Briefs

NAREX has developed an artificial-intelligence program that quantifies collectibility -- it tells which bad-debt accounts are likelier to pay than others, and the best way to make them pay up.

"We can match up what kinds of human interaction works best with which type of consumers," said NAREX founder and CEO Bernhard Nann, who worked for credit-ratings agency TRW before founding NAREX in 1995.

Before NAREX, "The collections industry was morphing (credit ratings) scores into a form where we could gauge collectibility," said NAREX customer Jason Howerton, the chief operating officer of ARS National Services is Escondito, Calif. "The problem was no one really was diving headfirst into creating a score on recoverability."

Nann said NAREX improves on static recoverability models that have eight to 12 variables with fixed weights. Instead, NAREX crafts customized algorithms for each customer drawing on 80 to 200 variables, weighted differently for each consumer. And NAREX uses fresh data to continually update the models.

Customers include the credit grantors, who must decide which agencies get which accounts, and the agencies themselves, who must decide which methods to use.

Nann won't get more specific about how the models work, calling the NAREX way proprietary. But he says NAREX -- which runs 90 servers with 19 terabytes of storage -- can improve collections by 5 to 20 percent.

That's a big deal in a big industry. Maryland-based consultant Kaulkin Ginsberg Co. estimates $135 billion in delinquent consumer debt was farmed out for collection in 2000, and there were 6,500 collection agencies and 1,600 credit reporting agencies chasing it. It's still going up: Fitch Ratings said its chargeoff index rose in March to its highest level since July 1998.

See http://www.narex.com/