Comment: Reevaluating Strategies to Increase Credit Card Recoveries

A $25 billion problem is getting a lot of headlines these days, but much less analytic attention than it deserves. And often the attention it does get is not remotely the right kind.

The problem is unrecovered credit card losses. At present, the average credit card company picks up just 10 cents in net recoveries per dollar charged off. But if some key weaknesses in current collection processes were recognized and corrected, recoveries could rise measurably, perhaps as much as 20%.

The chief weaknesses of collection processes are twofold. First, the processes are undifferentiated.

Virtually every delinquent cardholder gets the same treatment, regardless of whether it is needed, will work, and is cost-effective.

Second, even if this cookie-cutter formula results in the right treatment for a given customer, it is usually inefficiently administered, crimping the issuer's eventual net return-recoveries minus expenses.

To attack the first problem, credit card companies need to understand that they are being schizophrenic. On the one hand, the best of these companies employ increasingly so-phisticated models to predict who is likely to use a credit card profitably and therefore should be targeted for an origination effort.

On the other hand, the majority of issuers, including industry leaders, use few quantitative tools to determine which customers, once they become delinquent, require a traditional collections effort and which are likely to "cure" without such an effort-one, incidentally, that almost always serves to rupture what might have been a salvageable customer relationship.

Companies need to replicate on the back end of the credit process (collections) something like what has served the industry leaders so well on the front end (customer acquisition).

Specifically, they need to employ scoring tools based on behavioral data that digest both external information (such as that provided by credit bureaus) and, more importantly, proprietary internal information (e.g., transactions data) to rank each account according to its likelihood of repayment. The resulting scores provide the basis for segmenting the accounts into discrete buckets, each of which gets a tailored collection effort.

Consider one such possible bucket-call it the retention group. These are usually long-lived customers with low outstanding balances who are temporarily out of pocket. They want to pay and eventually will.

Indeed, an analysis of a representative sample of 1994 delinquents who, the models indicate, belong in this bucket reveals that within one year they repaid almost all outstanding balances.

So, how to treat these folks when they stop paying? A hint: Don't cancel their cards, and send them off to a collection agency.

The proportion of canceled cards held by retention-bucket customers could be as high as 30%.

A mature, customer-hungry business can't afford to cast aside so many people. There are other proven ways to facilitate repayment without risking alienation of this basically attractive customer group.

On the other hand, let's consider a less palatable case: call it the deadbeat group. Current practice, being inflexible, routes these people through the usual time-based process: in-house efforts, then agencies, then maybe legal action.

But since the scoring models can predict with a high degree of accuracy which delinquents will be deaf to appeals, why waste money making these appeals? Clearly, the need is for quick and summary action.

Good segmentation requires that the issuer understand not only when to extend the gloved hand but also when to apply the mailed fist.

Unfortunately, it is difficult for the less-advanced credit card issuers to transition to a sound segmentation strategy. The main reason is that while some in the industry are adept at modeling, most are not.

This, in turn, is the result of bad data (missing elements, inconsistent data definitions, and unconsolidated data bases); bad systems (can't score accounts using untraditional variables); and bad analysis (existing staff is not familiar with advanced statistical techniques).

The solutions to these problems are reasonably straightforward. They include: new and revised procedures for data capture and warehousing; systems modifications; and, most important, the development of an adequately staffed modern marketing function.

Besides organizing to "make the punishment fit the crime"-i.e., defining the collection strategy that is most appropriate to a given customer segment-well-managed card issuers are beginning to raise collection yields by rethinking operational approaches to all collection channels-internal, agency, or legal network.

There is, in fact, tremendous scope for improving gross recoveries (now averaging around 15 cents) and reducing the expenses of recovery (about 5 cents) through improved management of the steps in the collections process.

At the very minimum, one can reduce in-house collection costs by more aggressive sourcing of purchased components-e.g., telecommunications and letters. But what is appropriate to do in-house may be inappropriate when dealing with outside collectors. Here, increasing rather than reducing expenditures may be necessary.

Standard practice, for example, has long been to pay collection agencies a flat contingency fee of, say, 25 cents on the dollar. Practice has also been to mandate strict work rules-e.g., a collection call a day, a collection letter a week. Our analysis of these practices reveals that they are massively inefficient, resulting in limited agency or network incentives and useless spending.

Superior management of outside collectors could mean less micromanagement. In other words, don't tell the agency what to do and how to spend money. Simply recast incentives to elicit better performance and higher net yields.

This means compensating agencies at anywhere from 50 cents to as much as 80 cents per dollar recovered above valid historical collection benchmarks. Graduating payments in this manner turns out to be the only foolproof method of preventing "skimming"-the practice whereby collectors supposedly work the client's paper only enough to achieve a predetermined collection norm before passing on to another client.

To summarize: It is possible to raise net yields by 20% while preserving many more customer relationships than is now the case, provided issuers:

Transition from a process that does not differentiate among customers to a segmentation approach that aligns the individual customer with the optimal recovery channel.

Revamp policies and procedures for all collection channels in order to lift incentives and reduce rampant waste. u

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER