Mismanagement of Customer Loyalty

TheMismanagem ,

"^CustomerLoyalty

by Werner Reinartz and V. Kumar

Much of the common wisdom about

customer retention is bunk. To get

strong returns on relationship programs,

companies need a clearer understanding

of the link between loyalty and profits.

86

HE BEST CUSTOMERS, We're to!d, are loyal ones.3 They cost less to serve, they're usually willing to

-X pay more than other customers, and they oftenact as word-of-mouth marketers for your company. Winloyalty, therefore, and profits will follow as night followsday. Certainly that's what CRM software vendors-and thearmies of consultants who help install their systems-areclaiming. And it seems that many business executivesagree. Corporate expenditures on loyalty initiatives arebooming: The top 16 retailers in Europe, for example, col-lectively spent more than $1 billion in 2txx>. Indeed, forthe last ten years, the gospel of customer loyalty has beenrepeated so often and so loudly that it seems almost crazyto challenge it.

But that is precisely what some of the loyalty move-ment's early believers are starting to do. Take the case ofone U.S. high-tech corporate service provider we studied.Back in 1997, this company set up an elaborate costing

HARVARD BUSINESS REVIEW

scheme to track the performance of its newly institutedloyalty programs. The scheme measured not only directproduct costs for each customer but also all associated ad-vertising, service, sales force, and organizational expenses.After nmning the scheme for five years, the company wasable to determine the profitability of each of its accountsover time. Executives were curious to see just what pay-off they were getting from their $2 million annual invest-ment in customer loyalty.

The answer took them by surprise. About half of thosecustomers who made regular purchases for at least twoyears-and were therefore designated as "loyal"-barelygenerated a profit. Conversely, about half of the mostprofitable customers were blow-ins, buying a great deal ofhigh-margin products in a short time before completelydisappearing.

Our research findings echo that company's experience.We've been studying the dynamics of customer loyalty

using four companies' customer databases. In additionto the high-tech corporate service provider, we studied alarge U.S. mail-order company, a French retail food busi-ness, and a German direct brokerage house. Collectively,the data have enabled us to compare the behavior, reve-nue, and profitability of more than 16,000 individual andcorporate customers over a four-year period.

What we've found is that the relationship between loy-alty and profitability is much weaker-and subtler-thanthe proponents of loyalty programs claim. Specifically, wediscovered little or no evidence to suggest that customerswho purchase steadily from a company over time are nec-essarily cheaper to serve, less price sensitive, or particu-larly effective at bringing in new business.

Indeed, in light of our findings, many companies willneed to reevaluate the way they manage customer loyaltyprograms. Instead of focusing on loyalty alone, companieswill have to find ways to measure the relationship between

JULY 2002 87

The Mismanagement of Customer Loyalty

loyalty and profitability so that they can better identifywhich customers to focus on and which to ignore. Herewe present one way to do that-a new methodology thatwill enable managers to determine far more preciselythan most existing approaches do just when to let go ofa given customer and so dramatically improve the returnson their investments in loyalty. We'll also discuss strate-gies for managing relationships with customers who havedifferent profitability and loyalty profiles. Let's begin,though, by reconsidering the evidence for the link be-tween loyalty and profitability.

Is Loyalty Profitable?To answer this question, we looked at the relationshipbetween customer longevity and companies' profits. Weexpected to find a positive correlation, so our real ques-tion was how strong would it be. A perfect correlation(that is, 1) would mean that marketers could confidentlypredict exactly how much money there was to be madefrom retaining customers. The weaker the correlation(the closer it was to zero), the looser the association be-tween profits and customer tenure.

The results were hardly a ringing endorsement of theloyalty mantra. The association was weak to moderatein all four companies we studied, with correlation coeffi-cients of 0.45 for the grocery retailer, 0.30 for the corpo-rate service provider, 0.29 for the direct brokerage firm,and just 0.20 for the mail-order company.

But did the weakness of the overall correlation be-tween profitability and longevity conceal any truth in thespecific claims about the benefits of loyal customers? Tofind out, we tested the three claims usually advanced byloyalty advocates, the ones we started with at the begin-ning of this article: that loyal customers cost less to serve,that they are willing to pay more for a given bundle ofgoods, and that they act as effective marketers for a com-pany's products. We tested each of these hypotheses forall four companies by looking at several cohorts of cus-tomers at each who had begun doing business at the sametime, tracking the profitability of each member of eachgroup. In this way, we saw how these customers' purchas-ing patterns and the level of service the companies ac-corded them evolved over time.

