Credit scoring

In plain words

Scoring is a way to make a lending decision quickly and consistently. Instead of a person reading every application, a model looks at the borrower’s data and gives a number. The higher the score, the lower the expected risk. The lender then sets a cut-off: above it the loan is approved, below it declined, and in between it goes to manual review or gets a smaller amount.

What goes into a score

Typical inputs are the credit history, current debts, income and its stability, the age of the borrower’s accounts, the number of recent applications and data from the application itself. Lenders also build behavioural scores for existing clients, based on how they paid earlier loans. Collection teams use a similar idea to rank overdue clients by the chance of payment.

Why it matters to an owner

The cut-off is a business choice, not a technical one. Lower it and the approval rate rises along with future overdue debt. Raise it and losses fall along with volume. An owner should see both sides at once: how many applications came in, how many were approved, and how those approved loans paid back several months later.

Common mistake

Treating the model as permanent. Scoring is trained on past borrowers, and when the market, the product or the traffic source changes, its accuracy drops quietly. A score that worked a year ago may now approve the wrong people. Models need regular checks against real repayment. In some countries borrowers have a right to know why they were declined, so rules on automated decisions are worth checking with a lawyer.