Credit scorecard (Scorecard)

What a scorecard does

A credit scorecard is a statistical model that turns information about an applicant or customer into a single score representing the likelihood of a particular outcome, most often the probability that they will repay as agreed. It takes a set of input variables, applies a weight to each, and combines them into a number the lender can rank and act on. A higher score typically indicates lower risk, and the lender sets cut-offs that translate scores into decisions such as accept, decline or refer.

The appeal of a scorecard is objectivity and consistency. Rather than relying on the varying judgement of individual underwriters, a scorecard applies the same logic to every case, which makes decisions faster, more comparable and easier to audit. It also lets a lender measure and tune its lending precisely, because the relationship between score and outcome can be tracked over time.

Inputs, weights and types

The variables that feed a scorecard are drawn from the data available at the point of decision. Application scorecards use information supplied at the point of applying together with credit reference data, covering factors such as credit history, existing commitments and public record information. Behavioural scorecards, used on existing customers, draw on how the account has actually been run, such as payment patterns and balance behaviour. Increasingly, affordability signals from Open Banking data, where the customer consents, can enrich the picture with a direct view of income and spending.

Each variable carries a weight derived from how strongly it predicted the outcome in the development data. The model is built on historical cases where the result is already known, so the weights reflect genuine statistical relationships rather than opinion. The output is a score, often mapped to an easily understood scale or to a set of risk grades.

• Application scorecards: applicant-supplied and credit reference data
• Behavioural scorecards: how an existing account is actually run
• Affordability inputs: consented Open Banking income and spending data

Governance, fairness and monitoring

A scorecard is not a build-once asset. Its predictive power decays as the economy, the customer base and borrowing behaviour shift, a phenomenon usually called drift. Lenders therefore monitor scorecards continuously, checking that the population still resembles the one the model was built on and that the score still separates good outcomes from bad as expected. When performance degrades, the scorecard is recalibrated or rebuilt.

Fairness and explainability matter as much as accuracy. Firms must be able to explain the main reasons for a decision, avoid using data that would produce unfair or discriminatory outcomes, and ensure the scorecard supports, rather than undermines, a proper creditworthiness and affordability assessment under CONC. The Consumer Duty raises the bar further, expecting firms to show that automated decisioning delivers good outcomes and does not cause foreseeable harm. A well-governed scorecard sits inside a broader decision framework, with policy rules, affordability checks and human referral all playing their part.