Editor’s note: Gautam Bose is senior risk analyst at Banc One Corporation, Columbus, Ohio, and Erick Haskell is a senior consultant with Deloitte Consulting in Minneapolis. The authors are not writing in their professional capacities and the opinions contained in this article do not necessarily represent those of their respective employers.

In the past decade, two important business realities have factored prominently in the development of marketing strategies for most retail banks. First, a small percentage of customers, perhaps 20 percent, generally account for an overwhelming percentage of profits, often more than 80 percent. This is commonly referred to as "the 80-20 rule." Second, it is far more costly to acquire new customers than it is to retain existing ones.

With these facts in mind, banks have spent considerable time, money, and effort identifying their most profitable customers and developing strategies to retain them. Such strategies have included differentiated service, premier products, and preferred pricing (fee waivers, rate reductions, etc.) for a bank’s most profitable customers. While such strategies have often proven effective in the retention of targeted customers, many banks’ profitability metrics may not accurately identify the most high value customers, particularly if the customer has a loan or credit product. This is due to the failure of most banks’ profitability metrics to adequately incorporate the individual risk of each customer.

Current approaches for incorporating consumer credit risk

Currently, most banks include a charge for loan loss provision in their customer profitability metrics for credit products. The charge for loan loss provision is generally an average determined at the product level. Each customer holding a particular product is allocated the same average charge for loss provision. Some banks use this same process for allocating a capital charge to customers as part of their profitability measurement.

Then, in an effort to incorporate risk, many banks will develop a matrix relating customer profitability to individual risk. As the graphic below shows, customer profitability is on one axis and risk is on the other, with risk typically represented by a credit bureau score.

Managers can then pursue specific strategies for customers within each quadrant. Obviously, a bank’s highest priority is to retain those customers who are most profitable with the lowest risk (highest credit bureau score). Those customers who are profitable but have low credit bureau scores should be re-priced so that their risk to the bank is compensated by higher fees or a higher interest rate. Unprofitable customers with high credit bureau scores should be redirected to low-cost channels or targeted for cross-sell opportunities. Finally, unprofitable customers, with low credit bureau scores should be strategically introduced to competitors if it is determined that they will not become profitable in the future.

While this method attempts to incorporate individual risk in the determination of customer profitability, it is still insufficient for accurately identifying high value customers, and it fails to make optimal use of the rich customer data now resident in most banks. The matrix approach does not incorporate important drivers of customer behavior in the decision making process, such as other relationships with the bank, time on books, and demographic, psychographic, and socioeconomic factors.

The solution: risk-adjusted customer profitability model

Banks usually have a rich repository of customer information which, if utilized appropriately, can be very predictive of future customer behavior. Using advanced statistical analysis, tools may be developed that will allow multiple variables to be used for decision making. Specifically, to address the shortcomings of the matrix approach, banks should make individual risk adjustments at the customer account level within their customer profitability models.

The logic for making such an adjustment is no different than that which has been driving improvements in customer profitability measurement in recent years. Just as banks have diligently measured other elements of the profitability equation at the customer account level (e.g., cost of funds, account fees, transaction costs, etc.), they should now focus on attributing risk at the individual level. Below are the steps required to develop a risk-adjusted customer profitability model.

1. Identify significant predictive variables. In addition to credit bureau and behavior scores, banks have a wealth of customer data that can be used to increase the accuracy with which they estimate the likelihood that their customers will default on their credit obligations. Advanced multivariate regression techniques can easily identify the most significant variables from the vast number of predictor variables available on the customer databases. The dependent variable is credit default and potential independent variables come from the bank’s entire repository of customer data, including account information (balance, time on books, other products held), behavioral data (payment history, credit utilization), and available demographic and psychographic data. Regression analysis will identify those variables that have the power to explain the probability of credit default. Most banks will be surprised to discover the predictive power of the data residing in their own customer information files.

2. Build probability of charge-off model (scorecard). After identifying the variables most predictive of customer default, the next step is to build a probability of charge-off model - or scorecard. The scorecard uses the predictive variables, along with historical customer charge-off data, to arrive at a probability of charge-off for every customer. Specifically, the model is developed using statistical software tools to query the customer information file with data for the past 12 or 24 months. The query returns the average incidence of default for every combination of the predictive variables. Visually, the model would appear as a multidimensional object with each cell representing the probability of charge-off for customers possessing particular characteristics represented by the predictive variables. For example, a customer with a home equity line of credit, who has been on the books for 14 months, who frequently draws the line to its limit, and who also has a checking account at the bank, may have a 1.17 percent probability of charge-off over the next 12- or 24-month period.

3. Incorporate risk adjustment. After developing the scorecard, the charge-off probabilities should be incorporated into the customer profitability calculation by applying a probability of charge-off to each customer account. The mechanics of this step are generally the same for most profitability metrics, including customer contribution, net income after capital charge (NIACC), or lifetime value. Essentially, the risk adjustment is carried out by substituting an account level loan loss provision and an account level capital charge for the their product level counterparts in the existing profitability calculation.

For customer Jane Doe, account level loan loss provision is equal to Jane’s probability of charge-off multiplied by the outstanding balance on her loan (or the dollar limit on a revolving credit product) minus the expected recovery for secured products. For example, if Jane has a probability of charge-off equal to 3.39 percent, an outstanding loan balance of $7,555, and an expected recovery rate of 80 percent, her account level loan loss provision is equal to $51.22 ( [$7,555 x 3.39%] x [1-80%] ), or $4.27 on a monthly basis. This amount replaces the product-driven loan loss provision in most profitability models.

To derive the account level capital charge, each customer must be assigned a risk factor, which is also derived from the probability of charge-off model. The risk factor is calculated by dividing the customer’s account level loan loss provision by the total outstanding balances (less expected recovery) for that product. The resulting risk factor represents the percentage of total risk for that product represented by each customer. Next, account level capital charge is derived by multiplying the risk factor by the total capital allocated to that product.

It is important to note that although this process may seem like a substantial undertaking, it is likely that other departments within the bank are probably already performing some of these steps for other functions. For example, during the loan approval process, most banks use scorecards akin to the probability of charge-off model described above. Similarly, the bank often analyzes the existing customer base for risk management purposes. Therefore, adding a customer level risk adjustment to the existing customer profitability model may simply involve leveraging existing techniques in other parts of the bank.

It is also important to note that there are several critical prerequisites for successfully implementing a risk adjusted profitability model. These include:

  • Mastering the basics. Guarantee that the fundamentals of profitability analysis, such as activity-based costing, funds transfer pricing, and risk-adjusted capital allocation, are correct.
  • Warehousing the data. Design a customer database with accurate data that is regularly updated and validated.
  • Acting on the information. Ensure the availability of technology that will communicate the customer knowledge from the Risk Adjusted Profitability Model to the front-line customer service representatives so that it is used at the time of customer contact.

Competitive advantage

Successful retail banks are devoting as much time and effort to customer retention as they are to new customer acquisition. Retaining profitable customers and devising successful strategies to increase the value of unprofitable customers will be the single largest source of competitive advantage in the coming years. Achieving this competitive advantage requires that banks have the ability to accurately identify their high value customers. Traditionally, customer profitability analysis has not adequately incorporated credit risk at the individual level. By developing risk-adjusted customer profitability metrics that take full advantage of new technology and the wealth of customer data available in most institutions, banks will ensure that their retention efforts are truly targeted at the most profitable customers.