Credit Scorecards: Weight of Evidence, Information Value & Points Scaling
Coarse classing, WoE, Information Value, monotonic bins, points with factor/offset and PDO, scoring applicants, cut-off strategy and population stability: 8 hands-on labs on one fictional lender's synthetic applications.
The first lesson is free — no account needed.
About this course
A credit scorecard course that builds the application scorecard of one fictional lender, Larkspur Lending, on synthetic data in a spreadsheet, for credit risk analysts, model validators and finance students. You group ten fine classes of debt-to-income into five coarse classes and count goods and bads, calculate and chart the weight of evidence of each class, rate eight characteristics by information value with the usual IV bands, merge classes to fix a reversal in the bad rate, scale a WoE logistic model to 600 points at 30:1 odds with 20 points to double the odds, score six applicants with lookups and turn scores into default probabilities, pick the cut-off that meets a 3% bad-rate target, and check the population stability index after launch. Synthetic data, no lender or bureau affiliation. Real scorecards must also meet fair-lending rules. Educational only, not credit or financial advice. The first lab is free.
Lessons
- Coarse Classing: Binning a Characteristic into Good/Bad CountsFree
- Weight of Evidence: LN of the Good/Bad Distribution Ratio (with a Chart) Pro
- Information Value & Predictive Power Bands Pro
- Monotonic Bins: Spotting a Reversal and Merging Classes Pro
- Points per Attribute: Factor, Offset, PDO & Base Odds Pro
- The Scorecard Table: Scoring Applicants with Lookups Pro
- Cut-off Strategy: Approval Rate vs Bad Rate (with a Chart) Pro
- After Launch: Population Stability Index & Characteristic Analysis Pro