New Book: Redesigning Credit Risk Modeling

Next Monday, Dec 2nd, my new book will be available on Amazon.

You might notice how inexpensive it is. You'll be getting the author's price, meaning no royalties. I just want to make it as accessible as I can.

This is my little book of ideas about how to structure credit risk analytics to break free of the 1960s scoring paradigm and target the real problems in lending. There is almost no math. It's a book of concepts, but I leave out the 200 pages of fluff stories in most business books. It just has the answers.

If you're just interested in hearing the point and saving the $9 (Kindle price), I'll save you some coffee money. Each week, assuming I don't forget, I will post a short synopsis of the next chapter.

Chapter 1: AI / ML are not the answer. This book started when a CRO friend said, "We've done all the machine learning stuff and it hasn't moved the needle. What are we doing wrong?" Whether logistic regression or machine learning, most groups have good credit scores, if you're measuring Gini or K-S. Adding a few points of lift does not move the needle on how to maximize yield and minimize volatility. How do you determine pricing or optimize cut-off scores through the credit cycle and economic cycle? Fixed outcome window credit scores, LR or ML, don't integrate with cash flow calculations and financial metrics, so you're left with magic in the middle, and there just isn't enough fairy dust to go around.

Our first goal is to stop creating rank-order credit scores and directly connect credit risk to finance so that guesswork can be replaced with account-level (or segment-level), economic scenario-based optimization of pricing and cut-off scores.

By the way, GenAI might save you some operational costs, but it does nothing to help solve the key problem in lending. This is not a book on GenAI.

More next week...

Joseph Breeden
Posted on LinkedIn