Customer Profitability: Irrelevant for Decisions?

By Peter Carroll and Madhu Tadikonda

Historical customer profitability reports offer scant help in making crucial sales, cross-sell and service decisions. Better tools focus on the expected values of key decisions for each individual customer.

One of the most influential retail banking metrics to emerge in recent years is customer profitability. Melding new financial methodologies and technology, bankers are calculating the net contribution of each customer relationship — and they are continually amazed by the radical profitability skews surfacing in the data. In a typical retail portfolio, 20% of accounts contribute profits equaling 200% of the overall return, while up to half of the accounts generate losses.

Swept up by the power of these findings, bankers are revolutionizing the way they make decisions. At the structural level, they are weaving customer profitability considerations into long-term decisions about market positioning and major customer groups. To lift unprofitable relationships, for example, they are ushering people into service packages containing price incentives for using automated teller machines and other low-cost channels. At the operational — or day-to-day — level, banks are basing marketing solicitations on customer value data, selectively offering fee waivers and rate breaks to retain high-value clients, and varying service responses based on customer profitability profiles.

Clearly, the most promising growth and profitability payoffs from sophisticated customer information spring from improved operational decision-making about sales, cross-selling and service. Yet, this is also where the use of historical customer profitability numbers is potentially dangerous. Many retail banks take it for granted that such data provide strong guidance for daily decisions. But historical profitability profiles may not signal how customers will behave in the future — even with currently held products, never mind for new ones being offered. In fact, customer profitability statistics are of limited usefulness in decision support.

A far better approach is to support operational decisions with a suite of determinative tools crafted to handle individual customers and situations. This capitalizes on customer information but avoids the trap of regarding last year's profit as the most important predictor of next year's behaviors in a different context.

Here, the emphasis is on the expected long-term values of an array of key decisions for each individual customer. Account-level data is used to forecast customer behaviors, such as purchase propensity, service needs and repayment patterns. These projections are then coupled with bank-specific cost data to calculate the net present value of the lifetime cash-flow consequences of a particular decision, such as attempting to cross-sell a given product.

Such refinements are essential if companies are to make productive use of expanded customer data. The key question for the decision-maker is: "If I take this action, what will be the outcomes, expressed in terms of growth or erosion of shareholder value?" The historical profitability of a customer, perhaps surprisingly, provides very little in the way of an answer — and that can spell trouble for institutions that rely unduly on this type of information.

Missing the Mark

Consider a decision on whether to cross-sell a home equity line of credit to two checking account customers, Mrs. Smith and Mr. Jones. On a rolling 12-month basis, each may each have generated a $55 profit for the bank. A simple profitability-based decision framework would suggest that these customers be treated identically on potential home equity loan solicitations, though perhaps differently from other customers in distinctly higher or lower profit deciles.

However, this framework ignores the numerous ways to generate $55 in profit on most standard checking accounts. The $55 profit from Smith may stem from the interest income earned on her high average account balance, partially offset by the costs associated with the large number of checks she writes per month and her frequent balance inquiries. The $55 profit from Jones may be driven by his perennially low account balance, high incidence of returned checks and corresponding payment of overdraft fees.

In this case, the identical $55 profit figure obscures important behavioral differences that may have a substantial impact on the relative attractiveness of these customers for a loan solicitation. The details of how each $55 return was generated may suggest vast differences between Smith and Jones — in their relative propensities to respond an offer, draw down a credit line, or default on a loan. All of these factors can dramatically affect the merit of making the loan solicitation, but they are lost in the summary characterizations of both Smith and Jones as "$55 profit customers."

This narrative suggests the serious limitations of a profitability-based decision-making framework. In fact, three drawbacks undercut the utility of customer profitability data for operational decision-making:

Descriptive, not predictive. Profitability figures broadly indicate which customers have been economically valuable over a given period. As suggested by the Smith and Jones narrative, however, such data does not necessarily provide good guidance for cross-sell and service decisions for a specific customer. It may not suggest how that customer will behave in the future, especially in the context of the sales/service action under consideration.

Myopic. Virtually all retail financial services products have a specific life cycle. With certain types of loans, for example, customers may have a negative profit contribution in the early years and become more profitable later on. As a result, profitability recorded in a given performance period may incompletely describe the value of a particular relationship.

Retrospective. Historical profit data is a lagging metric. It does not adequately reflect anticipated changes in customer holdings or changes in usage patterns. Furthermore, a retrospective metric may incorporate specific market conditions (e.g. interest rate environment, competitive intensity) which no longer exist.

Putting Decisions First

Senior managers should review how they themselves use measurement tools at the structural level when considering how to construct the ideal operational decision support that answers the question "If I do this, what happens, and does it add value?" When making structural decisions, managers first carefully articulate the main options at hand and then build customized analyses to support their deliberations. Typically, specific business plans and budgets are generated for each strategic option. The models define what is to be done and handicap the outcomes. They incorporate all of the relevant investments, expenses, revenues, risks and hurdle rates associated with a specific decision. Once these elements have been quantified, the business plan cash flows are discounted to yield the action's absolute and proportionate returns.

Tellingly, strategists marshal a broad range of information to generate the economic projections underpinning the business plan.

For example, a bank may decide to redesign checking account packages to address a sharp profitability skew in the customer portfolio. First, checking account customer behavior records (transaction volume, balance levels, automated teller machine usage, etc.) are used to define a set of behavioral segments. Then, customer profitability information is used to compute the average profit per accountholder in each segment under current terms and conditions. Based on this aggregate information, managers make choices about target segments and the design of product/service/price combinations to be positioned against those segments. Such choices are guided by the needs and economics of each customer group and often aug- mented by external market research on these customers.

