Eugene W. Anderson, Claes Fornell, & Donald R. Lehmann

Customer Satisfaction, MarketShare, and Profitability:Findings From Sweden

Are there economic benefits to improving customer satisfaction? Many firms that are frustrated in their efforts to im-prove quality and customer satisfaction are beginning to question the link between customer satisfaction and eco-nomic returns. The authors investigate the nature and strength of this link. They discuss how expectations, quality,and price should affect customer satisfaction and why customer satisfaction, in tum, should affect profitability; thisresults in a set of hypotheses that are tested using a national customer satisfaction index and traditional account-ing measures of economic returns, such as return on investment. The findings support a positive impact of qualityon customer satisfaction, and, in tum, profitability. The authors demonstrate the economic benefits of increasing cus-tomer satisfaction using both an empirical forecast and a new analytical model. In addition, they discuss why in-creasing market share actually might lead to lower customer satisfaction and provide preliminary empirical supportfor this hypothesis. Finally, two new findings emerge: First, the market's expectations of the quality of a firm's out-put positively affects customers' overall satisfaction with the firm; and second, these expectations are largely ra-tional, albeit with a small adaptive component.

D i>es higher customer satisfaction lead to superior eco-nomic retums? Widespread acceptance of this relatit)n-

Khip is evident in the growing popular literature on qualityand customer satisfaction, the increasing number of consult-ing and marketing research Finns ihat promise to improve aclient's ability to satisfy customers, and—perhaps most per-suasively from a market-oriented perspective—the numberof organizations actively using some form of customer sat-isfaction measurement in developing, monitoring, and/orevaluating product and service offerings, as well as for eval-uating, motivating, and/or compensating employees.

However, at the level of the nrm. recent empirical evi-dence casts doubt on whether companies' efforts to im-prove customer satisfaction and quality through implemen-tation approaches such as total quality management (TQM)actually are having the desired effects. Specifically, severalsurveys ptiinl lo the failure of TQM to enhance either eco-nomic retums or competitiveness. A study by the AmericanQuality Foundation and Emst & Young suggests that many

Eugene W. Anderson is an Assistant Prolessor of Marketing and Claes For-nell is the Donald C. Cook Prolessor ol Business Administration. NationalQuality Research Center. School of Business Administration, The Univer-sity of Michigan, Donald R, Lehmann is the George E. Warren Professorof Marketing, Graduate School of Business, Columbia University, The au-Ihors gratefully acknowledge the data provided through Ihe funding of theSwedish Post and the support of the National Quality Research Center atthe University of Michigan Business School. Funding (or the anaiysis wasprovided by the Markeling Science Institute's Market-Driven Quality re-search competition. This work has benefitted substantially from the com-ments ol Rick Staelin, John Hauser. Bari Weitz, Roland Rust, and partici-pants in the Customer Satisfaction Workshop at the University of MichiganBusiness School, The authors thank Jaesung Cha and Jay Sinha tor theirhelp with the data.

companies are wasting their efforLs in trying to impmve qual-ity (American Quality Foundation 1992). The consultingfirms of A.T Keamey and Arthur D. Little present equallydisappointing fmdings in two separate studies: (1) 80% ofmore than 100 British firms reptirled "no significant im-pact a.s a result of TQM" and (2) almost two-thirds of 500U.S. companies saw "zero competitive gains" (The Econo-mist 1992).

If frustration with attempts to improve quality leadsmany business firms to abandon the Quality MovementiNew.sweek 1992). the recent surge of interest in customersatisfaction is likely to follow the same path—unless it canbe demonstrated (hat there are positive economic retums toimproving customer satisfaction. Firms will appropriate re-sources for improving customer satisfaction only if the ef-fects are of sufficient size, as measured b̂ * traditional ac-counting methods.

In view of these facts, it is not surprising that there is re-surgent interest in understanding the links between quality,customer salisfaction, and firm performance (e.g.,. eco-nomic returns).' In a meta-analysis of sLralegy variables.Capon, Farley, and Hoenig (1990) identify 20 studies thatfind a positive relationship between quality and economic re-turns. For example, Buzzell and Gale (1987) and Phillips,Chang, and Buzzell (1983) each report a significant relation-ship between relative qualiiy—as perceived by the businessunit—and retum on investment (ROI) for firms representedin the PIMS database. In the last few years, researchershave started to elaborate on the process by which delivering

'In marketing, customer salisfailion l()ng has been recognized as a cen-tral concept, as well as an important goal uf all business activiiy. Satisfac-tion is a core concept in the American Marketing Association's ofl'icialdefinition of markeling.

Journal of MarketingVol. sa (July 1994), 53-e6 Customer Satisfaction, Market Share, and Profitability / 53

high-quality goods and services influences profitabilitythrough customer satisfaction. Building from the individual-level model of customer satisfaction proposed by Oliver(1980), several studies discuss and/or observe a strong linkbetween customer satisfaction and loyalty (Anderson andSullivan 1993; Bearden and Teel 1983; Boulding et al.1993; Fomeli 1992; LaBarbera and Mazursky 1983; Oliverand Swan 1989). Reichheld and Sasser (1990) discuss whyincreasing customer loyalty .should lead to higher pmfitabil-ity. Rust and Zahorik (1993) empirically demonstrate therelationship between customer satisfaction and profitabilityfor a health care organization.

Our purpose is to examine more closely the links be-tween customer-based measures of firm performance—such as customer satisfaction—and traditional accountingmeasures of economic returns. Although there have been afew firm-specific studies (e.g.. Rust and Zaborik 1993). thisarticle represents the first large-scale examination of therelationship.

A unique feature of our empirical work is ihe set of cus-tomer-based performance measures for firms participatingin the Swedish Customer Satisfaction Barometer (SCSB)(see Fornell 1992 for a description). The SCSB providesyearly firm-level indices of quality, expectations, and over-alt customer satisfaction for major competitors in a varietyof product and service industries. Imptirtantly. each firm'sset of indices is an estimate based on an annual survey ofcurrent customers rather iban a set of unstandardized num-bers drawn from multiple "independent" sources (e.g.,trade press, consumer advocates) or based on an internal,self-reported measure of quality. The SCSB provides a stan-dard set of customer-based performance measures thai canbe matched to economic performance measures, such as mar-ket sbare and ROI.

Prediction of economic returns is one of the central pur-poses of tbe SCSB. The index is constructed using a meth-odology that maximizes the relationship lietween customersatisfaction and the likelihood of repeal purchase. It is im-portant to note Ihat this methodology distinguishes theSCSB measures from other common approaches used tocombine tbe facets of cuslomer satisfaction into a singleindex^unit weighting schemes or some variation of factoranalysis (e.g.. the J.D. Power Index for automobiles). Thelogic behind the SCSB methodology is to derive theweights with respect to a proxy for economic returns (e.g.,customer loyalty), providing a better chance of predicting ac-tual economic returns (Fornell 1992).

