Every number in Risk Analytics starts here. A grade for the borrower. A probability of default where the data permits.
Three model families do the work. Heuristic, where credit-expert judgment carries the weight. Statistical, where historical data does. Combined, where both meet. The choice depends on the counterpart and the data on hand.
Phase 1 is where the grade is built, calibrated, validated, and used. Not a model-room exercise. The grade the credit committee acts on tomorrow morning.
Five activities. The output is a grade the credit team trusts, the auditor recognises, and the regulator accepts.
Quantitative, qualitative, external. Define the default event.
Identify discriminatory indicators. Filter the noise.
Multivariate. Logistic regression. Weight the blocks.
Map scores to default probabilities. Bucket into grades A to E.
Quantitative back-testing. Use-test. Annual review.
The PD Scoring Calculator runs the statistical block: indicator inputs through the logistic-regression scoring function to a default probability. The Rating Calculator combines the statistical block with the heuristic and external blocks to produce the final grade. Both are working previews from Course 1 of the Risk Management training program.
Enter financial indicators. The scoring function runs a logit model on calibrated coefficients and returns a default probability with the matched rating bucket. Demonstrates how a statistical model translates raw data into a PD figure that credit committee can act on.
Launch the calculator →The full heuristic model. Quantitative block (60% weight) feeds from financial ratios. Qualitative block (30%) captures management and sector judgment. External block (10%) carries the sector and macro index. Composite score maps to a final rating grade and an approval routing decision.
Launch the calculator →Phase 1 instruments combine the rigor of statistical model design with the discipline of heuristic rating. The output is a grade the credit team trusts and the regulator accepts.
Heuristic block (credit-expert questionnaire) plus statistical block (logistic regression on historical data). Weights calibrated to portfolio segment. Five grades A to E with clear delineation criteria.
View the toolThree families: quantitative (financial ratios, cash-flow indicators), qualitative (management quality, sector position), external (macro and sector index). Each indicator carries its discriminatory power and its survey-effort score.
View the toolMaps composite scores to default probabilities using a logit function on calibrated coefficients. Buckets PD ranges into rating grades. Reviewed annually against realised defaults.
View the toolRoutes new applications by grade: Auto Approval, Accelerated Approval, Standard Approval, Alert, Decline. For existing clients, supports limit adjustments based on payment behaviour and rating migration.
View the toolAUC and Gini, calibration tests, override pattern analysis. Documentation that satisfies internal audit, external auditors, and Basel-aligned supervisory expectations.
View the toolWhen the rating became a combined model, our committee stopped arguing about the grade and started discussing the drivers behind it. For the first time the audit trail showed why a borrower was rated as it was, not just the final number.
Head of Credit Risk, regional commercial bank
Two places Rating & Scoring shows up next.
The full method behind heuristic, statistical, and combined model design: data requirements, univariate and multivariate analysis, calibration, validation, and use-test integration.
Six-module program covering rating, expected loss, RAROC, and the supporting validation discipline. Both calculators on this page are drawn from the course materials.
Every downstream calculation in Risk Analytics depends on the rating. Expected loss multiplies PD by LGD by EAD. RAROC allocates capital against the rating-driven PD. The Risk Appetite Statement sets thresholds in PD-weighted terms. A weak rating model contaminates the entire stack. A strong one disciplines it.
Thirty minutes. We compare your data inputs, scoring function, calibration, and validation against what we have built and recalibrated across more than a hundred projects.
Curated news with practitioner commentary. One Deep Dive in rotation. One applied tool. Read in fifteen minutes. Used the same week.