Claim 1: It costs less to serve loyal customers. Manyadvocates of loyalty initiatives argue that loyal customerspay their way because the up-front costs of acquiringthem are amortized over a large number of transactions.But, of course, that argument presupposes that the cus-

Werner Reinartz is an assistant professor of marketing atInsead in Fontainebleau, France. V. Kumar is the ING ChairProfessor at the University of Connecticut's School of Busi-ness Administration in Storrs, Connecticut, and the executivedirector of the school's ING Center for Financial Services.

tomers are profitable in those transactions. A more plau-sible argument forthe link between loyalty and decreasedcosts can be built on the idea that loyal customers will bemore familiar with a company's transaction processes.Since they need less hand-holding, the company shouldfind it cheaper to deal with them. Loyal-and therefore ex-perienced - customers of software products, for example,should be able to resolve problems on-line without need-ing the direct intervention of a technical assistant.

Our analysis, however, offers no evidence to back thatargument. It is certainly true that within any one com-pany, the monthly cost of maintaining a relationship withan individual customer-not just for the actual transac-tions but also for communications through mailings,telephone, and so forth-vary enormously, sometimes bya factor of 100 or more. But in none of the four companieswe tracked were long-standing customers consistentlycheaper to manage than short-term customers. In fact,the only strong correlation between customer longevityand costs that we found-in the high-tech corporate ser-vice provider-suggested that loyal and presumably expe-rienced customers were actually more expensive to serve.

That last finding isn't without precedent. There's a siz-able body of academic research documenting the oftenpoor profitability of long-standing customers in business-to-business industries. These customers, who almost in-variably do business in high volumes, know their value tothe company and often exploit it to get premium serviceor price discounts. Indeed, we discovered that in its effortsto please the regulars, the corporate service provider haddeveloped customized Web sites for each of its top 250clients. At the click of a button, these customers could ob-tain personalized service from dedicated sales and serviceteams. The maintenance of these teams, not to mentionthe Web sites, cost the company $10 million annually.

What surprised us more was the weakness of the cor-relation between customer loyalty and lower costs in theother three companies, where we had expected to findservice costs falling over time. In the mail-order company,for example, it had seemed reasonable to assume thatlong-standing, experienced patrons might be happy toswitch their purchases from the phone to the company'sWeb site, a move that would significantly reduce commu-nication costs. Yet the communication cost-to-sales ratiofor this company's long-standing clients is barely differ-ent from what it is for the newer ones; in both cases, ittook a bit more than six cents (6.3 cents versus 6.5 cents,to be exact) spent on marketing communication to gen-erate a dollar's worth of sales. It turned out that cus-tomers who processed their own orders through a Website expected lower prices, which offset any cost savingsthe company may have garnered by using a cheaper chan-nel. The disparity between the cost-to-sales ratios for re-cent and longtime customers at the French grocery chainand the German brokerage firm was also smaller than

HARVARD BUSINESS REVIEW


we had expected. At these companies, too, customers ex-pected something in retum for their loyalty.

These findings suggest that, at the very least, the linkbetween loyalty and lower costs is industry specific. Nodoubt there are industries in which the oldest customersare the cheapest to serve, but as we've shown there are alsoothers in which they are more expensive to satisfy.

Claim 2: Loyal customers pay higher prices for thesame bundle of goods. If loyalty doesn't necessarily lowercosts, then perhaps it generates revenue. Many propo-nents of the loyalty movement argue that customers whostick to one company do so because the cost of switchingto another supplier is too high. They will, therefore, bewilling to pay higher prices up to a point to avoid makingthe switch.

This argument sounds reasonable, but the logical con-clusion is less obviously so-namely, that if loyal custom-ers are worth pursuing because they'll pay higher prices,then companies will charge them higher prices. Thisseemed to us to be highly implausible in most corporatecontexts, where customers regularly guarantee greaterfrequency of purchase in return for lower prices. But wedid think it could describe many consumer markets. Mail-order customers, for instance, might well pay a little morefor using a catalog they could find their way around. In-deed, charging established customers more is the normin some industries. Credit card com- ^ ^ ^ ^ ^ _ ^ , ^panies routinely lure in customerswith promises of low initial interestrates, only to raise them later.

