Make the most of what you've got," is the mantra in the credit card business these days, as new lending is all but shut down and consumers are spending and defaulting in ways legacy models are not equipped to predict.
"The new account market is really in suspended animation," says Dennis Moroney, research director at Mastercard-owned TowerGroup. "Mail volume fell off a cliff and nobody in their right mind is trying to acquire."
Institutions are instead leveraging analytics, reweighted models and decision optimization technology to maximize retention, enhance portfolio management and minimize losses.
Credit card acquisition mailings were down 29 percent in the third quarter of 2008, compared with the same period in 2007, according to Mintel Comperemedia; the average maximum credit amount offered declined 27 percent in the third quarter of 2008 from the first. Federal Reserve numbers on consumer credit indicate total outstanding revolving credit declined 3.4 percent to $974 billion in November.
Those responsible for credit strategies and portfolio management know the reality behind the rapidly declining economic landscape: consumer credit conditions are changing faster than existing models and bureau scores can account for, and the time it takes to put new credit models into production means their value dramatically erodes not long after they're implemented.
"In today's market if a scorecard is only indicative for a year, and it takes you six months to get it, the payback period on that card is so small," says Tom Johnson, vp of new products at Zoot Enterprise, which sells credit decisioning and loan origination products.
Because of the enormity of the undertaking, most issuers aren't completely rebuilding models, rather they're keeping the structure of the models stable but tweaking the weights and parameters. "Over the last six months we've seen more demand for credit line decreases. [But] people are trying to be clever about the way in which they implement those strategies because they understand it's very easy for them to backfire," says Andrew Jennings, svp and CRO at Fair Isaac Corp.
Retrenching on credit strategy is not as simple as raising score thresholds because the performance of any given score changes as economic conditions fluctuate. Issuers with less sophisticated tools may be relegated to this strategy, but others are finding that decision optimization technology can help make the most of credit pullback. The largest issuers have been utilizing optimization for years, but the technique has trickled into the mid-market says Dan Gellar, evp of Market Rate Insights.
Fair Isaac, using its consortium credit card transaction data, recently conducted an optimization experiment with an objective of decreasing the credit lines on 10 percent of active accounts. In the first scenario, the objective was accomplished using standard blunt instruments - decreasing customers with lower scores, higher utilizations, etc. The outcome of the test: profits on this group decreased 10 percent.
In the second scenario, the same objective was accomplished using optimization, taking into account how the decrease would impact customer's balance, probability of delinquency, and probability of attrition. Using this model, the same credit line decrease decisions increased the group's profitability by three percent.
But it's not enough to reduce the total amount of credit outstanding. Many issuers are looking for ways to stay a step ahead of consumer behavior by predicting which accounts will default, and reaching a settlement before consumers know they're in danger of default. And while bureau scores based on historic data still perform well when ranking risk, says Moroney of TowerGroup, they're not proving to be adept at forecasting losses. This leaves issuers looking for real-time insight into each consumer's position. "Issuers are trying to figure out surrogates for that, other types of metrics they might use," Moroney says, such as checking account balance information.
The other variable issuers are struggling to account for in credit lending strategies are economic metrics like unemployment and inflation. Adding economic indicators doesn't happen in a scoring model, but rather in offline testing of how economic conditions would affect a given credit strategy. Much of this involves taking the concept of stress testing and applying it to the individual account level.
"It's one of those areas that I think it would be foolish to say, 'Here's the model, just plug in some economic forecast and out comes the answer.'" Jennings says. "What you're trying to do is get a sense of the range of possibilities."