We begin by defining and discussing the links betweenquality, expectations, customer satisfaction, and profitabil-ity, as well as the relationship between customer satisfac-tion and market share. Next, the data and methodology arediscussed. Finally, we present tbe findings and discuss theirimplications.

Customer Satisfaction andProfitability

How does satisfying current customers affect profitability?How do market expectations and experiences affect cus-

tomer satisfaction? Tn this section, we develop a conceptualframework linking customer-based measures of firm perfor-mance (e.g., customer satisfaclion) witb traditional account-ing measures of economic returns, sucb as ROI.

Before proceeding, it is important to make clear whal ismeant by "customer satisfaction" in the context of thisstudy. Al lea.st two different conceptualizations of customersatisfaction can be distinguished: transaction-specific and cu-mulative (Boulding et al. 1993). From a transaction-specificperspective, customer satisfaction is viewed as a post-choice evaluative judgment of a specific purchase occasion(Hunt 1977; Oliver 1977. 1980. 1993). Behavioral research-ers in marketing bave developed a rich body of literature in-vestigating tbe antecedents and consequences of this typeof customer satisfaction at the individual level (see Yi 1991for a review). By comparison, cumulative customer satisfac-tion is an overall evaluation based on the total purchase andconsumption experience with a good or service over time(Fomeli 1992; Jobmon and Fornell 1991). Whereas transac-tion-specific satisfaction may provide specific diagnostic in-formation about a particular product or service encounter, cu-mulative satisfaction is a more fundamental indicator of thefirm's past, current, and future performance. It is cumula-tive satisfaction that motivates a Hnn's investment in cus-tomer satisfaction. Because the focus here is on the relation-ship between customer satisfaction and economic returns,our theoretical framework treats customer satisfaction ascumulative.

Whal is quality aiid how is it distinct from customer sat-isfaction? In this study, perceived quality is taken to be aglobal judgment of a supplier's current offering(Steenkamp 1989). This is similar in spirit to tbe positiontaken by Zeithaml (1988. p. 3) in summarizing an extensivereview of the literature on quality: "Perceived quality canbe defined as tbe ct)nsumer's judgment about a product'soverall excellence or superiority." However, it is worth not-ing that there are several distinct conceptualizations of qual-ity (Holbrook 1994). In marketing and economics, qualityoften has heen viewed as dependent on the level of productattributes (e.g., Hauser and Sbugan 1983; Rosen 1974). Inoperations management (e.g., Garvin 1988; Juran 1988),quality is defined as having two primar>' dimensions: (1) Fit-ness for use—Does the product or service do what it is sup-posed to do? Does it possess features tbat meet tbe needs ofcustomers? and (2) Reliability—To what extent is the prod-uct free from deficiencies? In the services literature in mar-keting, quality is viewed as an overall assessment (e.g.. Par-asuraman. Zeiihaml. and Berry 1985). Service quality inthis context is helieved to depend on gaps between deliv-ered and desired service on certain dimensions.

The theoretical framework presented here views cus-tomer satisfaction as distinct from quality for several rea-sons. First, customers require experience with a product todetermine how satisfied tbey are with it. Quality, on theother hand, can be perceived without actual consumption ex-perience (Oliver 1993). Second, il has been long recognizedthat customer satisfaction is dependent on value (Howardand Sbeth 1969; Kotler and Levy 1969). where value canbe viewed as the ratio of perceived quality relative to price

54 / Joumal of Marketing, July 1994

or benefits received relative to costs Incurred (Dodds.Monroe, and Grewa! 1991; Holbrook 1994; ZeithamI1988). Hence, customer satisfaction is also dependent onprice, whereas the quality of a good or service is not gener-ally considered to be dependent on price. Third, we viewquality as it pertains to customer's current perception of agood or service, whereas customer satisfaction is based onnot only current experience but also all past experiences, aswell as future or anticipated experiences. Finally, there isample empirical support for quality as an antecedent of cus-tomer satisfaction (e.g., Anderson and Sullivan 1993;Churchill and Suprenant 1982; Cronin and Taylor 1992; For-nell 1992; OUver and DeSarbo 1988).

Overview of the Theoretical Framework

The theoretical framework developed in the remainder ofthis section can be summarized in the general set of equa-tions presented in Table I. Profitability at time t is posi-tively affected by customer satisfaction, as well as other fac-tors (e.g., past values of the dependent variable, economicconditions, rirm-specific factors, luck, error). Customer sat-isfaction, in turn, is positively affected by market expecta-tions and experiences. Finally, current market expectationsare positively related to both historical expectations and themarket's experiences with quality in the most recent period.The nature of each of these relationships is discussedsubsequently.

How Does Customer Satisfaction AffectProfitabiiity?

Fornell (1992) enumerates several key benefits of high cus-tomer satisfaction for the firm. In general, high customer sat-isfaction should indicate increased loyalty for current cus-tomers, reduced price ela.sticities, insulation of current cus-tomers from competitive efforts, lower costs of future trans-actions, reduced failure costs, lower costs of attracting newcustomers, and an enhanced reputation for the firm. In-creased loyalty of current customers means more customerswill repurchase (be retained) in the future. If a firm hasstrong customer loyalty, it should be reflected in the firni'seconomic returns because it ensures a steady stream of fu-ture cash flow (Reichheld and Sasser 1990).

The more loyal customers become, the longer (hey arelikely to continue to purchase from the same supplier. Thecumulative value of a loyal customer to a firm can be quitehigh. For example, consider the lunch habits of three col-leagues that regularly patronize a restaurant close to theirworkplace. If the average price of a meal is $6 and the triovisits the restaurant three times a week, the annual revenuereceived by the establishment is in the neighborhood of$2,700. One hundred similarly loyal customers would beworth $90,(KK). This group would be worth ahnosl a half mil*lion dollars over the next five years—even if they all col-luded to keep the restaurant a secret from other potential cus-tomers. The net present value of the expected margin fromthese customers reflects their asset value to the restaurant. In-creasing customer satisfaction increases the value of afirm's customer assets and future profitability.

TABLE 1General Specification of the Conceptual Modei

EXPECTATIONS, = f, (EXPECTATIONS,.,,

SATISFACTION, = f2(0UALITY,, PRICE,,

PROFITABILITY, = f j (SATISFACTI0N,.C3,)

where d = vector of other factors (e.g., environmentaltrends, firm-specific factors, error)

Customer satisfaction should reduce price elasticitiesfor current customers (Garvin 1988). Satisfied customersare more willing to pay for the benefits they receive and aremore likely to be tolerant of increases in price. This implieshigh margins and customer loyalty (Reicbheld and Sasser1990). Low customer satisfaction implies greater turnoverof the customer base, higher replacement costs, and, due tothe difficuity of attracting customers who are satisfieddoing business with a rival, higher customer acquisitioncosts. Decreased price elasticities lead to increased profitsfor a firm providing superior customer satisfaction.

i4igh customer satisfaction should lower ihe costs oftransactions in the future. If a fimi has high customer reten-tion, it does not need to spend as much to acquire new cus-tomers each period. Satisfied customers are likely to buymore frequently and in greater volume and purchase othergoods and services offered by the finn (Reichheld and Sas-ser 1990).

