The Fallacy of Customer Retention
For the past year or so, bankers have been assailed with a proposition that can best be described as a nostrum or panacea. The proposition holds that they have only to increase their customer retention rates in order to banish profitability problems.
This viewpoint has, from time to time, insinuated itself into the pages of this newspaper. Its most comprehensive formulation, however, is to be found in the Harvard Business Review article of last fall, entitled "Zero Defections: Quality Comes to Services." The article says, "Companies can boost profits by almost 100% by retaining just 5% more customers." Of all the businesses referenced, it was branch deposit gathering and credit cards that came closest to fulfilling the boast.
Unfortunately, the argument is flawed and its prescriptions dangerous. This article will examine the argument and expose the flaws. It will also show how, despite these flaws, a focus on customer retention can still fit, tactically, within a broader framework of customer management.
The retention argument is as follows:
1. It costs you money to bring a customer into the bank.
2. Year-by-year customer cash flows are increasingly positive.
3. The loss of an account consequently deprives you of those positive (albeit fairly highly discounted) future cash flows.
4. The cost of account losses is hard to "see" in conventional accounts or in period-based management information systems. A "present value of cash flow" approach is more revealing.
5. Increasing the customer retention rate has the mathematical effect of raising average account life span.
6. An increase of just a few percentage points in the retention rate for deposit product customers, therefore, through its effect on average account life, translates into a large percentage increase in the net present value of average customer relationships.
A Seductive Appeal
This logic is presented as more than a mere observation of economic interest. It is presented as a revolutionary insight: "Customer defections have a surprisingly powerful impact on the bottom line. They can have more to do with a service company's profits than scale, market share, unit costs, and many other factors usually associated with competitive advantage." And elsewhere: "Low-defection strategies can overwhelm low-cost strategies."
The argument is very seductive. It provides a link to the world of manufacturing, where Japanese methods that stress quality appear to pay off. If a slavish devotion to quality turns out to be economic in manufacturing, doesn't it seem possible, even likely, that the same will be true in services? In banking?
The service quality theme in banking is already well established. The retention argument is so seductive precisely because it seems to validate and quantify this service quality theme. Until now, it had been more or less a working assumption that investment in service quality pays off. There is a clear possibility that improved service quality will reduce customer defection.
Since the benefit of reduced customer defection has been quantified, it may be possible to justify the cost of improving service quality via the economic gains of lower customer defection. If a 5% reduction in the defection rate improves profits by 85%, the amount we are justified in spending on service quality enhancements is potentially enormous.
Correlation, Not Causation
When something seems too good to be true, it usually is. This kind of one-pill-cures-all-ills argument has been heard before. For example, it was once observed that company profitability and market share were closely linked. Managers seeking to raise profitability were encouraged to focus on the more tangible goal of increasing market share. Since the two were linked, the achievement of the market share goal would ensure the achievement of the profit goal.
In a classic misstatement, Bruce Henderson (then at the Boston Consulting Group and a prime mover in the market share school of thought) said, "It doesn't matter why market share and profitability are linked, only that they are." This is reliable advice only to those who are predisposed to see causation in correlation.
Today's version of the old aregument is that profitability and the customer retention rate are correlated: Increase the retention rate and you will increase profits. The basis for this assumption is equally shaky and for the same reason.
The authors of the HBR article make two good points that should not be lost. First, a present-value approach to the evaluation of customer accounts is more revealing and more accurate than the conventional period-accounting approach. Second, within such a present-value formula, the length of the relationship is an important variable. Often, the longer the average life of the relationship, the higher is the net present value. But not always and, at many institutions, not even in the majority of cases.
Unfortunately, the argument is flawed in just about every other sense. One's first suspicion upon reading that a "simple fix" can raise profits by 85% is that profits must currently be so low that a trivial actual dollar increase raises them by a large percentage. Retail profits are often understated because of inappropriate measurement systems. Understated profits are easier to raise, at least in percentage terms.
This is a mistake that is easily made if one has a relative lack of experience in the financial services area. To see the profitability of deposit gathering clearly, you must usually revise transfer price procedures (in order more accurately to value the deposits, which are notionally "sold" by the branch network to the bank's treasury department), expense allocation methodologies, and capital allocation rules.
There are some nice judgments to be made. What is the effective maturity of checking deposits; what yield curve do you transfer price from; exactly how much equity do you need to support a deposit-gathering activity? It is very easy and quite common to underestimate the profits generated by deposit gathering and, therefore, very easy to exaggerate the power of profit gains. I strongly suspect that the retention argument rests, in part, upon mistakes in measuring the profitability of deposit gathering which tend to exaggerate the gains to be made if account life is lengthened.
