Independent mortgage banker Countrywide has been way out front in terms of technology for a long time. We interviewed Chief Technology Officer Richard Jones to find out how the company knew to build its own automated underwriting system 10 years ago, and what spoils have come from having a proprietary AU.
EARLIER THIS YEAR, MORTGAGE BANKING INTERVIEWED RICHARD JONES, chief technology officer at Countrywide Home Loans, Inc., Calabasas, California, an industry leader in technology adoption. We asked him about the company's decision to build its own automated underwriting system (AUS), the benefits and drawbacks of the "build" decision and what the future of automated underwriting (AU) development is going forward.
Jones' answers illustrate how Countrywide's size and technology prowess continue to keep it at the top of the list of successful mortgage banks. In fact, it seems the company has had a clue about technology for quite a while, even before it built its groundbreaking CLUE [TM] (Countrywide Loan Underwriting Expert System).
Q: Can you talk about when and why Countrywide first decided to develop its own automated underwriting engine?
A: We first decided to do this in 1991. At that time [there] was a refinance boom, kind of like what we're having right now. There was really nothing in existence at that time... no other automated underwriting system, and there was a shortage of underwriters. We were doing a huge amount of volume and we needed to automate. This is consistent with Countrywide's philosophy. All along, [our idea] was to use technology to gain a competitive advantage, and that was our thinking in 1991.
Q: You mentioned there wasn't really any other technology out there--but did Countrywide consider buying pieces of the technology and putting it together, or just waiting for someone else to come out with some automated underwriting technology, rather than going through the expense and the time of Countrywide building it itself?
A: Well, we always [do a] "build" versus "buy" decision whenever we look at a technology project. That's exactly what we did in that case. There wasn't any GSE [government sponsored enterprise] solution. There was no LP [Loan Prospector [R]] or DU [Desktop Underwriter [R]]. Our solution predated those solutions. We realized that we needed to have something that would have a very sophisticated decision-making capability. We looked at neural network technology. We looked at a number of different technologies, and decided on a combination of a rules engine and a statistical scorecard methodology. We also looked at decision tree--type [software]. The problem with the neural networks and the decision tree--type software was that it was too hard to explain how the answer was derived. It was like a black-box kind of situation. So while they could come out with interesting answers to the underwriting decision, it was very hard to account for exactly how the decision was arrived at.
The way we did it with GLUES was, we bought a rules engine software product to help us define and automate, but we also have built statistical scorecards [to help us] look at the history of certain loans and certain situations and the probability of default. Then we built a historical scorecard around all that information.
Q: What were some of the advantages Countrywide thought it would get by building its own system rather than either waiting for someone else to build one or using some of this other technology that you mentioned?
A: There are several advantages that we thought [of]. We have a large range of products that we offer customers. We need to have the flexibility to create new product [and] respond with an automated underwriting decision n that reflects those new products. We're able to maintain a one-month release cycle, so if a new product is invented by marketing, we're able to support that with [an] automated underwriting decision within a one-month release cycle.
Another big advantage, we felt, was that it wouldn't be just a black box. We would know what was inside the black box. We would know all the elements of the risk decision that was being made and understand exactly how that decision process was arrived at by the software. As opposed to just throwing it to a black-box decision-maker and it comes back with an answer and you don't really know all the factors that went into making that decision.
Q: Last year, the matter of black-box systems became a huge issue in the industry, particularly relative to credit scoring and also in relation to Fannie Mae's and Freddie Mac's automated underwriting systems. Do you look back and think, "We really saw something far ahead that's coming to be an issue right now"?
A: I think Countrywide demonstrated a lot of foresight when it [developed its own automated underwriting system.] Certainly Fannie Mae and Freddie Mac realized that their DU and LP products are...