Consistently providing goods and services that satisfycustomers should increase profitability by reducing failurecosts. A finn that consistently provides high customer satis-faction should have fewer resources devoted to handling re-turns, reworking defective items, and handling and manag-ing complaints (Crosby 1979; Garvin 1988; TARP 1979.1981).

The costs of attracting new customers should be lowerfor firms that achieve a high level of customer satisfaction(Fomell 1992). For example, satisfied customers are reput-edly more likely to engage in positive word of mouth, andless likely to engage in damaging negative word of mouth,for the firm (Anderson 1994b; Howard and Sheth 1969;Reichheld and Sasser 1990; TARP 1979, 1981). Mediasources are also more likely to convey fwsitive informationto prospective buyers, Customer satisfaction claims maymake advertising more effective, and high customer satisfac-tion may allow the firm to offer more attractive warranties.

An increase in customer satisfaction also should en-hance the overall reputation of the firm. An enhanced repu-tation can aid in introducing new products by providing in-stant awareness and lowering the buyer's risk of trial (Ro-bertson and Gatignon 1986; Schmalansee 1978). Reputa-tion also can be beneficial in establishing and maintainingrelationships with key suppliers, distributors, and potentialallies (Anderson and Weitz 1989; Montgomery 1975).Reputation can provide a halo effect for the firm that posi-tively infiuences customer evaluations, providing insulationfrom short-term shocks in the environment. Customersatisfaction .should play an important role in building other

Customer Satisfaction, Market Share, and Profitability / 55

important assets for the firm, such as brand equity (Aaker1992; Keller 1993).

The first hypothesis of the model can be stated asfollows:

H,: Customer satisfaction has a positive eflect on ecoQomicretums.

Although there are many compelling reasons to con-clude that higher customer satisfaction leads to higher prof-itabiiity. it is, nevertheless, not always the case. At somepoint there must be diminishing retums to increasing cus-tomer satisfaction. For example, many companies seek to in-crease customer satisfaction by investing in quality control.There are many economic benefits associated with this ac-tivity (e.g., less rework, iower warranty costs). Yet qualitycontrol is likely to have its greatest impact when reliabilityis relatively low. and there may come a point when thecosts associated with reducing the probability of defectswill be greater than the benefits to the firm.

There is also evidence that conformity to specificationsis not as important in determining overall customer satisfac-tion as the design of a product or service in meeting cus-tomer needs (Anderson and Sullivan 1993). Given that in-creasing quality and customer satisfaction by design (e.g..adding features, increasing the quality of raw materials and/or level of features, increasing the level of personal service,providing greater variety by differentiating the product lineto meet needs) is likely to increase costs at an increasingrate (Shugan 1989), it is likely that there are diminishing re-tums to efforts lo improve product or service quality and cus-tomer satisfaction.

Firm-Level Antecedents of Customer Satisfaction

As Table 1 indicates, customer satisfaction is affected byoverall quality, price, and expectations. At the individualcustomer level, several studies have shown that perceivedquality affects customer satisfaction (Anderson and Sulli-van 1993; Bearden and Teel 1983; Bolton and Drew 1991;Cadotte. Woodruff, and Jenkins 1987; Churchill and Supre-nant 1982; Fornell 1992; Oliver and DeSarbo 1988; Tseand Wilton 1988). This relationship is intuitive and funda-mental to all economic activiiy. Aggregated to the firmlevel, customers' current experience with a supplier's offer-ing also should have a positive influence on their overall as-sessment of how satisfied they are with that supplier.

In addition, price plays an important role in this relation-ship. The received value of a supplier's offering—that is,quality relative to price—has a direct impact on how satis-fied customers are with that supplier (Anderson and Sulli-van 1993; Fornell 1992; Sawyer and Dickson 1984; Zei-thaml 1988). Anterasian and Phillips (1988) discuss therole of value in driving overall business performance. Inboth our conceptualization and measurement of quality, itis important to consider the relationship between quaiityand price. In our empirical work, in view of the propositionthat price affects satisfaction and the possibility of confound-ing effects of a price-quality relationship—as wel! the needto compare the hedonic value of quality across categories(Lancaster 1979; Rosen 1974)—each construct is measured

relative to the other (Fornell 1992). The resulting indexmeasures the value received by customers. The expected re-lationship between quality and satisfaction can be summa-rized as follows:

H;: The current level of quality as perceived by the marketshould have a positive effect on overall customersatis facUon.

Expectations about the quality of goods and servicesalso should have a positive impact on customer satisfaction.At the aggregate level of analysis here, expectations cap-tures the accumulated knowledge of the market concerninga given supplier's quality. Just as current quality is ex-pected to have a positive influence on overall customer sat-isfaction, so should all past experiences with quality, as cap-tured by expectations. In addition, expectations contain in-formation based on not actual consumption experience butaccumulated information about quality from outsidesources, such as advertising, word of mouth, and generalmedia. Like past experience, positive Information aboutpast quality should affect customer satisfaction positively.

In addition, in forming expectations, consumers usepast experience and nonexperiential information to con-struct forecasts of the supplier's ability to deliver quality inthe future. This role of expectations is important becausethe nature of the ongoing relationship between a firm andiu customer base is such that expected future quality is crit-ical to customer satisfaction and retention as it relates tolong-term relationships with customers (Bateson 1989;Czepiel and Gilmore 1987; Gronroos 1990; Lovelock1984; Shostack 1977). In durable goods categories, cus-tomer satisfaction depends on both whether the currentlyowned product will continue to meet customer needs andthe anticipated quality of future service. In service indus-tries, client satisfaction with the vendor depends on the an-ticipated quality of future service as well as the ability oftbe service to provide for future needs. This forecast compK)-nent of expectations also argues for a positive impact of ex-pectations on satisfaction.

The preceding factors suggest that we should expect theaggregate measure of expectations used here to have a posi-tive impact on overall customer satisfaction. Although theremay be individual differences affecting expectations at theindividual level, such differences should cancel out in the ag-gregate (Katona 1979). In aggregating expectations acrosscustomers to the level of the firm, expectations should re-flect more accurately both a firm's reputation for providinghigh (or low) quality and its ability to do so in the future.