Customers Are Like Cheese
One's second suspicion upon reviewing the claims of the retention argument concerns the conerstone assumption of the retention-leads-to-profit argument. Customers are said to be like cheese: They ripen with age. An exhibit in the HBR article shows the net cash flows for up to 20 years of customers' account life. As time goes on, the cash flow balance becomes more and more strongly positive. This is a powerful and prevalent myth in consumer banking. Unfortunately, it is not true for the majority of customers, as we shall see. This is the fatal flaw in the retention argument.
The retentionists have not, I suspect, tracked portfolios of accounts through 20 years of experience. This is very difficult to do. They have looked at a portfolio of customers, sorted them by account age, and analyzed the gross cash flows for each "vintage." This is much easier to do. On this basis, customers do appear to improve with age. Each vintage is found to have superior annual profitability to the younger vintages.
One may therefore conclude that customers get better with age. The assumption being made, however, is that the average profitability of today's five-year-old accounts is a good predictor of the future profitability of today's one-year-old accounts. This assumption turns out to be invalid because it ignores a second, more powerful, effect.
The powerful effect is the wide array of individual results within a portfolio of customers. Properly measured, the annual cash flows on a sample of retail banking customers is very skewed. As many as 60% of customers actually have a negative annual contribution. As few as 20% have a positive contribution that is also equal to or greater than the required rate of return on capital employed. That such a portfolio can have an acceptable aggregate return is a minor miracle.
As a shorthand, let's refer to these customers as follows:
"A" customers: acceptable annual contribution.
"B" customers: unacceptable but positive contribution.
"C" customers: negative annual contribution.
When you examine the claims made for customer retention in this light, a very different picture emerges.
The reason groups of customers get better with age is that they contain a rising proportion of A customers. The reason the proportion of As rises is that the Cs drop out. It is not that the Cs become As.
Look at it this way: In the first quarter of 1987, you put 1,000 new customers on the books. In the second quarter, 500 of these accounts lost money, 200 made a little but not much, and 300 were strong contributors. By the end of 1989, from this group of 1,000 you have 500 customers left.
Of the 500 who were initially unprofitable, only 150 remain. Of the 200 who made a little, 100 remain. Of the 300 who made strong contributions in 1987, 250 remain. The mix has shifted toward As, who now constitute 50% of the total, whereas in 1987 they made up only 30%. The resulting improvement in profitability has been caused by the departure of Cs, not by the change of status of Cs to Bs and Bs to As.
Such improvements in profitability can be marked. If an A is worth $250 per year, a B is worth $50 per year, and a C is worth ($150) per year, the average annual contribution goes from $10 per customer in early 1987 to $90 in late 1989.
Why do the Cs leave at a higher rate than the As? Probably for much the same reason that they are Cs: They may be particularly price sensitive; they may be oriented toward products that tend to be lower profit; they may be quick to take offense.
Whatever the reason, Oliver, Wyman & Co. analyses show that the ratio of average balances of stayers to leavers is 2:1 for young accounts (one to three years), 1.5:1 for older accounts (four to seven years) and 1:1 for old accounts (seven to twelve years). The ratio only becomes less than 1:1 for the very oldest accounts: 0.95:1 for accounts aged fifteen to twenty years. Lower-balance, less profitable accounts defect more quickly than higher-balance, more profitable accounts.
If you examine the lifetime profitability of today's older customer, you will find that their economic status (A, B, or C) has not changed much. (obviously, individuals can and do change - Horatio Alger was an individual - but the migration of Cs to A status is not a broadly observable, statistical phenomenon.)
A Breathless Announcement
So, the cornerstone of retention analysis is a wobbly foundation indeed. This did not prevent the president of the Council on Financial Competition (a Washington-based for-profit enterprise) from declaring breatlessly, "For the past 90 days, the Council has been collecting data on the success of some financial institutions in better retaining and recovering retail and corporate customers. The dimension of success and the concomitant effect on corporate-wide profits are almost unprecedented in Council reserch. A 5% gain in customer retention rates can result in an 80% increase in retail bank profits." ("Zero Defections: Perfecting Customer Retention and Recovery." The Council on Financial Competition. He was not referring to research conducted by the Council but to the work of the firm of Bain & Co., which underpins the HBR article.)
In fact, retail bank customers do not get particularly more profitable with age. The appearance that they do is a misreading of the available data on the cash flows of groups of accounts. Active programs to retain defectors until they become more profitable are likely to be highly counterproductive.