As we had expected, the evidencefrom the corporate service providerdid not support the claim: The long-term customers consistently paidlower prices than the newer custom- ^ ^ _ ^ ^ ^ ^ _ ên did - between 5% and 7% lower,depending on the product category. What was surprisingwas that we found no evidence that such loyal customerspaid higher prices in the consumer businesses. Indeed, wefound that like corporate clients, consumers also expect,and get, some tangible benefits for their loyalty. At themail-order company, for instance, it turned out that regu-lar customers actually paid 9% less than recent customersin one category of products. At the French grocery chain,there was no significant difference in prices paid in anyproduct category. In that case, any willingness on the partof loyal customers to pay higher shelf prices was probablycanceled out by the discounts many got from using theloyalty cards they were entitled to. At the brokeragehouse, all customers were charged the same fee - a per-centage of their trade volume-regardless of their historywith the company.

In general, then, it seems that a loyal customer -whether corporate or consumer - is actually more pricesensitive than an occasional one. A number of theories

To identify the true apostles,companies need to judgecustomers by more than

just their actions.

could explain this phenomenon. First, loyal customersgenerally are more knowledgeable about product offer-ings and can better assess their quality. That means theycan develop solid reference prices and make better judg-ments about value than sporadic customers can. This wascertainly in evidence at the mail-order company; ioyalcustomers typically would choose cheaper product alter-natives-a lower-priced blender, say-in the catalogs thanwould those who were less familiar with the company.

Perhaps more fundamental, though, is the fact thatcustomers seem to strongly resent companies that try toprofit from loyalty. Surveys consistently report that con-sumers believe loyal customers deserve lower prices. Thismay well explain why U.S. telecom companies, which rou-tinely offer customers special deals initially only to raiseprices later, all experience high rates of customer chum.Finally, it's just plain impossible these days to get awaywith price differentiation for any length of time. Remem-ber how close Amazon came to destroying its brand whenit attempted to charge different prices to different cus-tomers for the same DVDs.

Claim 3: Loyal customers market the company. Theidea that the more frequent customers are also the stron-gest advocates for your company holds a great attractionfor marketers. Word-of-mouth marketing is supremelyeffective, of course, and many companies justify their^ ^ ^ ^ ^ ^ ^ ^ ^ investments in loyalty programs by

seeking profits not so much fromthe loyal customers as from the newcustomers the loyal ones bring in.

To test whether regular custom-ers of the French grocery chain wereactually more effective marketersthan infrequent customers, we asked

^ ^ ^ ^ ^ ^ ^ ^ ^ a sample of the company's custom-ers two questions. First, to gauge the

extent of passive word-of-mouth marketing, we inquiredwhether they named the company when asked to recom-mend a particular grocery retailer. Then, to measure thelevel of active word-of-mouth marketing, we asked whetherthey ever spontaneously told friends or family about pos-itive experiences with the company. We then identifiedevery customer's actual level of loyalty, as measured by hisor her recorded purchase behavior (that is, how often, howmuch, and how many different sorts of items were pur-chased). Finally, we solicited their own subjective mea-sure of loyalty, their "attitudinal loyalty," in a telephonesurvey, in which we asked them if they felt loyal to thecompany, how satisfied they were with it, and whetherthey had any interest in switching to another company.

Overall, the link between customer longevity and thepropensity to market by word-of-mouth was not thatstrong. But when we looked at attitudinal and actualloyalty separately, the results were intriguing. Customersof the grocery chain who scored high on both loyalty

2002 89


measures were 54% more likely to be active word-of-mouth marketers and 33% more likely to be passive word-of-mouth marketers than those who scored high on be-havioral loyalty alone. The results from a survey of thecorporate service provider's customers produced similar ifless striking results: Customers who exhibited high levelsof both behavioral and attitudinal loyalty were 44% morelikely to be active marketers and 26% more likely to bepassive word-of-mouth marketers.

Although it's perhaps not surprising that people whotalk more positively about a company are also more likelyto sell others on the company, our findings are importantfor loyalty managers because most measure loyalty purelyon the basis of purchasing bebavior and do not conductattitudinal surveys like ours. But if managers are invest-ing in a loyalty program for its supposed marketing ben-efits, then they are looking at a potentially misleadingindicator. Customers may well buy all their groceries atthe same supermarket out of inertia and convenience. Toidentify the true apostles, companies need to judge cus-tomers by more than Just their actions.