The U.S. auto industry provides an interesting exampleof the effects of expectations on customer satisfaction. Thereputation of Detroit's products suffered in the 1970s and agood portion of the 198C)s. Past negative experiences,broadly disseminated through word of mouth and mediasources, contributed to lower overall expectations with theproducts and service that accompanies them. It is likely thatoverall customer satisfaction in the late 1980s was thereforelower due to nol only customers' experiences in the 1970sand 1980s, but also anticipated lower quality. A case inpoint is the Mercury Tracer and Mazda 323. two virtuallyidentical cars. Mazda customers were more satisfied over-

56 / Joumal of Marketing, July 1994

all, ceteris paribus, because Mazda customers had higher ex-pectations than Mercury customers (e.g., continued reliabil-ity, durability, positive service encounters). This, of course,is eontradictory to the pervasive belief that firms thai ex-ceed their customers' expectations will enjoy an immediateincrease in customer satisfaction, but it is consistent withthe cumulative notion of satisfaction.

The arguments advanced here differ from those associ-ated with the transact ion-specific conceptualization of cus-tomer satisfaction, In a transaction-specific situation, wemight expect an increase (decrease) in a consumer's expec-tations to lead to a short-term fail (rise) in thai consumer'ssatisfaction with a specific transaction. In Ihe context of cu-mulative customer satisfaction, the Iong-ruti effects of in-creased (decreased) expectations should outweigh this short-term effect and lead to a rise (fall) in overall customer satis-faction. Overall customer satisfaction aggregates custotnerexperiences over time, and we expect the effect of any tem-porary disconfirmation of expectations to be marginal (An-derson and Sullivan 1993), Our firm-level measures of cus-tomer satisfaction also aggregate across customers, and. un-less disconfirmation is systematic and widespread, positiveand negative experiences should cancel out and their effecton overall satisfaction should be marginal. Due to this aggre-gation process, we expect overall satisfaclion lo be rellec-tive of actual past levels of perceived quality or deliveredvalue and forecasted future quality, rather than dominatedby the effects of any perceived gap between current qualityand expectations. This argument is persuasive from a com-petitive perspective as well, because expectations and per-ceived quality cannot remain out of sync for very long in amature, competitive market. If expectations are t(X> low, thefirm will nol attract customers and. consequently, new saleswill not develop. If expectations are too high, customerswill buy. become dissatisfied, and switch to competitiveproducts, and, again, the firm will have deficient sales. Atany given time, therefore, the difference between actual qual-ity and expectations al the aggregate level should be small.

Although the present aggregate-level study does notallow us to evaluate the efficacy of these arguments com-pletely—^such as comparing the relative importance of thenegative infiuence of expectations on satisfaction by a per-ceived gap between quality and expectations versus the im-portance of the ptwitive direct impact nf expectations on cus-tomer satisfaction—we nonetheless expeci to find thai thelatter effect is stronger and the impact of expectations on cu-mulative customer satisfaction is positive. At the sametime, it is worth noting that the conclusions reached previ-ously are not without support. The preceding argumentleads to the same conclusion reached by Boulding and col-leagues (1993) in an individual-level study of the effects ofexpectations on overall judgtnenLs. Their argument for how"will expectations" (expectations of what quality servicewill be like, as distinct from what quality should be like) at-fect quality perceptions is based (in an adaptation mecha-nism (Helson 1964; Oliver 1980), in a manner analogous toan assimilation effect (Anderson and Sullivan 1993). Themarket level argutnent presented here is different in that theeffect of the market's expectations on customers' overall sat-

isfaction at time t also depends on a forecast of what qual-ity will be like in t + 1, t + 2 as well as the impact of allpast quality experiences fmm t - I. t - 2 Overall cus-tomer satisfaction with a particular firm is a function of allpast, current, and future experience:

H3; The market's expectation of the quality of a supplier's of-fering should have a positive effect on overall customersatisfaction.

Customers' Expectations of the Firm's Quality AreAdaptiveThe experiences of customers in a previous period t - 1should have a positive influence on buyer's expectations ofquality in the current period t. Customers are likely to up-date expectations on the basis of both past experience andother types of nonexperiential information. This updatingprocess is consistent with the notion of adaptive expecta-tions found in both psychological and economic research. Ol-iver and Winer (19871 provide a comprehensive review ofdifferent approaches to modeling the updating of expecta-tions. Johnson, Anderson, and Fomeli (1994) compare alter-native approaches for modeling expectations and find thatexpectations are very nearly rational in character but thatthey are adjusted over time in an adaptive manner (Lucas1973; Muth 1961; Taylor 1979). That is. the market consid-ers all available infomiation conceming quality and contin-ually updates expectations in an efficient manner save for"imperfections" (e.g., uncertainty, costs) that impede theflow of information and result in a small updating effectthat gives the appearance of being adaptive.

The relative size of the adaptive updating effect is impor-tant and depends on both production and consumption fac-tors (Anderson 1994a; Anderson and Sullivan 1993). Onthe production side, greater temporal variation in qualityshould imply a greater updating effect. For example, a highrate of innovation or technological change may provideshocks that require the market to revise expectations. Qual-ity also may change because of period-to-period fiuctua-tions in materials, production, or the service delivery sys-tem (e.g., business cycles). Conversely, there should be lessof an updating or leaming effect when there is greater stabil-ity. In this case, the market's expectations (based on similarpast experiences) should mirror the level of quality experi-enced in the current period.

On the consumption side, the market's degree of uncer-tainty regarding a particular product or service encountercan infiuence the size of the updating coefficient. For exam-ple, where there is iittie familiarity or expertise among cur-rent customers, it is more likely that the updating effecl willbe large. The mix of newly acquired versus repeat custom-ers consequently can infiuence the size of the updating co-efficient, as can frequency of purchase, the stage of marketevolution, or shifting sociodetnographic factors. For someproducts, market infomiation conceming the quality of thegood or service may be costly or difficult to obtain withoutexperience (Darby and Kami 1973; Nelson 1970; Zeithaml1981). In attracting new customers, advertising itself alsocan influence the size of the u[>dating effect. Although adver-tising may not neces.sariiy distort expectations (e.g., puff-

Customer Satisfaction, Market Share, and Profitability / 57

ery), it is unlikely to provide complete information. Custom-ers may be attracted by a limited set of particular benefitsstressed in advertising, but they must experience the prcxl-uct or service to leam more fully about quality—and thenmay revise expectations accordingly. A similar argumentmight be constructed for the efficiency of word of mouth inconveying information about quality.