As noted, the reality is that many or most of the bank's retail customers lose money. Practicing retention analysis and intervention on the losers will only compound the problem: You will pay more money to retain people who will then continue to lose you money annually. You will pay to perpetuate a customer mix that is rich in C customers.
There are many other aspects of the retention argument that are invalid: for example, the six reasons given for supposed progressive improvement in customer profitability. Each is incorrect, is correct but inadequate, or is an example of specious reasoning.
Reason No. 1: High first-year customer acquisition costs. In a cash-flow picture of an account's life, the acquisition costs usually render the first-year cash flows negative. This is a perfectly good reason, therefore, why cash flows in all subsequent years might, on average, be higher. It does not explain why cash flows in years two to seven should each be progressively higher.
Reason No. 2: Individual account balances grow across time. Such a phenomenon could explain rising profitability, but this explanation turns out not to be true, as we have seen. Group average balances rise as low-balance customers defect. Individual balances do not grow much.
Reason No. 3: Number of accounts grows across time. If the number of accounts per customer grew with time and if all accounts were profitable then this could, indeed, explain rising profitability. In fact, the growth in the number of accounts per household over time is largely a function of the same mix-shift effect that causes balances to appear to grow. Besides, the number of accounts is obviously irrelevant without a corresponding increase in balances; having more accounts without an increase in balances is more likely to reduce profitability than increase it.
Reason No. 4: Cost of serving customers may drop as customers better understand institution (and vice versa). This reason is cited but referred to as "largely undocumented." Even if correct, this effect could clearly not be sufficient to explain profit improvements of the magnitude claimed.
Reason No. 5: Loyal customers' conversations yield referrals. Again, if true, this could be a plus, but not enough to make a big difference. In the data cited by the Council on Financial Competition, the profit improvement attributed to this effect is roughly equal to the year 0 negative cash flow. The implication is that a loyal customer a responsible for bringing in one representative new customer per year, at no cost to the bank. If this were true, the rate of growth would be astronomical - like a chain letter.
With an 80% retention rate, nearly 50% of all customers would be loyal (year four or older) customers. If 50% of all customers each bring in one new customer per year, the rate of growth would be 30% per annum from this source alone (50% from the new referrals, less 20% attrition on the original customer base).
Reason No. 6: Loyal customers are less price sensitive. Meaning what? That the longer customers stay with you, the less price sensitive they become? If that were true, you could raise prices - but banks don't do that. Does it mean that the only customers who are, in fact, less price sensitive to begin with? This may well be true but it is another mix-shift phenomenon.
These reasons are thin. If they were all true, the only one that could possibly explain results of the magnitude claimed is the second: Account balances grow across time. In fact, they do not. They only appear to grow crop of new customers ages, it gets more exclusive and more profitable with the departure of the unprofitable members of the group.
Why Retail is Tough
Many bankers are reluctant to accept this fact. In the absence of relationship profitability reporting, it is easy to overlook it. Yet a macro analysis confirms it. According to an Oliver, Wyman analysis of Federal Reserve data, the median U.S. household represents a total potential of about $275 per year in available revenue (net of interest paid) from deposit and savings products. But a typical retail bank will spend from $250 to $300 per year per relationship on operational and maintenance activity related to deposit and savings products. Since most households do not even concentrate their banking business with a single institution, there is clearly a large fraction of the population that represents a losing proposition for its banks.
There are three primary implications for banks of the great profit skewness that is found in their retail relationships.
1. Pricing. In general, fees and spreads must be raised for many customers and probably reduced for a few. The prevalence of indiscriminate fee-waiving must be cut back: Some banks waive as many fees (outside of authorized programs) as they collect. Some will argue that stiffer fees and spreads will only send depositors to mutual funds. On the one hand, so be it. Mutual funds provide a terrific set of customer values for a modest fee. Why offer subsidies to customers to stay in other products if those subsidies have to be so large that you consistently operate at a loss? On the other hand, the greatest degree of defection to mutual funds that has occurred to date is probably from A customers, who are subsidizing the remaining customer base. In any move to restructure fees and rates, this A group should probably get price relief (or better service) and so may become less inclined to move to mutual funds.
2. Marketing. In a world of A, B, and C customers, marketing has two jobs. First, use sales efforts to attract new A customers. The A, B, C mix of new customers should be richer in As than the current portfolio. An expected value approach is invaluable to the marketing function in doing this job. In practice, this means using a combination of product cash-flow models and empirical data on customer behavior and characteristics. The models show hypothetical present values and reveal which customer-behavioral characteristics have the greatest effect upon present value. An analysis of variations in those behavioral characteristics then reveals the richer sources of As.
