Quant Memo
Core

Credit Scorecards and Internal Rating Models

Banks don't wait for a public rating agency to grade every borrower — they build their own scorecards, weighting financial and qualitative factors into a single score that maps to an internal rating grade.

Prerequisites: Credit Ratings and the Agencies, Leverage Ratios and Interest Coverage

Most companies in the world are never rated by Moody's or S&P — they're too small, too private, or in a market those agencies don't cover. Yet banks still need to lend to them, price the risk, and set loan terms. The tool built to fill that gap is the credit scorecard: an internal, standardized model that turns a borrower's financials and qualitative traits into a single number, and that number into a rating grade.

A credit scorecard assigns points to a set of financial ratios and qualitative factors (leverage, coverage, industry, management quality, and more), weights and sums them into a total score, and maps that score to an internal rating grade with an associated probability of default. It's how banks rate the thousands of borrowers no public agency ever will.

Building the score

A typical scorecard combines quantitative factors — leverage, interest coverage, profitability margins, revenue size — with qualitative overlays like industry cyclicality, management track record, and competitive position. Each factor is scored on a sub-scale (say, 0 to 20 points) and multiplied by a weight reflecting how predictive that factor has historically been of default, then summed into a total score out of 100. Banks calibrate these weights by backtesting the scorecard against years of historical defaults, checking that borrowers who actually defaulted scored systematically lower than those who didn't.

Total Score=iwi×si\text{Total Score} = \sum_i w_i \times s_i

In words: each factor's raw sub-score is multiplied by its weight, and the weighted sub-scores are added up into one final number, which is then mapped onto a fixed rating scale, such as internal grades 1 through 10.

Worked example

A mid-sized manufacturer is scored on three factors: leverage (weight 40%, sub-score 60/100 — leverage is moderate), interest coverage (weight 35%, sub-score 75/100 — comfortably covered), and industry cyclicality (weight 25%, sub-score 50/100 — a cyclical sector).

  1. Leverage contribution. 0.40×60=240.40 \times 60 = 24.
  2. Coverage contribution. 0.35×75=26.250.35 \times 75 = 26.25.
  3. Industry contribution. 0.25×50=12.50.25 \times 50 = 12.5.
  4. Total score. 24+26.25+12.5=62.7524 + 26.25 + 12.5 = 62.75 out of 100.

If the bank's internal scale maps scores of 60–70 to internal grade "BB-equivalent," this borrower lands solidly in that band, which in turn maps to a pre-calibrated one-year probability of default the bank uses for loan pricing and capital allocation.

leverage: 24 coverage: 26.25 industry: 12.5 score = 62.75 maps to internal grade "BB-equivalent"
Weighted sub-scores sum to a single number, which the bank's mapping table converts into a rating grade and default probability.

What this means in practice

Internal scorecards let banks lend to the vast unrated middle market on a consistent, auditable basis, and regulators require validated internal rating models under frameworks like Basel's internal-ratings-based approach before a bank can use its own estimates for regulatory capital.

A scorecard's weights are fit to historical data, which means it can systematically misjudge borrowers whose risk profile doesn't resemble the historical sample it was calibrated on — a fast-growing tech company scored on a model built from stable industrial defaults, for instance, may get a misleadingly comfortable or harsh score.

Related concepts

Practice in interviews

Further reading

  • de Servigny and Renault, Measuring and Managing Credit Risk (ch. on scoring models)
ShareTwitterLinkedIn