Knowing When to Lose a CustomerOur empirical findings are clear-cut. The link betweenloyalty and profits is weaker than expected, and none ofthe usual justifications for investing in loyalty stands upwell to examination. But that doesn't mean we believeinvestments in loyalty are doomed. In our opinion, thereason the link between loyalty and profits is weak hasa lot to do with the crudeness of the methods most com-panies currently use to decide whether or not to maintaintheir customer relationships.

The most common way to sort customers is to scorethem according to how often they make purchases andhow much they spend. Many tools do that; one of themost familiar is called RFM (which stands for recency,

frequency, and monetary value). Mail-order companies inparticular, including the one in our study, rely on this toolto assess whether a customer relationship merits furtherinvestment.

To understand how RFM and methods like It work, let'simagine for the sake of simplicity that a company focuseson just two dimensions, recency and frequency of pur-chase. This company measures recency by finding outfrom its database if the customer bought something inthe iast six months, between six months and a year ago,or more than a year ago, assigning a higher score the morerecent the purchase. It then measures how frequentlythe customer made purchases in each of those three timeframes-twice or more, once, or never-assigning a scorein a similar way. Then it adds the two scores together. Ingeneral, the more items a customer purchases and themore recent the transactions, the higher tbe overall scoreand the more resources the company lavishes on the per-

son. In actual application, many companies weight thescores in favor of recency.

Unfortunately, our study of the mail-order companysuggests that scoring approaches of this kind result in asignificant overinvestment in lapsed customers. Takea look at the graph "The Cost of Keeping Customers On."It plots the profits earned from one particular segmentof customers - those who tumed out to have purchasedvery intensively for a brief period and then never again.The profits from those customers were tracked for 36montbs-the full time the company treated them as activecustomers because their initial high volume of purchaseskept their RFM scores high even after they'd stopped buy-ing. As the graph shows, the company started to incurlosses on these customers after about 20 months. We es-timate tbat the total cost to the company of misinvest-ments of this kind amounted to about $1 million a year.

So just why is RFM such a ptx)r way to measure loyalty?One problem is that pattems of buying behavior for fre-quently bought goods are quite different than those forinfrequently bought goods. But RFM can't distinguish be-tween them-that is, it ignores the pacing of a customer'sinteractions, the time between each purchase. To under-stand the importance of pacing, imagine that your com-pany has two hypothetical customers - Mr. Smitb andMs. Jones - who both start to buy goods in month one.Over the course of the first year, they purchase at differ-ent rates: Smith buys after short intervals, making pur-chases again in the second, sixth, and eighth months,whereas Jones takes far longer to buy again, waiting a fullseven months before she purchases again in month eight.

A simple RFM evaluation might suggest that Smithis likely to be more loyal than Jones-his purchases aremore frequent and as recent and thus more deservingof investment But an RFM evaluation would fail to takeinto accoimt the fact that Smith usually buys something,on average, every 2.3 months, and yet by month 12 hehasn't bought anything for four months. Jones, too, hasn'tbought anything for four months. But she normallydoesn't buy anything for seven months, so she's wellwithin her historic range. On that basis, the probabilitythat Jones will purchase in the future is actually higberthan it is for Smith, so she is more likely to be a safe betfor further investment.

The model of buying behavior we're describing here isa special case of "event-history modeling," a statisticaltechnique witb a long and strong history. Like most sta-tistical models, it figures the probability that some futureevent will occur based on statistical pattems observed ei-ther theoretically or empirically in the past. Other exam-ples of event-history modeling are the occurrence of hur-ricanes over time and tbe recurrence of diseases within apopulation. In our case, the "event" is purchasing, and thepast pattems are taken from the empirical data our fourcompanies have collected in their customer databases.

90 HARVARD BUSINESS REVIEW


There are more and less complex ways to use the event-history model to compute the probability that a customerwill keep on purchasing. But in its simplest form, the for-mula is t", where, for Smith in our example, n is the num-ber of purchases he made in the entire time period (inthis case, the whole year), and f is the fraction of the pe-riod represented by the time between his first purchaseand his last one.

Let's use the formula to assess the probabilities thatSmith and Jones will each remain active, that is keep onmaking purchases. Smith has made four purchases, thelast being in month eight, so n is 4 and t is 8+12 or 0.6667-That makes Smith's probability of still being active(0.6667)", or 0.198. In other words, there's about a 20%probability that Smith will keep on purchasing. Jones alsomade her last purchase in month eight, so her t is 0.6667as well, but she bought only twice, so her probability is(0.6667)^, or 0.444, nearly 45%. Jones, therefore, is morethan twice as likely as Smith to remain an active cus-tomer. Unlike RFM, this approach is particularly good atpredicting how quickly a customer's purchasing activitywill drop off, as the probability of their being active in thefuture drops steeply with time, so it clearly has the po-tential to prevent heavy overinvestment in profitable butdisioyal customers.