If marketing is to implement the sales side of its A-rich approach, it must use three major techniques. In order of importance, they are:
Branch location policy. For deposit and savings products, branch location policy is the single most important determinant of new customer acquisition.
Account acquisition intiatives. These include credit card solicitations, mortgage originators calling on real estate companies.
Local promotion. This should be conducted in and around existing branch locations.
Reorienting the Cross-Sell
The second job for marketing is incremental sales to existing customers - cross selling. Cross-selling takes on a totally new dimension when you recognize the profitability skewness in customer relationships. An analysis of customer-account profitability within relationships is extremely revealing. If a relationship has multiple product accounts, there is a very high degree of correlation between the profitability status of each account (i.e., its A, B, C status) and that of the relationship. In other words, if I told you that a customer had two products and was a C customer for one of them, you could conclude with a probability that ranges from 0.85 to 0.99 (depending on the products) that he would be a C customer for the other product and that he would be a C relationship overall. A similar prediction can be made for As.
Selling products to C customers is, therefore, a lose-lose proposition. It costs you to promote the sale. If you get a response, you lose on the new piece of the relationship.
This suggests strongly that indiscriminate cross-selling is inefficient. It also suggests that the use of cross-sell marketing to raise the number of products per household and thereby change Cs into As is a singularly misdirected effort. Narrowly focusing on the As for cross-sell purposes is likely to be far more economically rewarding.
There is a small proportion of C customers who are really As. You simply have a fraction of their total relationship. But this small subgroup can be filtered out reasonably easily via analysis. Mass-mailing the Cs, hoping to convert a few to As, is an activity with a negative present value. It is also a tactic heavily promoted by mail-houses posing as objective advisers!
Both of marketing's jobs (A-rich de novo selling and highly selective cross-selling) will tend to cut the amount of marketing and sales effort substantially while focusing it and improving the economic results enormously.
3. Customer Management. Rather than treat all customers the same once they come through the door, a bank should deliver a level of service that is graduated according to the needs and the revenue potential of the customers.
As noted earlier, banks have embraced service quality with a vengeance. business schools teach service industries. Pop management authors preach on the subject, and business periodicals keep gossipy success stories circulating.
What is missing from the enthusiasm over service quality is an economic foundation. The implicit assumption is that investments in service quality (whether capitalized or expensed) necessarily pay off in some mysterious way. One bank enshrined its position on service quality as follows:
"We will differentiate ourselves by delivering service that is more than the customer expects, greater than the situation warrants."
This is, of course, one of the ways in which the retention argument appeals to a management fantasy. It says that raising service quality will stop defection, which will raise profits. The cost of raising service quality is not detailed, but with 85% profit gains to be made one would presumably be churlish even to ask about cost. The proponents of service quality (among them one of the authors of the HBR article) are delighted to see a quantitative justification for an argument that had previously relied mainly upon anecdote and a kind of misguided humanism for its persuasive power.
Putting service quality on a somewhat firmer economic foundation can be achieved via four tactics, all of which rely upon a well-conceived management information system that can measure customer relationship profitability accurately and report it quarterly:
* Establish discrete service levels in steps that range from bare-bones banking to, if necessary, full-service private banking.
* Manage the average cost-to-serve at each level to match the revenue potential of the customers who qualify for the particular level of service; e.g., by controlling the quantity and caliber of relationship management time per customer.
* Move customers among service levels, up or down, based on their actual economic performance, i.e., on their A, B, or C status. For example, a relationship that is a persistent C in a given service level should either be moved to a lower cost-to-serve method of serving or be repriced.
* Implement some form of retrospective pricing for customers in service levels with highly variable customer service needs. Total fees would be set in a period to reflect the relationship's return in prior periods.
Retention's Reduced Role
It is within the rubric of customer management that customer retention has a role to play. You definitely do not want to lose A customers if you can avoid it. Avoiding it may involve changing the level, quality, and most of service to these customers. It may involve fee and rate changes. Overall, it may lessen the total contribution from these As, but it will also tend to keep them.
The underlying fact that gives rise to these observations is the extreme profit skewness of retail bank customers. The economic significance of the implications and actions outlined briefly above is truly enormous. Customer retention has an important but relatively minor role to play in all of this.
Obviously, any customer retention effort that makes the assumption that all customers are equally profitable, or even that all customers are profitable, will not be very productive and could be downright damaging.