Uncertainty also can arise if it is difficult for customerslo predict what their consumption experiences will be likeover time. This may be the case if a product or service hasimportant experience attributes (attributes that must be expe-rienced to be evaluated) or credence attributes (those thatare very difficult to evaluate and force the customer to relyon the product's reputation to evaluate them) (Darby andKami 1973; Nelson 1970; Zeithaml 1981). If certain as-pects of quality are unobservable or difficult to anticipate, ilmay be problematic for the market to predict future quality.Expectations of quality for a particular firm would be up-dated as infomiation about actual quality becomes availa-ble. For example, automobile customers leam about durabil-ity, reliability, and quality of service over time. Personalcomputer users are likely to encounter unanticipated bene-fits and difficulties as new applications are identified andcomplementary products develop. Customers may have loadapt as the nature of an offering becomes apparent. In con-trast, if infonnation is relatively complete and easy to ob-tain, the period-to-period updating effect at tbe market levelshould be small. Similarly, there may be less updating if var-iation in production or consumption is indistinguishablefrom white noise. This might be the case if a product or ser-vice is difficult to standardize or quality is difficult for buy-ers to evaluate (Anderson 1994a; Anderson and Sullivan1993; Deighton 1984; Hoch and Ha 1984).

It can be surmised from these statements that the size ofthe updating effect depends largely on the rate at which qual-ity changes over time and the market learns. The rate ofleaming or adjustment by the market is not likely to be in-stantaneous—as it might be if (be market were perfectly ef-ficient—due to the cost of acquiring infomiation and the ef-fects of uncertainty discussed previously. Another implica-tion is that the updating effect should be small relative tothe cumulative effect of all past information. In Sweden, asin other industrialized nations, most industries are mature.In more mature markets, production-side factors are suchthat quality is relatively stabie—even though the mosthighly evolved (or complacent) competitors in these indus-tries certainly have been forced to change during the periodof the study. Customers in mature markets may havegreater experience with and knowledge of quality (Johnsonand Fomell 1991). This implies that the updating coeffi-cient, representing the relative weight given by the marketto the most recent information about quality, should besmall relative to the size of the coefficient of lagged expec-tations, that is, the relative weight given by the market to allpast information about quality.

We argue that the processes described previouslyshould lead to a similar finding at the fimi level, just asBoulding and colleagues (1993) fmd evidence fora small up-dating effect at the individual level. The competitive argu-

ments advanced in the previous section also provide a com-pelling argument for a relatively small updating coefficient.The difference between the market's expectations and ac-tual experiences with quality cannot be great for Iong peri-ods of lime or the firm will not survive.

The preceding arguments can be summarized asfollows:

H4: The marketplace has adaptive expectations concemingthe quality of a supplier's offering. The size of the adap-tive updating effect should be small.

Relative Importance of Ouality and ExpectationsIf both current quality and expectations have a positive im-pact on customer satisfaction, then which effect should weexpect to be stronger? If expectations primarily representpast quality experiences and/or nonexperiential quality infor-mation, we would expect current quality to have a greaterimpact for several reasons. First, current quality experi-ences should be more salient and take precedence over pastquality experiences in determining customer satisfaction. Ac-tual experience with a good or service should outweighother information, especially in the aggregate. In addition,perceived quality is measured in our study as perceived qual-ity relative to price and contain.s additional information thatexpectations do not contain. Finally, Oliver (1989) arguesthat transaction-specific satisfaction for ongoing consump-tion activities (durable goods, services, and repeatedly pur-chasing packaged goods) should be primarily a function ofperceived perfonnance. Expectations should be passive andhave a minimal effect on satisfaction under these conditions(Bolton and Drew 1991; Oliver 1989). In such situations,the level of and even degree of variation in quality is wellknown to customers. This same argument has even greaterforce when the focus is on cumulative customer satisfac-tion. Cumulative customer satisfaction is based on many ex-periences. Customer knowledge, particularly in relativelymature and stable markets, should be such that expectationsshould accurately mirror current quality. The contributionof expectations lo customer satisfaction should be mainly inthe form of predicting future quality. Unless there is uncer-tainty with regard to future quality, the contribution of ex-pectations to overall customer satisfaction should be mini-mal (Anderson 1994a). In the extreme, expectations pro-vide no additional infomiation.

Sweden's economy is well developed. The selected cat-egories are mature, even though these categories are compel-itive and subject to change^—as well as perceived with a lim-ited degree of uncertainty—and information flows rela-tively freely. Accordingly, just as we expect the updating ofexpectations from period lo period to be small, we arguethe following:

H5: The impact of perceived quality on overall customer sat-isfaction should be relatively greater than Ihe impacl of en-pectalions about quality.

Customer Satisfaction and Market ShareIntuitively, customer satisfaction and market share might beexpected to go band in hand. Buzzell and Wiersema(1981a, b) find relative quality and market share to be posi-

58 / Journal of Marketing, July 1994

tively related for firms in the PIMS database (though recentwork by Szymanski, Bharadwaj, and Varadarajan 11993]suggests this may be the case only for PIMS data or whenthe employed methodology does not control for "unobserv-ables"). The same type of relationship might be expectedfor customer satisfaction. For example, high customer satis-faction should help in attracting as well as retainingcustomers.

However, it is not clear that high customer satisfactionand high market share are always compatible, Fornell(1992) and Griffin and Hauser (1993) discuss the possibil-ity of a negative relationship between customer satisfactionand market share. They argue that whereas a small market-share firm may serve a niche market quite well, a large mar-ket-share firm must serve a more diverse and heterogeneousset of customers. At least two primary forces are at work indetermining whether the relationship between customer sat-isfaction and market share is positive or negative. First, in-creasing market share, at least up to a point, can produceeconomies of scale. This, for example, may allow the fimito charge lower prices, thus increasing the value of thefirm's offering and consequently increasing customer satis-faction. By contrast, there may be a dilution of effort thatgoes with trying to serve an increasing number of custom-ers and/or segments. This dilution could lead to low-qualityservice and is likely to occur in industries in which cus-tomer preferences are heterogeneous and/or personal ser-vice is important. In undifferentiated industries with homo-geneous customer preferences, it is more likely ihat cus-tomer satisfaction and market share are positively related, es-pecially in the long run.

It is instructive lo examine these arguments for thecases of firms pursuing different "generic" strategies—differentiation, niche, and low-cost leadership—as origi-nally categorized by Porter (1980). Firms lollowing pureniche strategies are likely to be more successful at satisfy-ing customers than those pursuing other strategies. Al-though it is true that firms can differentiate their offeringsto meet the needs of multiple segments, it may become dif-ficult or costly to do so without diluting the quality of whatis provided (e.g.. personal service). As a linn grows bybringing in customers with preferences further away fromthe firm's target market, the overall level of customer satis-faction is likely to fall.

It is worth noting that this situation is complex becauseof the dual impact of quality and price on satisfaction. Forexample, in a market in which there is a relatively largeprice-sensitive segment with homogeneous needs, a low-cost leader may provide a level of value that creates a rela-tively high level of customer satisfaction. There is clearly aneed for understanding the trade-offs in such situations(e.g., price elasticity versus quality elasticity of relums), ifthere are conditions under which customer satisfaction andmarket share are negatively related. If lowering price can at-tract customers that become less satisfied while increasingthe satisfaction of the current customer base, then what arethe marginal effects of the additional customers on overallsatisfaction and profitability?