In practice, of course, our calculations are much moresophisticated than the foregoing example and can takeinto account any number of variables, including demo-graphics, amount of spending, and type of products pur-

chased.^ Given enough historical data, we can estimatethe probability of another purchase out into several fu-ture time periods. But no matter how complex the soft-ware a company uses to do the math, the analysis is veryeasy to implement, since all such probability models de-pend on just three simple pieces of information that anycustomer database stores: When did the customer buy forthe first time? When did she purchase last? and When didshe purchase in between?

The second main drawback of scoring methods likeRFM is that the monetary-value component is almostalways based on revenue rather than profitability. Forexample, the mail-order company classifies the revenuegenerated from a customer into the following four cate-gories: $50 or less, $51 to $150, $151 to $300, and morethan $300. But the decision to continue investing in cus-tomer relationships needs to be based on customers'profitability, not the revenue they generate. The cost ofservicing customers who buy only small quantities of low-margin products may exceed the revenue they bring in.That profile tumed out to fit fully 29% of the mail-ordercompany's customers.

Instead of looking at revenue, therefore, we will need toincorporate profitability into our probahility calculation.Specifically, we need an estimate of the average profitearned on each customer in any typical purchase period.For Smith and Jones, that's the average monthly profit fig-ure, but the choice ofthe time period is generally drivenby an industry's natural purchase cycle. In the mail-order

The Cost of Keeping Customers On

Just because a group ofcustomers was profitable inthe pa5t, doesn't mean it willcontinue to be so in the future.Many nonloyal customers canbe very profitable initially,causing companies to chaseafter them in vain for futureprofits. Such is the case illus-trated in this graph, whichtracks the profitability of thatsegment of one company's cus-tomers. Once these customersceased their buying activity,they became unprofitable be-cause the company kept invest-ing in marketing to them.

profit in$thousands

-2

month 18 24 30 36

JULY 2002 91


business, marketers think in terms of months or quarters.In retailing, the period is a week.

An estimate of per-period profitability is not hard to ob-tain, especially in today's information-rich age. Our cor-porate high-tech service provider, for example, was easilyable to calculate the historical profitability of each of itscustomers from its sales data, and we have been able tocalculate the profitability of individual customers for theother companies we studied as well. To estimate a cus-tomer's future profitability, you simply multiply his aver-age periodic profit figure by the number we previouslycalculated, the probability that the customer will still beactive at the end of that period.

To see how this works, consider how a simple version ofour approach could help the high-tech corporate serviceprovider decide whether and how to invest in two ongo-ing customer relationships during the next year. From itssales data, the provider determines that the first account,Adam Incorporated, yielded an average profit of $5,500per quarter over the last two years, while the second ac-count. Eve Limited, yielded an average profit of $i,cxx>.Using the formula, we estimate the probability that AdamIncorporated will remain active is 85% in the first quarter,60% in the second, 35% in the third, and only 22% in thefourth. Probabilities for Eve Limited are only slightlylower, starting at 80% in the first quarter and declining to

50%, 27%, and 15% in the subsequent quarters. For each ac-count, we now multiply the probability figure for each pe-riod by the historical average profit number, and the sumof those calculations gives us the estimated profit in dol-lars for each customer over the next year.

Both accounts are clearly profitable: Adam Incorpo-rated is likely to generate $11,110, while Eve Limited willlikely produce $1,720. But how much should the companyinvest in maintaining each relationship so that it will ac-tually deliver the numbers? Given that a visit by the fullsales team costs our company $5,000 and a single sales-person's visit costs $2,000, it's clear that Adam Incorpo-rated merits the full treatment. Eve Limited, however,doesn't deserve even a single visit. Ifthe account stays ac-tive, that's obviously good news, but it's not worth ourcompany's time to chase the sale. Even loyal and prof-itable customers don't always deserve to be courted.

When tested on real customer databases, our approachproduced a nuanced picture of the relationship betweenprofitability and loyalty. About 40% of the service pro-vider's profitable customers turned out to be not worthchasing, being unlikely to buy anything in the future, andalmost the same percentage of the loyal ones were un-profitable. Fully 30% turned out to be neither profitablenor loyal. (See our results for all four companies in thechart "Which Customers Are Really Profitable?").

