Credit scores get treated as a general measure of financial trustworthiness. People talk about a good score the way they'd talk about a good character reference.

That's not what they measure. A credit score is a statistical prediction of one narrow thing, and understanding what that thing is explains most of the behaviour that otherwise seems arbitrary.

What it predicts

A credit score estimates the probability that a borrower will become seriously delinquent on a credit obligation within a defined future period, typically the next couple of years.

That's it. It doesn't measure wealth, income, savings, or whether you handle money sensibly. It measures likelihood of missing payments on borrowed money.

Which explains the most counterintuitive feature of the system: somebody who has never borrowed has no score, or a poor one. Not because they're a bad risk, but because there's no data from which to predict. A model with no observations cannot make a confident prediction, and lenders treat uncertainty as risk.

Similarly, someone with substantial savings and no credit history may score worse than someone with modest income who has serviced several loans reliably. From a prediction standpoint that's correct. From a common-sense standpoint it's obviously strange.

What actually moves it

The specific weightings differ between scoring models and countries, and the broad structure is consistent.

Payment history is the largest single factor. Missed payments, defaults and insolvency events matter more than anything else and persist for years.

Amounts owed relative to available credit — utilisation — is the next largest. Using a high proportion of available credit predicts difficulty, so it lowers the score even if you pay in full each month.

This produces the frequently misunderstood situation where paying off and closing a card can lower your score, because it reduces total available credit and raises your utilisation ratio on what remains.

Length of history matters, which is why closing your oldest account is generally a mistake.

New applications produce hard searches, which lower the score slightly and temporarily. Multiple applications in a short period look like distress.

Credit mix has a small effect — having managed different types of credit demonstrates more.

What doesn't affect it

Worth stating explicitly because the misconceptions are widespread.

Your income doesn't appear in most credit scores. Lenders consider it separately, and it isn't in the score itself.

Your savings and investments aren't in it.

Checking your own score doesn't affect it. That's a soft search and is invisible to lenders.

Your partner's credit doesn't affect yours unless you hold joint accounts or a financial association exists, which is worth checking if you've separated from someone.

And in most systems, day-to-day transactions, overdraft usage patterns and where you shop don't feed into the score, though lenders may see some of this separately when you apply.

The behaviours that backfire

Several perfectly sensible financial habits produce worse scores, and it's worth knowing which.

Avoiding credit entirely. Prudent, and it leaves you unscoreable when you eventually need a mortgage.

Paying off and closing accounts. Reduces available credit and history length.

Using a debit card exclusively. Generates no data.

Making a large purchase on a card and paying it immediately. Depending on when the balance is reported to the bureau, this can register as high utilisation even though you paid it the same week.

That last one is a genuine quirk: the reported balance is a snapshot on a particular date, not an average. Paying before the statement date rather than after can noticeably change the reported figure.

The errors problem

Credit files contain errors at a non-trivial rate, and studies examining consumer reports have found meaningful proportions containing inaccuracies, some serious enough to affect lending decisions.

Common problems: accounts that aren't yours, settled debts still showing as outstanding, incorrect addresses linking you to someone else's file, and duplicate entries.

You're generally entitled to access your file, and correcting errors is free. Given the effect a single incorrect default can have on borrowing costs over years, checking periodically is one of the higher-return uses of twenty minutes available.

The honest framing

A credit score is a tool built by lenders to serve lenders. It answers their question well.

It's become a general-purpose gatekeeping mechanism used for things it wasn't designed for — tenancy, some employment screening, insurance pricing in some markets — and its suitability for those uses is genuinely questionable.

The practical advice is to treat it as a game with published rules rather than as a verdict on your character. Hold an old account, use a small proportion of your limit, pay on time, don't apply for things you don't need, and check the file occasionally.

None of that will make you better with money. It will make the number go up, and the number is what the system reads.

Building one from nothing

Practical guidance for the situation the system handles worst: having no history at all, whether because you are young, newly arrived in a country, or have simply never borrowed.

The standard route is a small amount of credit used lightly and repaid in full — a low-limit card used for a regular small expense, paid off each month. It generates data without generating cost.

Registering on the electoral roll, where that exists, is a surprisingly significant factor in some countries because it confirms identity and address stability.

And time simply matters. A file needs months of history before it produces a useful score, which means anyone planning a mortgage application should start considerably earlier than feels necessary. Six months is the minimum worth having; a couple of years is better.