In summary, the relationship between customer satisfac-tion and market .share is an emerging issue in need ofgreater understanding. Achieving success in one may lowerperformance in the other. Market share can be gained by at-tracting customers with preferences more distant from thetarget market. Service capabilities also can be overextendedas volume grows. Market share effects on profitability areequally problematic (see Szymanski, Bharadwaj. and Vara-darajan 1993 for a review of the market share-profitabilityrelationship). Clearly, there can be situations in which in-creasing one and/or the other is not profitable for the fimi.For example, an extreme approach for maximizing cus-tomer satisfaction would be to eliminate all but one cus-tomer and direct all resources to that individual. Obviously,it would be a rare set of circumstances under which itwould be profitable to do so. Conversely, a high marlcetshare or "one size fits all" strategy is likely to be profitableonly if enough customers have similar preferences. It is alsopossible that differentiation may fail to provide higher sat-isfaction due to the difficulty of serving multiple customerswithin each segment and the dilution of effort that comesfrom serving multiple segments. A firm that manages bothto provide high customer satisfaction by customizing its of-fering to each customer and maintain a large market sharewould have to enjoy very high economies of scope andscale. Another way to think about this issue is to considerwhat the small niche firm has to do to be successful. Provid-ing superior customer satisfaction is critical for its survival.

Data and MethodologyDescription of the DataAnnual indices of firm-level expectations, quality, and cus-tomer satisfaction are made available by the SCSB. Initi-ated in I9K9. the SCSB is an ongoing project managed bythe National Quality Research Center (NQRC) at the Univer-sity of Michigan Business School and the International Cen-ter for Studies of Quality and Productivity (ICQP) at theSt(Kkholm School of Economics. The 77 firms included inour NQRC study are all major competitors in a wide varietyof industries: airlines, automobiles, banking (consumer andbusiness), charter travel, clothing retail, department stores,fumiture stores, gas stations, insurance (life. auto, and busi-ness), mainframe computers (business). PCs (business),newspapers, shipping (business), and supermarkets. Thecompanies surveyed in each industry are the largest sharefinns such that cumulative share is approximately 70%. Sev-eral state-owned monopolies are also measured by theSCSB but are not included in this study.

The measurements in the SCSB begin with a computer-aided telephone survey designed to obtain a representativesample of customers for each firm. Potential respondentsare selected on the basis of recently having purchased andused a company's product. To participate, each respondentis required to pa.ss a battery of screening questions. The ques-tionnaire employs 10-ptiint scales to collect multiple meas-ures for each construct. For example, for the quality con-struct, resptmdents are asked to evaluate quality given priceand price given quality in two separate questions. This pro-

Customer Satisfaction, Market Share, and Profitability / 59

cess results in approximately 25,(X)0 observations per vari-able (for each year) from which indices are constructed. For-nell (1992) describes a latent variable approach to estimat-ing the indices.

The SCSB measures are combined with economic re-turns data for the publicly held firms. Specifically. ROI foreach firm (that is. return on assets located In Sweden) isused as a measure of economic retums. Unusual or extraor-dinary retums are treated as outliers. To make the ftillest pos-sible use of Ihe available data, missing values are treated ashaving the same correlation as the values present in the dataset. In other words, the distributions of the variables aretreated as censored and a covariance matrix is created as abasis for estimation.

Clearly, there are difficulties in combining the data sets.For exampie. the ROI data are for a business as a wholerather than a specific product measured by the SCSB. Al-though this is not a serious issue for the retail and servit:esectors, it is a concem for firms with more diverse productlines, such as the automobiles. Although ROI is commonlyused in studies of the impact of strategic variables, it is notan ideal measure of economic retums. Capital market data(stock prices) would have been another interesting measureif a large portion of the SCSB firms were actively traded inSweden or transacted most of their business there.

Testing the Hypotheses

The system of equations to be estimated is presented inTable 2. In keeping with the arguments advanced in the pre-vious section, expectations are influenced by past quality,customer satisfaction is influenced by both quality and ex-pectations, and economic retums are influenced by satisfac-tion. Obviously, there are other variables besides customersatisfaction that affect economic returns. The effects ofthese variables are captured in the lag structure, the errorterm, and a trend term. If the marketplace has adaptive ex-pectations, then we should expect the coefficient for the im-pact of past quality QUAL^_, on expectations EXP, to bepositive 1 > Pi2 > 0. (To test the adaptive expectationsmodel, we restrict the coefficients such that p , ; = I - pi,.)For customer satisfaction SAT, we expect the impact ofboth curTent quality QUAL, and EXP, to be positive. P22 -*0 and (B23 > 0. The effect of current quality on customer satis-faction should be greater than that of expectations, P22 >^23- The impact of SAT, on profitability as measured by re-tum on assets ROI, is expected to be positive. P̂ 2 ^ ^- Th''*latter relationship is predictive in that the survey measuringcustomer satisfaction is conducted in the first half of the fis-cal year and economic retums are based on year-end finan-cial reporting. As logarithms are taken of each variable, theestimated coefficients are interpretable as elasticities.

Specification

To account for heterogeneity in the cross-section of indus-tries (e.g., differences in accounting practices, industrial or-ganization considerations) and possible unobservable ef-fects (e.g.. firm strategy, pioneering advantage), the systemis formulated as state dependent (Aniemiya 1985; Boulding1990; Jacobsen 1990a, b; Maddala 1977). This formulation

TABLE 2System of Equations Underlying

the Conceptual Framework

Lagged Dependent Specification

EXP, = ai -̂ p,tEXP,_, +

SAT, = ag +

ROI, = 03 +

,_, + P22QUAL, + P23EXP, +

,.! + 1332SAT, + pggTREND +

First Differences Specification

+ P13TREND + e,,

ASAT, = P22AQUAL, + P23AEXP, + 324TREND +

AROI, = 332ASAT, + 333trend + E3,

refiects the expected persistence of the benefits of customersatisfaction for the firm (consistent with the overall or cumu-lative nature of satisfaction focused on in this study). Thisspecification is also consistent with the argument that themarketplace has adaptive expectations. Finally, it fits withthe intuitive notion of Ricardian Rents resulting from highcustomer satisfaction (Montgomery and WemerfeU 1988).Accordingly, the endogenous variable in each equation is re-gressed on its lagged quality and a set of independent varia-bles capturing the appropriate effects.^ In view of the exis-tence of simultaneity and expected correlation between theerrors of the equations, three-stage least squares is used toestimate the model.

It is worth noting that a common—and conservative—correction for controlling for heterogeneity and unobserva-bles in short cross-sectional time-series data is to transformthe data through first differences (Maddala 1977). (This spec-ification restricts pn = P21 = P^i = - ' • ) I' should bepointed out that this specification is more consistent with atransaction-specific conceptualization of customer satisfac-tion. It implies that short-temi changes in quality and expec-tations have immediate rather than long-term consequencesfor customer satisfaction and ultimately profitability. Wetherefore expect expectations to have a negative effect oncustomer satisfaction in this specification.

