want reported activity, ensure a small balance exists on the statement date and then clear it — never carry interest-bearing debt for scoring purposes.
"Closing an unused card is good housekeeping."
The claimFewer cards means less risk and a tidier file.
Where it breaksClosing a card removes its limit from your aggregate, which raises utilisation immediately, and it eventually removes its history, which can shorten average account age. Both effects push the wrong way. But the reverse advice — never close anything — is also overstated: closed accounts in good standing typically remain on the file for years, an annual fee on a card you never use is a certain cost against an uncertain scoring benefit, and an unused open card is an unmonitored fraud surface.
The borderKeep no-fee, long-held cards open with occasional small use. Close a fee-paying card you do not value, ideally after asking to downgrade it to a no-fee product in the same family, which keeps the account age and the limit. Never close anything in the months before a major credit application.
"Checking your credit hurts your score."
The claimCredit checks damage your score, so look as rarely as possible.
Where it breaksIt conflates two different things. Checking your own report is a soft enquiry and has no effect at all. Applications generate hard enquiries, which have a small and temporary effect — and even there, rate-shopping for a single loan within a short window is typically consolidated into one enquiry, so the fear of comparison shopping is largely misplaced. The myth's real cost is that it discourages people from reading their own file, which is where errors and early signs of identity fraud are found.
The borderCheck your own reports often and at every bureau; it is free and it is the main defence against errors. Be deliberate about applications, and concentrate loan shopping into a short window so the enquiries group.
We describe mechanics common to mainstream scoring models without naming versions, because lenders use different models and generations of the same model simultaneously — which means we cannot tell you which score a specific lender will pull or how it weights your file. Scoring systems also differ fundamentally between countries, so readers outside the model we describe should treat the structure as illustrative. And the whole framework optimises a number, when the thing that actually matters is the credit you are approved for at the price you are offered; the two correlate strongly but not perfectly.
Frequently asked questions
How fast can a score realistically improve?
Utilisation changes can show up within one billing cycle — 20 to 60 points is common for someone moving from high to single-digit utilisation. Repairing missed payments takes far longer, because the model is measuring consistency over time.
Do credit repair companies work?
They perform disputes you can file yourself for free. Nobody can lawfully remove accurate, timely negative information. Be sceptical of any firm promising deletion of legitimate items or charging large upfront fees.
Why do my scores differ across apps?
Different scoring models and different bureau data. Lenders also use industry-specific versions — an auto lender may see a different number than the one in your banking app. Track direction over time rather than obsessing over a single figure.
Does paying off a loan early lower my score?
It can dip slightly if it was your only instalment account, because credit mix narrows. That is not a reason to keep paying interest — a few points are worth far less than the interest saved.
The bottom line
Pay every obligation on time without exception, keep reported utilisation in single digits by paying before the statement closes, protect your oldest accounts, dispute genuine errors with documentation, and cluster rate shopping into a two-week window. Those five behaviours account for nearly all of the score movement available to you — and none of them require paying anyone for help.
Revision log. Jul 2026: added utilisation model and rewrote the 90-day sequence. Reviewed twice yearly.