How a credit file is built, and what actually moves a score
What lenders can see, what they cannot, and which of the things people worry about matter.
The short answer
A credit file is a record of how you have handled borrowing, assembled by a credit reference agency from monthly reports sent in by lenders and a small number of public sources. A score is not part of that record and no lender is obliged to use it: it is a summary the agency calculates from the file, while lenders run their own scorecards on the same data. What moves it is dull and mechanical — whether payments arrived on time and how much of your available credit is in use — while salary, savings and much of what people worry about are not on the file at all.
What sits on a credit file
A credit file is narrower than many people expect. It holds four kinds of thing.
Identity and address history. Name, date of birth, current and previous addresses, and in some countries whether you appear on the electoral roll. It exists mainly so a lender can be confident the file belongs to the person on the form.
Credit accounts. Every account a reporting lender has opened in your name: the type, the date it opened, the credit limit or original amount borrowed, the balance outstanding, and a month-by-month payment status going back years. That last part is the spine of the file. Each month gets a marker — paid as agreed, one month late, two months late — and the markers stay visible long after the account closes.
Public records. Money judgments made by a court, insolvency proceedings, bankruptcy orders and similar formal decisions. These arrive from official registers rather than from lenders.
Searches and links. A log of who has looked at the file and why, plus any financial associations — the link created when you hold a joint account or joint borrowing with somebody else, which lets a lender see the other person's file when assessing yours.
What is not there matters just as much. A credit file does not hold your salary, employment history, current-account or savings balances, marital status, ethnicity, religion or nationality, medical history, or any criminal record other than money judgments. Nor does it record what you bought — a lender sees a $2,400 card balance, not whether that balance is groceries or a holiday. Whether tax debts, rent, utilities and telecoms appear varies by country and by agency, and in many systems only once something has gone seriously wrong.
Because there is no single national file, there is no national blacklist either. Nobody records a decision to refuse you credit. A declined application leaves a search footprint, not a verdict, and the next lender cannot see that the first one said no.
Who reports it, and how often
Reporting is voluntary in most markets, and each lender chooses which agencies it supplies: a card issuer might report to all of them, a small lender to one, some lenders to none. So you do not have one credit file. You have one per agency, and they rarely match.
Updates arrive in a monthly cycle tied to each account's statement date rather than to the calendar. A lender takes a snapshot on the statement date, batches it, and sends it on. The agency then loads it. From the moment something changes in the real world to the moment it appears on a file, an illustrative lag of two to six weeks is normal.
That lag explains a lot of confusion. Paying a card to zero the day after the statement is cut leaves the old, higher balance on the file for most of a month. A closed account can still show as open. A file is a stale photograph, not a live feed.
Agencies are obliged to correct genuine errors. A dispute raised with the agency is put to the lender that supplied the entry; if the lender confirms it, the entry generally stays. Where the facts are agreed but the context is missing, most systems allow a short written note on the file, which a human underwriter can read but an automated score ignores.
What a score is actually for
The number an agency shows you is a consumer-facing indicator. It is produced by that agency's own model, on that agency's own scale, from that agency's own copy of your data. Scales differ, so a figure from one agency cannot be read against a figure from another, and a gap between two of your own scores usually means the underlying files differ rather than that one is wrong.
Lenders mostly do not use it. A lender pulls the raw file into a scorecard built on its own past customers, alongside the application form and any history you already hold with it. A bank that has watched your current account for a decade knows things no agency does, which is why two lenders can reach opposite decisions on the same file in the same week.
There is also a second test that has nothing to do with scores. Creditworthiness asks whether you have repaid reliably; affordability asks whether the new repayment fits your income and outgoings. A strong file with thin income can fail on affordability, and no amount of score improvement changes that.
The factors that carry real weight
No agency publishes its model in full, and the weightings shift over time. But the broad ordering is consistent, and it is worth seeing the factors ranked rather than listed.
| Factor | Illustrative weight | What it measures |
|---|---|---|
| Payment history | 30–40% | Whether payments arrived on time, and how recently one did not |
| Amounts owed and utilisation | 25–35% | Balances against limits — $900 on a $1,000 limit reads differently from $900 on a $9,000 limit |
| Length of history | 10–20% | Age of the oldest account, and the average age of all of them |
| Recent searches and new accounts | 5–15% | How much new credit has been sought recently |
| Mix of account types | 5–10% | Evidence of handling more than one kind of borrowing |
Two of those deserve unpacking. Payment history dominates because a missed payment is the closest thing in the data to the event lenders are trying to predict. Recency matters within it: a late marker from last month weighs far more heavily than one from three years ago, even though both remain visible.
