What it means
A credit score is produced by taking several factors, weighting them according to how well each predicts non-payment, and combining them into one figure. For businesses, typical factors include payment history, financial strength, leverage, trading history, profitability and sector risk.
For individuals the inputs differ, but the underlying logic is identical. The appeal is consistency and speed.
A scorecard applies the same rules to the tenth application of the day as to the first, removes the influence of who happens to be assessing the file, and lets a business automate low-risk decisions so that human attention goes where it is genuinely needed. It also makes decisions explainable, which matters when a customer asks why their limit is what it is.
Scores are usually mapped to bands rather than used as continuous numbers. A business might set standard 30-day terms and a full limit for scores above 80, half the limit and a quarterly review between 60 and 79, prepayment below 60, and automatic decline below 40.
The bands, not the score, are what actually drive behaviour. Scorecards need maintenance.
A model built on data from a benign period will underestimate risk in a harsher one, and factors that predicted default well five years ago may not do so now. Sensible practice is to track outcomes by band, checking that the businesses scoring 50 really do default more often than those scoring 80, and to recalibrate when they do not.
The main limitation is that scoring describes the average behaviour of similar cases rather than the case in front of you. It handles volume well and edge cases badly, so most credit policies allow a documented manual override for applicants whose circumstances the model was never built to capture, such as a newly incorporated subsidiary of a substantial parent.
In practice
Real-world examples
Example 1
An equipment hire company automates approvals for anything scoring above 75 with a requested limit under $20,000, which clears roughly 60% of applications the same day. The credit team's time shifts to the harder third of the book.
Example 2
A telecoms reseller notices that customers scoring between 55 and 65 are defaulting far more often than the model predicts. It recalibrates the sector risk weighting upwards for two industries where its own losses have concentrated.
Example 3
A packaging manufacturer receives an application from a newly formed subsidiary of a large listed group, which scores 22 because it has no history at all. The credit manager overrides the score on the strength of a parent company guarantee and documents the reasoning on the file.
Think of it
Credit scoring is very much like a video game reputation meter. As you complete tasks honestly and on time, your reputation goes up, unlocking better tools and trust. If you fail quests or miss deadlines, your reputation drops, and people hesitate to work with you.
Formula
Calculation
A weighted scorecard works as: total score = sum of (factor weight x factor score as a percentage of maximum), with the weights summing to 100. The resulting score is then mapped to a policy band that sets the limit and terms.
A wholesaler scores a prospective customer on five factors. Payment history carries a weight of 35 and the applicant scores 80% of maximum, giving 35 x 80% = 28. Financial strength and leverage carries a weight of 25 with a score of 60%, giving 25 x 60% = 15. Trading history carries a weight of 15 with a score of 100%, giving 15. Profitability carries a weight of 15 with a score of 40%, giving 15 x 40% = 6. Sector risk carries a weight of 10 with a score of 70%, giving 10 x 70% = 7.
The weights total 35 + 25 + 15 + 15 + 10 = 100, and the total score is 28 + 15 + 15 + 6 + 7 = 71.
The wholesaler's policy places 70 to 84 in band B: standard 30-day terms, a limit of $150,000 and a six-month review. Had profitability scored 80% instead of 40%, that factor would have contributed 15 x 80% = 12 rather than 6, lifting the total to 77, which stays in the same band and would not have changed the outcome. That is a useful check on which factors actually move a decision.Case study
Seen in the real world
Cobalt Industrial Supply is a fictional business used purely as an illustrative example of scoring done well and then done lazily. It built a five-factor scorecard, mapped it to four bands, and cut its average account-opening time from six days to one while write-offs stayed flat.
Three years later, in this illustrative account, write-offs had roughly doubled without any change to the scorecard. Reviewing outcomes by band showed the problem: applicants in band B were defaulting at almost the same rate as those in band C, because the payment history factor relied on data that had become considerably less current for smaller customers.
Cobalt reweighted the model, added a simple sector adjustment, and introduced a quarterly report tracking actual default rates against each band. The fictional lesson is that a scorecard is not a one-off build; without outcome tracking, a model quietly stops describing reality while continuing to produce confident-looking numbers.
Watch out
Common mistakes
- Treating the score as the decision. A score sets a starting point, but limits, security and terms should also reflect concentration, margin and the size of the order in front of you.
- Never testing the model against outcomes. If band B and band C default at similar rates, the scorecard has stopped discriminating and needs recalibrating.
- Refusing all manual overrides in the name of consistency. Some applicants, such as new subsidiaries of strong parents, are genuinely outside the model's experience, and a documented override is better than a bad automatic answer.
Questions
People also ask
Is business credit scoring the same as personal credit scoring?
The mechanics are similar but the inputs differ, with commercial models drawing on filed accounts and trade payment data rather than consumer borrowing history.
How often should a scorecard be reviewed?
Most organisations track band performance quarterly and recalibrate annually, or sooner if default rates move sharply or the customer mix changes.
Can a customer improve its score?
Yes, mainly by filing full accounts on time, paying suppliers within terms and reducing leverage, since those factors carry the heaviest weights in most commercial models.