Which Customers Are Really Profitable?

When customers are sortedaccording to their profitabilityand longevity, it becomes clearthatthe relationship betweenloyalty and profits is by nomeans assured. Here, a sizablepercentage of long-standingcustomers in all four compa-nies are only marginally prof-itable, whereas a large percent-age of short-term customersare highly profitable. It is thesesegments that drive down theoverall correlation betweenloyalty and profitability.

Highprofitability

Lowprofitability

corporate serviceprovider

grocery retail

mail-order

direct brokerage


grocery retail

mail-order

direct brokerage

percentageof customers

20%

15%

19%

18%


29%

34%

29%

33%


grocery retail

mail-order

direct brokerage


grocery retail

mail-order

direct brokerage


30%

36%

31%

32%


21%

15%

21%

17%

Short-termcustomers

Long-termcustomers



As valuable as segmentation is, even more valuable iscorrect identification at the individual level. Knowing that60% of your loyal customers are profitable is useless if youdon't know which ones to court with what level of service.At the corporate service provider, for example, we wereable to predict how profitable and how loyal any particu-lar customer would be with 30% more accuracy than weobtained using traditional methods like RFM. That kindof misinformation carries a high price. Our mail-ordercompany, for instance, was sending mailings to pieople itshould have ignored, ignoring people it should have beencultivating, and sending the wrong material to people.

From Measurement to ManagementSo what is the next step? After analyzing your custom-ers' profitability and the projected duration of their re-lationships, you can place each of them into one of fourcategories, as shown in the matrix "Choosing a LoyaltyStrategy." Now, what kind of relationship managementstrategies should you apply to the different segments? Forthe customers who have no loyalty and bring in no prof-its-we call them "strangers"-the answer is simple: Iden-tify early and don't invest anything. But for customers inthe other three quadrants, the choice of strategy willmake a material difference to the segment's profitability.

We've found that the challenge in managing customerswho are profitable but disloyal - the "butterflies"- is tomilk them for as much as you can for the short time theyare buying from you. A softly-softly approach is more ap-propriate for profitable customers who are likely to beloyal-your "true friends." As for highly loyal but not veryprofitable customers-the "bamacles"-the emphasis hasto be on finding out whether they have the potential tospend more than they currently do.

Turning True Friends into True Believers. Profitable,l(âl customers are usually satisfied with existing arrange-ments. At the mail-order company, for instance, we foundthat they tended to retum goods at a relatively high rate,refiecting their comfort in engaging with the company'sprocesses. They are also steady purchasers, buying regu-larly, but not intensively, over time.

In managing these true friends, the greatest trap isoverkill. At the catalog company, for instance, we foundthat intensifying the level of contact through, for exam-ple, increased mailings, was more likely to put off loyaland profitable customers than to increase sales. Peoplefiooded with mail may throw everything out withoutlooking at it. Sent less mail, however, they are more likelyto look at what they get. Indeed, the mail-order companyfound that its profitable, loyal customers were not amongthose who received the most mailings.

Choosing a Loyalty Strategy

When profitability and ioyalty

are considered at the same

time, it becomes clear that

different customers need to

be treated in different ways. Highprofitability

Lowprofitability

Butterflies

• good fit between company'sofferings and customers' needs

• high profit potential

Actions:• aim to achieve transactional

satisfaction, not attitudinal loyalty

• milk the accounts only as long asthey are active

• key challenge is to cease investingsoon enough

Strangers

• little fit between company'sofferings and customers' needs

• lowest profit potential

Actions:* make no investment in these

relationships

• make profit on every transaction

True Friends

• good fit between company'sofferings and customers' needs

• highest profit potential

Actions:• communicate consistently but

not too often

• build both attitudinal andbehavioral loyalty

• delight these customers tonurture, defend, and retain them

Barnacles

• limited fit between company'sofferings and customers' needs

• low profit potential

Actions:• measure both the size and share

of wallet

• if shareof wallet is low, focus onup-and cross-selling

• if size of wallet is small, imposestrict cost controls

Short-termcustomers

Long-termcustomers

I ULY 2002 93


What's more, companies need to concentrate on find-ing ways to bring to the fore their true friends' feelingsof loyalty, hecause "true believers" are the most valuablecustomers of all. At the grocery retailer, for example, wefound that customers who scored high on both actual andattitudinal measures of loyalty generated 120% moreprofit than those whose loyalty was observed throughtransactions alone. It wasn't just a business-to-consumerphenomenon, either: Those of the corporate service pro-vider's customers who exhibited loyalty in both thoughtand deed were 50% more profitable than those who ex-pressed their loyalty through action alone.