Utilisation is the percentage of available revolving credit in use, and it is the fastest-moving number on the file. General guidance often quotes a threshold below 30% as comfortable, which is illustrative rather than a rule — the models read utilisation as a slope, not a cliff. Because it is recalculated from each monthly snapshot, it can move a score within one cycle, which makes it the only major factor that responds quickly to anything.
What a search does, and does not do
Searches come in two forms, and conflating them causes needless worry.
A soft search is a look at the file that does not indicate an application: identity checks, eligibility checkers, quotation searches, account reviews by an existing lender, and your own inspection. Soft searches are recorded, but only you can see them, and they have no effect on scoring.
A hard search is recorded when you formally apply for credit. It is visible to other lenders, and a cluster in a short window is read as a signal — not because applying is bad, but because people applying repeatedly are sometimes being turned down. A single hard search typically has a small, short-lived effect; on illustrative timelines, hard searches are visible to lenders for around a year and sit on the file for roughly two.
Rate-checking tools that promise no impact are doing a soft search against a lender's rough criteria. The result is an indication, not a decision, and the full hard search still happens if you proceed.
How long adverse markers last
Adverse information does not stay forever, and it does not weigh the same throughout its life. Most systems set a fixed retention period, after which the entry is deleted outright. Long before that, its influence fades as newer months of good payment history accumulate.
| Entry | Illustrative time on file | When the weight fades |
|---|---|---|
| Late payment marker | Around 5–7 years | Most of the effect within 1–2 years |
| Default | Around 5–7 years from the date recorded | Gradually, and settling it does not remove it |
| Court money judgment | Around 5–7 years | Marked as satisfied once paid, which lenders read differently |
| Insolvency or bankruptcy | Around 6 years or longer, depending on the order | Slowly, and often noted separately by underwriters |
| Closed account in good standing | Around 5–7 years | Not adverse — it contributes history while it lasts |
| Hard search | Around 2 years | Most of the effect within 12 months |
One detail catches people out. Paying off a defaulted account changes its status from outstanding to satisfied, which many lenders treat as meaningfully better, but the default itself remains until its retention period expires. The clock runs from the date the default was recorded, not from the date it was paid.
Myths that persist
Checking your own file damages it. It does not. Your own inspection is a soft search that no lender sees.
Closing old accounts tidies the file. Closing an unused card removes its limit from the available-credit total, pushing utilisation up on what remains, and eventually removes an account that was contributing history.
Carrying a small balance helps. Paying a statement in full is reported as paid as agreed, which is the strongest available marker. Leaving a balance to show activity adds interest without adding data.
Being debt-free means a strong file. Never borrowing produces a thin file — little to predict from — which some lenders treat cautiously. That is not a poor file, but it is not a strong one either.
A partner's problems contaminate your file. Only a genuine financial association, such as a joint account or joint loan, links two files. Sharing an address does not.
Income raises a score. Income is not on the file. It is central to affordability and irrelevant to scoring.
Files can be inspected directly with each agency, and access is generally free or low-cost because most jurisdictions give people a right to see what is held about them. Since agencies hold different data, an error corrected with one is not corrected with the others.
What this means in practice
The mechanism points to a handful of considerations rather than a checklist.
- The file is the object of interest; the score is a summary of it. Changes flow from the file outward, which is why nothing moves until a lender's monthly report lands.
- Only one factor responds quickly. Utilisation can shift within a statement cycle; history length, the fading of a late marker and the ageing out of a default move on a timescale measured in years.
- Differences between agencies are expected. Three scores that disagree usually mean three files with different contents, and each can only be checked and corrected separately.
- A decline is not recorded and is not permanent, but the search that preceded it stays visible for a while, which is why the interval between applications carries information.
- Passing a scorecard and passing affordability are separate tests, and a clean file cannot close a gap between a repayment and an income.
Where a decision turns on the contents of a particular file — a marker that looks wrong, a default whose date is disputed, an application with a deadline attached — the details matter more than the general pattern, and that is a conversation for a qualified professional who can see your situation. The general pattern is simple enough: files strengthen through payments arriving on time, modest use of the credit available, and age.