Strategic analyses also may require customized inputs from several bank departments. Operations, for example, may supply actual operating costs and potential adjustments for different product configurations; marketing may supply insight on product design.

Customer Context

The decision-making principles exercised at the structural level can be applied at the operational level. The key is to assemble an array of information relevant to each important tactical decision and tools translating that data into concrete decision support. In this approach, micro-business plans are created for specific operational decisions for each customer. In other words, the capital budgeting logic of a major structural decision is applied to a customer-level action (see chart below).

For every decision or action, the likely ensuing pattern of customer behavior and attendant cash flow and risk outcomes can be anticipated. For example, an effective decision-support tool for cross-sell solicitation includes estimates of all cash flows associated with the decision. Expenses would include up-front booking outlays, ongoing account maintenance and servicing, and potential credit losses. Revenues would include ongoing fee and spread income.

The decision-support tool will explicitly estimate the differences in these costs and revenues by predicting the key behaviors of Smith and Jones — not just in general but in the context of specific situations. As such, the micro-business plan integrates all of a customer's expected cash flows. These anticipated cash flows may then be discounted and summed to quantify the prospective value generated by a specific operational decision.

Continuing with the cross-sell solicitation example, the decision-support tool would estimate the cash flow consequences of soliciting Mrs. Smith and Mr. Jones for a home equity line of credit. Soliciting prospect Smith, based on specific response, utilization, attrition and default behaviors, may net $75. By contrast, Jones may pose a $20 net loss. And they had the same profitability last year! In each case, the integrated estimate of lifetime value can serve as the primary basis for selecting among possible actions.

This approach is vital for a typical retail bank, and it is markedly different from the subjective classification-and-treatment frameworks used by many managers (see sidebar).

Front-line personnel face many core decisions. Should they solicit a prospect for a credit card? Cross-sell a home equity line of credit to a demand deposit account customer? How should they handle a customer's complaint about a late fee? And what product should be pitched to a customer the next time he enters the branch? No single scheme can accurately characterize customers in all these contexts — and customer profitability, as we have seen, is singularly unable to do so.

Clearly, a decision tool is only as powerful as its estimates of customer cash flows. These cash flows are based on two discrete sets of information: 1) customer-level behavioral predictions, and 2) bank-specific financial inputs.

Individual behavioral predictions include factors such as response, utilization, attrition, default, transaction volume and channel usage patterns. Each behavioral prediction functions in roughly the same manner as a credit score — that is, account-level attributes are used to anticipate the outcomes of decisions based on past empirical data. For sales decisions, prospect attributes gleaned from credit bureaus and other external databases are used to predict behaviors. Such attributes may include external credit scores, total number of credit products held, and home and car ownership. For a cross-sell or service delivery decision, internal data about account balances, transactions and payment patterns may also be used.

These behavioral observations and predictions are then coupled with financial parameters such as funding costs and detailed servicing and operations costs to generate expected net cash flows.

An effective decision tool models the economic outcomes of a particular action. "If I offer this credit card, what is my best prediction of the resulting cash flows — given my current knowledge of this customer — and is the value of these cash flows high enough, both absolutely and in relation to alternatives, to justify going ahead?" This is akin to applying a capital budgeting method to expense-related decisions in the tactical/operational domain.

Salvaging Profitability Information

Fortunately, adopting a decision framework based on value models does not mean expensive investments in customer profitability reporting systems must be scuttled. For one thing, systematically tracking customer profitability is a useful way to monitor overall results. From a decision support perspective, however, banks will need to reorient and extend systems that merely produce customer profitability reports.

Discrete decision tools use many of the data feeds and analytical processes that underpin profitability reporting. For example, value models make extensive use of the same time-series data archives which support customer-level profit reporting and include detailed account behavior and transaction pattern information — although value models benefit from more extensive time-series information. Furthermore, value models rely on the same methodologies for activity-based costing, transfer pricing and risk-adjustment used in customer profitability computation.

What about the customer profitability metric itself? For less important operational decisions, that is, for choices that occur less frequently and/or have a lesser impact on customer value, customer profitability can serve as one of a small number of simple parameters used for decision support.

Consider a branch officer's decision on how to handle a bounced check. The decision on whether or not to waive a bounced-check fee for an individual customer could be supported by a full-fledged model that predicts the likely impact of the decision on customer behavior and value. However, the stakes and frequency of this particular decision may not be high enough to warrant the creation of such a model. In such instances, a simple rule built around customer profitability may suffice. For example, waive the fee if the customer's demand deposit account balance exceeds $2,500 and his 12-month profit exceeds $75.

The development of customer profitability data has by no means been a fool's errand. The only danger lies in simplistic reliance on such data for decision support. A far better approach is to: 1) make structural decisions by performing focused ad hoc analyses, which may include data on customer behavior as well as customer profitability information, and 2) support operational decisions with a suite of determinative tools crafted to handle individual customers and situations. By shifting to this two-pronged framework, and by building appropriate applications, the work that went into the customer profitability system will be put to good use.

Mr. Carroll is a managing director and Mr. Tadikonda a senior manager at Oliver, Wyman & Co., New York.

Copyright © 2003 by Banking Strategies, published by BAI.

back to top