Scorecarding was an approach used by our partner PiggyBank, and involves assigning a score - similar to that of a standard credit score - to an application in order to determine the level of risk.

PiggyBank used a standard 'scorecard' approach when assessing whether lending to an applicant represents an acceptable risk. A scorecard gives a score to an application - the higher the score, the lower risk the business considers the applicant to be. Each scorecard is built up from dozens of individual data points, taken from the application, and the applicant's credit file. For example, owning a credit card may be worth 50 points; having a verified previous address, 30 points; having had a recent defaulted store card may subtract 60 points, etc.

Scoring applications in this way has been standard practice across the industry for decades.

Originally, this was because scores had to be calculated by hand, so the process had to be simple. But even with the advent of powerful computers and more nuanced methods of assessing 'good' or 'bad' risk, the market still prefer scorecards; they are straightforward to build, and the resulting decisions are easily understood.

This is generally not true for more modern approaches of decisioning, e.g., neural networks: the underlying reasons for a decision may not be fully understood, even by the person who built it. It is true that machine learning (which includes neural networks) may make better lending decisions in the long run, but a good scorecard should be accurate enough to make the trade-off (better understanding versus a marginal drop in efficiency) acceptable.

The technical team behind the business’ products and services can help you build your own scorecard using the business’ generic model as a starting point.