Scorecard
A deliberately simple, inspectable model that maps a handful of weighted characteristics to a score a reviewer can read and defend.
Category
How a model is chosen and adapted
Related terms
A scorecard maps a small set of characteristics to points, sums them, and produces a score. It is one of the oldest forms in credit and remains widely used because of a property that is hard to buy any other way: a person can read it.
Why simplicity is the feature
An underwriter can trace the score's contribution in the same representation used for customer explanation, internal challenge, and regulatory review. Each characteristic's contribution is visible. A monotonic relationship can be enforced so the model cannot reward something the institution considers worse. Band-level monitoring can make the location of drift easier to diagnose.
A more complex model may outperform a scorecard on a particular held-out dataset. Whether measured lift survives representative out-of-time and segment testing — and earns the added explanation, calibration, stability, monitoring, implementation, and validation burden — is an institutional decision, not a conclusion one metric can make.
Where it sits in a larger system
A scorecard is a quantitative component, not a decision. The score is one input alongside affordability calculations, collateral, policy limits, exception rules, and the approval authority that applies. A threshold alone does not release a decision; the rules outside the model decide what may happen next.
The machine-learning models page explains the full lifecycle, and the financial institutions & fintech page places scorecards beside other sector model families. Compare the application scorecard use case with SME credit referrals: the first builds and validates a risk estimate; the second assembles models, policy, evidence, and authority around one decision.
Often confused with
How it differs from similar terms.
| scorecard | Compared with | What's different |
|---|---|---|
| Scorecard | Machine-learning model | Depending on the data and intended use, a scorecard may give up predictive lift in exchange for inspectability, simpler monitoring, and a lower validation burden. |
| Score | Decision | A score is an input. Policy, limits, evidence, and approval authority decide the outcome. |
Want these terms worked through on one of your own decisions? Start with a Decision Blueprint