Companies can do several things to make loyal cus-tomers feel rewarded for their loyalty. The French gro-cery chain lets loyal customers optin to e-mailings of special recipes,price promotions, and the like. It alsogrants them preferred access to com-pany-sponsored seasonal events. Forinstance, they get exclusive early ac-cess to semiannual, weeklong winefestivals in which they get to buymany of the better wines, which areavaiiabie only in limited quantities.Such measures are already having ^ ^ — ^ ^ ^ ^ ân appreciable impact on the pur-chasing volumes and profitability of loyal customers.

Enjoying Butterflies. The next most valuable groupcomprises customers who are profitable but transient, andsome industries are full of these kinds of purchasers.For instance, many of the direct brokerage company'smost valuable customers were what it called "movers,"investors who trade shares often and in large amounts.Aware of their value as customers, these people enjoyhunting out the best deals, and they avoid building a sta-ble relationship with any single provider.

The classic mistake made in managing these accountsis continuing to invest in them after their activity dropsoff. Any such efforts are almost invariably wasted; ourresearch shows that attempts to convert butterflies intoloyal customers are seldom successful - the conversionrate was 10% or lower for each ofthe four companies westudied. Instead of treating butterflies as potential truebelievers, therefore, managers should look for ways toenjoy them while they can and find the right momentto cease investing in them. In practice, this usually meansa short-term hard sell through promotions and mailingblitzes that include special offers on other products, anapproach that might well irritate loyal customers. Thecorporate service provider, for instance, telephones thoseit has identified as butterflies four or five times shortlyafter their most recent purchase and follows up with justone direct mailing six to 12 months later, depending onthe product category. If these communications bear nofruit, the company drops contact altogether.

No company should ever take

for granted the idea that

managing customers forloyalty is the same as

managing them for profits.

Smoothing Barnacles. These customers are the mostproblematic. They do not generate satisfactory returns oninvestments made in account maintenance and market-ing because the size and volume of their transactions aretoo low. Like barnacles on the hull of a cargo ship, theyonly create additional drag. Properly managed, though,they can sometimes become profitable.

The first step is to determine whether the problem isa small wallet (the customers aren't valuable to beginwith and are not worth chasing) or a small share of thewallet (they could spend more and should be chased).Thanks to modem information technology, which makesit possible to record the spending pattems of individuals,this is much less of a challenge than it once was. Our^ ^ ^ ^ ^ ^ ^ ^ _ _ French grocery chain, in fact, does

it rather well. By looking closely atPOS data on the type and amountof products that individuals pur-chase (say, baby or pet food), thecompany derives amazingly reli-able estimates ofthe size and shareof the individual customers' wal-lets it has already captured in eachproduct category. Then, a company

- ^ ^ ^ - ^ ^ ^ ^ — can easily distinguish which loyalcustomers are potentially profit-

able and offer them products associated with those al-ready purchased, as well as certain other items in seem-ingly unrelated categories. For instance, our corporateservice provider might sell add-on software or memoryupgrades for previously sold systems. Our mail-order com-pany might send a do-it-yourself catalog to a customerwho had previously bought a kitchen appliance.

There is no one right way to make loyalty profitable. Dif-ferent approaches will be more suitable to different busi-nesses, depending on the profiles of their customers andthe complexity of their distribution channels. But what-ever the context, we believe that no company should evertake for granted the idea that managing customers forloyalty is the same as managing them for profits. The onlyway to strengthen the link between profits and loyalty isto manage both at the same time. Fortunately, technologyis making that task easier every day, allowing companiesto record and analyze the often complex, and sometimeseven perverse, behavior of their customers. ^

1. A complete explanation of the actual model we use can be found in ourarticle, "On the Profitability of Long-Life Customers in a NoncontractualSetting: An Empirical Investigation and Implications for Marketing,"of Marketing (Octot)er 2000).

Reprint R0207FTo order reprints, see the last page of Executive Summaries.

To further explore the topic of this article, go towww.explore.hbr.org.


Documents

Mismanagement of Customer Loyalty