Credit Scoring Models: A Technical Description of the Design Architecture and Really Fun Reading

We rarely do this, but we thought it was super cool so we're posting John's article for Credit.com to the blog too!

I've been writing credit articles for the good people at Credit.com for almost 4½ years now. I've always tried to write them in such a way that my grandmother would understand them: simple and straightforward. I think that once every 500 articles, however, I should earn the right to really geek out and get technical on you. So here it is; enjoy!

Every credit scoring model, FICO®, VantageScore®  -- you name it – is actually a series of smaller models referred to as scorecards. Scorecards are models that are designed to best evaluate the credit risk of some sort of homogenous subpopulation, like those who have very young credit histories or have filed bankruptcy, for example. The purpose for this multi-scorecard architecture is to yield the most powerful credit-scoring tool regardless of the type of credit report being scored. 

When your credit file is pulled, and before it is scored, the credit-scoring model decides which scorecard it’s going to use to calculate your score. Don’t make this more complicated than it needs to be. Think of a bowling scorecard or a baseball scorecard. You add up points and at the end of the game you have a final score. This is no different, but instead of tallying runs and pins, the model is tallying the number of points you’ve earned for various credit characteristics. Oh, there I go...more technical. What’s a characteristic? 

A characteristic is a component of every scorecard. In fact, there are many characteristics in every scorecard. Think of a scorecard as a question that the model is asking your credit report. For example, “Mr. Credit Report, how many late payments do you have?” Another example could be, “Mr. Credit Report, how many inquiries do you have?” Those are all characteristics, and they each have an answer. The answer is referred to as a variable because the answer can vary from consumer to consumer and from credit report to credit report.

Each variable (answer) to the characteristic (question) is going to have a value or a weight. The weight is simply the number of points you earn for that particular question. Example: Let’s say you have five inquires on your credit report. That “variable” might equal 25 points (weight). To summarize, every credit report is broken down into characteristics, variables and weights. Once all of the weights are calculated, you end up with your final score. 

So there you have it: Credit Scoring Model Design 101. Thank you for allowing me to geek out on you. I promise I won’t do this more often than once every 4½ years!

John Ulzheimer – Credit scoring and credit reporting expert and author, John is the President of Consumer Education for Credit.com. Formerly with Equifax and Fair Isaac, John shares his unique insight of the inner workings of credit scoring models and the credit reporting industry on CreditBloggers.com.

Comments

[Jeannette Medlen]

So, let me ask you, if I find so very many inquiries into my credit history and they are from collection agencies or credit offerings, I suffer the possbility of them being weighted against me? How do I fight that?