The bigger the bank, the more likely they have a variety of custom models built within their app-processing systems and account management systems. Almost every single one of them uses FICO as an input.
Let's say that Leslie McFadden's $500,000 salary equals 50 points and the fact that she's lived at the same address for 10 years equals 50 points. The fact that she's been in the same industry for 15 years means that it's worth 50 points. The fact that her FICO score is worth 796 is also worth 50 points. You can give points for certain things. It's used as a variable within the model.
All of those models I just mentioned to you -- the bankruptcy score that Equifax builds, the variety of FICO scoring models, the behavior scores, the attrition scores, the collection scores, response models -- a lot of those are what are called shelf models, meaning they're on a shelf somewhere and all you have to do is tell one of the credit bureaus, "Hey, I want to start buying that score" and they'll just start putting it with your deliverable. They're built for general use, meaning that anybody can buy them and as such they're inexpensive -- there's no maintenance with them at all because either the bureau does the maintenance or FICO does the maintenance and they're very easy to use.
Someone like a credit union or a small regional bank -- they will depend almost exclusively on those prepackaged credit scoring models, like FICO, because they just simply don't have the money to go out and contract with someone to build a $3 million custom score and then have to rebuild it every 24 months because they get stale.
Now when you get into the big boys -- we all know who they are, the really big lenders and the credit card guys -- they don't want to just simply depend on a shelf model that everyone else is using, they want something that's customized specifically for them because of the performance. They may have a custom response model, they may have a custom risk score, they may have a custom attrition score, they may use a custom collection score, they may have a staff of people internally who build all these models themselves -- an entire operations, research and analytics team -- and some of them do that versus outsourcing it. Some of them would rather pay someone else to do it. It's just a matter of what their appetite is.