ResultsTable 3 presents three-stage lea.st squares estimates for thetwo specifications. The findings generally confirm the pat-tem of effects as hypothesized. Let us first discuss the find-ings relating quality and expectations to satisfaction andthen tum our attention to the effect on economic retums.With regard to the first equation of each specification, the co-efficients support the idea of adaptive expectations. The rel-ative size of the coefficients for the impact of past expecta-tions EXP,_| and past quality QUAL,_| on current expecta-tions EXP, is consistent with how one wouid expect afirm's reputation for quality to change over time. Although

tongcr time-series, poienlial methods of controlling for unob-servabtes are Ihc famity of error-component modets (Arnemiya 1985) andlalenls:lass pooling meihods (Ramasway, Anderson, and DeSarbo 1994).

60 / Joumal of Marketing, July 1994

TABLE 3Empirical Findings

All coefficients are three-stage least squares estimates.

Lagged Dependent Specification—Weighted R-square is .82

EXP, = .or + .91' EXP,_i + .09" QUAL,_, - .003* TREND

SAT, = -.12 + .44- SAT, , + .49* QUAL, + .10'EXP, - .003* TREND

ROI, = -1.10' + .75* ROI,_i + .40* SAT, + .002 TREND

First Differences Specification—Weighted R-square is .35

AEXP, = . i r AQUAL,., - .011" TREND

ASAT, = IT AQUAL, - .50* AEXP, + .000 TREND

AROl, = .76* ASAT, - .001 TREND

•indicales the coefficient is stgnificanl at ihe .01 level.

expectations are fundamentally stable, changes in the levelof quality prt)vided by a finii will enhance or erode the com-pany's reputation for quality over time. The estimates pre-sented in Tabie 3 suggest that though the market eventuallywill revise its expectations completely (the long-run elastic-ity of expectations with respect to changes in past quality isrestricted to be one), this will be a slow process for the typ-ical firm. Conversely, ihere appears to be considerable mo-mentum for the current levei of expectations. The stabilityof expectations suggests that a finn's reputation for provid-ing quality will noi change quickly.

As seen in the second equation of Table 3, both qualityand expectations have a positive impact on customer satis-faction.^ For the state-dependent or persistence formulation,these effects are not only in the predicted direction, but alsoof the predicted relative si/.e. In fact, the estimates suggestthat current customer satisfaction is primarily a function of(1) current quality and (2) pasl satisfaction. Quality has thegreatest Impaci on customer satisfaction, according to bothspecificatjons. The importance of current quality in determin-ing customer satisfaction is consistent with Ihe notion thatcurrent experience will be weighted more highly than pastor anticipated experience.

In the first specification, the size of the effect of laggedcustomer satisfaction indicates a strong carryover effect.For every percentage point change in customer salisfactionat t -1 , customer satisfaction al t changed by .44%. This sug-gests that high past salisfaction of current customers pro-vides a strong Indication that current and consequently fu-ture customer satisfaclion will be high. Interestingly, the es-timate of a carryover effect of .44 is very nearly identical tothe average carryover effect for sales, .468. as estimated inthe meta-analysis of Assmus. Farley, and Lehmann (1984).

•'il is worth noiing rhal the methodology here produces similar substan-tive results to other ineihtxls of amlrolling Tor fixed effects (e.g., instrumen-tal variables). The exception is ihe si7.e of the awrficienl for the effect olcustomer satistaction on ROI. Tliis coefficient is significantly larger wheninstrumental variable methods are used (Anderson, Fomell. and Lehmann1993).

A sizeable carryover effect supports the notion that cus-tomer satisfaction is indeed cumulative. The implication isthat high customer satisfaction insulates the ftmi from short-term changes in quality. The strong carryover effect of pastcustomer satisfaction aiso means that it is time-consumingfor firms wilh iow customer satisfaction to improve theirstanding in the market.

The eifect of expectations of quality on customer satis-faction is positive and significant, as well as relativelysmall. For every percentage point change in expectations,customer satisfaction changes by .10%. This is supportiveof the argument for adaptive expectations. Expectationsadapt slowly and provide incremental information to thatprovided hy C|uaiity. In particular, in modeling customer sat-isfaction as a long-term, dynamic phenomenon, the carry-over effect of past satisfaction naturally captures infornia-lion about past experience with quality, leaving expecta-tions with a relatively marginal effect that can be inter-preted as the effect of the market's forecast of future qual-ity on current satisfaction.

It is important to note that the sign ofthe impact of ex-pectations on customer satisfaction is reversed in the first-differences specification (i.e.. negative), which implies tliata short-term increase in expectations actually may lead to adecrease in customer satisfaction. That is, increasing cus-tomer expectations by overpromising is likely to be detri-mental to the firm in the short run. whereas increasing cus-tomer expectations through improving quality benefits thefirm in the long run.

Return on investment, a long-term measure of eco-nomic health, is strongly affected by customer satisfaction.This is tnje for both specifications. However, the inteq^reta-lion of the two specifications is different. The lagged-dependent variable specification implies that a change in cus-tomer satisfaction is not reflected all al once in returns.Rather, a percentage pt>int change in customer satisfactionin one period carries over to future periods, consistent withthe cumulative nature of customer satisfaction. The first-differences specification, on Ihe other hiincl. implies thaithere is a larger inimediale effect from a change in cus-tomer satisfaction, but that this advantage is short-lived andunsustainable. Nevertheless, both findings suggest Ihat pro-viding high quality and high customer satisfaction is re-warded by economic returns. Moreover, the log-linear formu-lation implies Ihat if the costs of providing high quality andcustomer satisfaction are increasing at an iticreasing rale,then there must be an optimal level of satisfaction. Obvi-ously, then, strategies thai seek lo maxiniize customer satis-faction are inappropriate.

How do these figures compare with other studies exam-ining ihe impact of marketing mix variables on ROI?Buizell and Gale (1987) report an impact coefficient for rel-ative quality on ROI of ,11. We can transform this valueinto an average elasticity of ROI wilh respect to quality byusing their mean values of ROI and quality. This calcula-tion yields an average short-run elasticity for ROI with re-spect to quality of .25. The coefficients in Table 3 can beused to compare our fmdings with this figure. To obtain anestimate of ihe shon-r\in impact of a change in quality on

Customer Satisfaction, Market Share, and Profitability / 61

FIGURE 1Returns due to Increased Satisfaction

One Point Increase In Each Year

dat

i

3

QSo

fDol

l

E:

Year

ROI, we calculate the elasticities in the chain from qualityto ROI. Here, the short-run elasticity of ROI wilh respect toquality is .49(.4O) = .196. Hence, we fmd an elasticity ofROI with respect to quality comparable to. though slightlysmaller than, that found in the PIMS databa.se.

