Fiscora's credit guides will tell you what moves a score: utilisation, payment history, file age, mix, inquiries. That advice is accurate and worth following. It is also, in a specific sense, beside the point — it teaches you to be legible to a system without teaching you what the system is.
This article is the other half. It is about the apparatus itself: what kind of decision it makes, why that decision resists challenge, and what standing you actually have when it goes against you.
What kind of thing a credit score is
A credit score is not a fact about you. It is the output of a proprietary statistical model, run on a file assembled by a private company, predicting one narrow thing — the probability that a lender loses money — and doing so on the basis of a population, not a person. It is a forecast about people who resemble you in the dimensions the model records.
Everything difficult about the category follows from that. A forecast is not wrong when it fails in an individual case; it is doing what forecasts do. But it is used as though it were a judgement about your character, and it carries consequences — housing, credit, sometimes employment and insurance pricing — that a probability estimate is not equipped to carry.
The register that governs this publication names the resulting harm algorithmic governance without due process: consequential decisions become unappealable because they became automatic. The causation in that sentence is the whole argument. Nobody removed your right to be heard. It simply became structurally impossible to exercise once the decision stopped passing through a person.
The four elements, applied
The governing instrument treats a determination as crossing an absolute prohibition when four things hold together. Applied to credit:
| Element | Credit scoring |
|---|---|
| Material consequence | Present. Housing, credit, sometimes employment and insurance. Squarely within the definition. |
| Automated determination | Usually present. Most decisions below a threshold are decided without a human reading the file. |
| No specific reasons | Usually present. Adverse-action notices name factor categories — "level of indebtedness" — not the specific fact and its weight. |
| No appeal to a human with reversal authority | Frequently present. You can dispute file data. You typically cannot appeal the score or the lender's decision to a person empowered to overturn it. |
Our reading is that mainstream credit scoring sits at or across this line more often than the industry acknowledges, and that the fourth element is the one most often technically satisfied and practically absent. A dispute process that can correct a wrong date but cannot revisit a decision is a data-hygiene mechanism, not an appeal.
Where you actually have leverage: the file, not the score
This is the most useful practical distinction in the category. You have weak rights over a proprietary model and considerably stronger rights over the data it consumes.
- Obtain the full file from every bureau operating in your jurisdictionNot the score — the file. Errors are common, and an error in the file is the one thing everyone agrees must be fixed.
- Dispute inaccuracies in writing, with evidence, keeping copiesA dated written dispute creates an obligation and a record. Verbal disputes do neither.
- Ask the lender for the specific factor, not the category"Which specific entry, and what weight did it carry?" You will often be refused; the refusal is itself a fact worth holding.
- Where a right to human review exists, invoke it by nameSome jurisdictions provide it for automated decisions with legal or similarly significant effects. Naming the right changes who handles the request.
- Re-check sixty days after any correctionCorrections are frequently reversed when the original furnisher re-reports. Verify rather than assume.
The inference layer you cannot seeBeyond the file sits a layer of derived inferences — stability, reliability, propensity — built from data you never supplied for that purpose. You usually cannot see these, cannot correct them, and cannot refuse them without losing access to the financial system. That combination is precisely what the register means by the human as product.
Exclusion by default: when no data is read as bad data
A thin file is not a neutral starting position. Scoring models treat the absence of a credit history as an absence of evidence of repayment, which functions as adverse. The people affected are recent immigrants, young adults, people leaving abusive financial situations, and people who avoided debt deliberately — including on religious grounds.
The register calls this exclusion by default: a provision removes a right from those who cannot reach it. It is worth being precise about the unfairness. The system is not making an error about these people. It is correctly reporting that it has no information, and the market has decided that no information means high risk. The defect is in the inference the market draws, not in the arithmetic.
Practical consequences for someone in that position:
- Building a file is a separate project from managing money well, and requires deliberately using instruments you may not otherwise want.
- Alternative-data products that score rent and utility payments help and come at a cost: they widen what is observed about you.
- The trade-off is genuine and personal. Financial legibility purchased with surveillance is a real exchange, not a free improvement.
The instrument beneath the score
There is a question this publication brackets in its credit guides and states plainly here: the product being scored is a loan at interest — a stipulated increase over principal, accruing to the lender as a condition of the deferral itself. For a substantial number of readers, that structure is not merely expensive but prohibited.
The licence governing this publication prohibits originating, servicing, pricing, brokering, securitising or automating such a transaction. Explaining one is not among those verbs, and general-purpose analysis is expressly outside the prohibition — so Fiscora's guides are within bounds. But being within bounds is not the same as having addressed the question.
The honest position: most readers hold these instruments and cannot exit them this year. Helping them pay less is a genuine benefit to a harmed party. That is harm reduction, and it is a defensible thing to do — provided it is not mistaken for an answer to the underlying question. The alternatives that are permitted under the same instrument — genuine profit-and-loss sharing, lease and sale structures carrying real asset risk, fee-for-service pricing, cost recovery at zero markup — are a separate body of knowledge, and one this estate intends to build out rather than gesture at.
What "improve your score" gets wrong
Not the advice — the framing. Three specific problems:
It presents a forecast as a verdict. Score-improvement advice teaches you to optimise a number as though the number measured your worth. It measures a modelled probability of lender loss. Internalising it as a personal grade is a category error with real psychological cost.
It makes the apparatus invisible by making it navigable. This is the counterproductivity problem in miniature: advice good enough to work removes the pressure that would otherwise be directed at the mechanism. The better the workaround, the longer the defect survives.
It individualises a structural position. Told to raise your score, you infer the score is a fact about your behaviour. For many people it is substantially a fact about their circumstances — where they live, what they inherited, whether anyone could co-sign.
Both things at onceFollow the score advice; it works and the consequences are real. And hold the second thought alongside it: you are complying with a mechanism, not being measured by a truth. Those are compatible positions, and holding both is the actual competence.
The freeze, and what it tells you about the system
A credit freeze prevents new accounts being opened in your name. It is free, does not affect your score, and works — because a lender who cannot pull your file will not approve an application.
Worth noticing what the freeze reveals. Its effectiveness depends entirely on the bureau's file being the gate through which all credit passes. The tool that protects you is the same concentration that makes you vulnerable: one private file, held by a handful of companies, determining access to the financial system. The register calls that centralisation and enclosure, and the freeze is best understood as a control you have been given over an apparatus you did not consent to.
Honest limits
Understanding does not improve access. Nothing here will get you a better rate. The score advice does that. This article gives you an accurate account of what is happening to you, which is worth having and is not a substitute.
Rights vary sharply by jurisdiction. The right to reasons, to human review, to free file access and to dispute resolution differ enormously between countries and sometimes regions. We describe a structure and questions that generalise; the enforcement behind them does not.
Some scoring is defensible. A steelman our credit guides would accept: statistical underwriting expanded access enormously compared to the relationship banking it replaced, which was discretionary and discriminated more or less openly. Automation removed a great deal of arbitrary human prejudice. That is a real gain, and the argument here is for reasons and appeals — not for returning to a bank manager's judgement.
Revision log. Sep 2026: first publication. Written as the mastery track for the credit desk's structural issues, referenced from the "More Details" section of Fiscora's credit guides. Reviewed annually.