Empirlcai Prediction of the Vaiue of a One-Pointincrease in Customer Satisfaction

What is the value of an increase in the cuslomer satisfactionindex for the typical Swedish firm represented in theSCSB? To illustrate this, let us consider the case of a firtnwith a five-year planning horizon. Suppose the fimi must es-timate the increase in ROI resulting from increasing its cus-tomer satisfaction index by a single point in each of thenext five years (cumulative increase of five points). Assum-ing the firm's ROI in the initial year is the same as the av-erage for the sample (10.83%), the estimates in Table 3imply incremental increases in ROI for the next five yearsof .07%. .18%. .33%. .51%. and .1\%, respectively, overwhat the firm's ROI would have been without increasingcustomer satisfaction. The fifth-year ROI of 11.54% repre-sents a 6.59% increase over the original ROI of 10.83%.The five-year cumulative increa.se of 1.8% (1.8 = .07 + .18+ .33 + .51 + .71) represents cumulative incremental retumsof 16.66% relative lo current ROI (16.66 = 100 11.8/I0.83|). The net present value for the incremental returnscan be calculated by a.ssuming that our "typical" finn hasan as.set base corresponding to the sample mean ($6(M} mil-lion), a policy of paying out all retums as dividends, and ap-plies a discount rate of 10,00%. As illustrated in Figure 1,the results of this calculation indicate incremental returnsover the next five years of $.357 million, $.888 million.$1,487 million. $2.09 million, and $2,66 million, respec-tively. This represents cumulative discounted returns of$7.48 million, or 11.5% of current ROI.

Although the preceding calculations may seem some-what modest in absolute size. It shouid be kept in mind thatthe prediction is based on a cross-sectional analysis and thatthe scale of a typical Swedish firm is much smaller thanthat found in an economy such as the United Stales'. For ex-

ample, if the same coefficients apply to a sample of U.S.firms (e.g., the Business Week 1000, with average assets of$7.5 billion and average ROI of 11%), the cumulative incre-mental retums from a continuous one-point increase in cus-tomer satisfaction over a five-year span would be $94 mil-lion, or 11.4% of current ROI.

The Value of Current Customer AssetsThe preceding empirical prediclion of the value of customersatisfaction can be supplemented by an analytical model. Ifimproving customer satisfaction increases the likelihood ofrepurchase, then we can illustrate the economic benefits ofsuch a change by con.sidering current customers as an assetto the firm and calculating their net present value to thefirm. A straightforward calculation might capture customerassets as a function of the likelihood or probability that a sat-isfied customer will remain loyal. PR(Loyal|SaUsfaction),the average gross margin per period G. the length of the av-erage repurchase cycle X. and a discount factor d. The asso-ciated net present value equation can be written:

NPV = 2 XG(Pr{Loyal|Satisfactionl/(l + 3))"*',1=1

We assume that there is a monotonic relationship be-tween customer satisfaction and repurchase intentions thatis linear for small changes in satisfaction. Anderson and Sul-livan (1993) estimate that a .0058 increase in repurchase like-lihoixl (on a scale from 0 lo I) will result from a one-pointincrease in customer satisfaction. Hence, if a firm's satisfac-tion index is on average 67 and undergoes an increase to70. the typical firm's repurchase probabilities wouldchange from the average of .75 to .7674. Given the averagegross margin for the firms in the SCSB ($65 million) and as-suming customers purchase an average of once per year, thenet present value of customer assets would rise $6.4 mil-lion, or 5.4%, from $118.8 million to $125.2 million.

Customer Satisfaction and Market Share

How are customer satisfaction and market share related?We have been able to obtain 1989-90 company-level mar-ket share data to match the customer satisfaction indices fora subsample of the SCSB firms. Plots of the raw data andyear-to-year changes in market share and customer satisfac-tion are shown in Figure 2. Both plots suggest downwardsloping, that is, inverse relationships between customer sat-isfaction and market share. The plot of raw satisfaction ver-sus raw market share shows that no firm has hoth high cus-tomer satisfaction and high market share. Moreover, year-to-year increases (decreases) in market share are likely to be as-sociated with decreases (increases) in customer satisfaction.The pearson correlation between raw market share and sat-isfaction is -.25 (/j-value of .03 with n = 77) and the corre-lation between year-to-year changes in the variables is -.37(p-value of .05). Regressing changes in the customer satis-faction index on changes in market share yields a coefft-cient of -.88 (jj-va]uc of .05).

62 / Journal of Marketing, July 1994

FIGURE 2Market Share and Satisfaction

1989 and 1990

10 20 30Market Share

40 50

1 0 -

f 5-

I 0 —

Changes From 1989 to 1990

A

S-1

AA

-3 -2

A

A

AA

A AA

A A

I < I I I I

- 1 0 1 2Change in Market Share

Figure 2 provides a preliminary indication, similar toGriffin and Hauser (1993), that increasing market share ac-tually may decrease customer satisfaction. This may indi-cate that a more differentiated strategy can lead to decreasesin market share. In addition, it may indicate that, at least inshort-run cross-sectional analyses, customer satisfactionand market share are not always compatible goals.

Summary and ConclusionsThe widespread belief in the intuitive relationship betweenquality, customer satisfaction, and economic returns, aswell as the growing frustration with attempts to improvequality, serve to underscore the importance of analyticaland empiricai work increasing our understanding of cus-tomer satisfaction and how it relates to economic returns.The frustration of many firms engaged in attempts to im-prove quality may be due to any number of factors, frompoor market data to the intransigence of functional silos orfixation with short-term results that may leave firms unableto wait for the benefits of investing in quality and customersatisfaction to materialize (Ettlie and Johnson 1994). Al-though we do not provide guidance for managers seeking ei-ther tools for improving quality (e.g., TQM) or guidelinesfor implementing quality programs, it does provide motiva-tion for continuing their efforts and overcoming any imped-iments encountered: Firms that actually achieve high cus-tomer satisfaction also enjoy superior economic returns. Anannual one-point increase in customer satisfaction has a netpresent value of $7.48 million over five years for a typicalfirm in Sweden. Given the sample's average net income of$65 million, this represents a cumulative increase of 11.5%.If the impact of customer satisfaction on profitability is sim-ilar for firms in the Business Week 1000, then an annualone-point increase in the average firm's satisfaction indexwould be worth $94 million or 11.4% of current ROI.Firms considering implementing or, in an increasing num-ber of cases, curtailing quality programs should consider

the benefits indicated by these findings in reaching theirdecisions.

Our findings also indicate that economic returns from im-proving customer satisfaction are not immediately realized.Because efforts to increase current customers' satisfactionprimarily affect future purchasing behavior, the greater por-tion of any economic returns from improving customer sat-isfaction aJso will be realized in subsequent periods. This im-plies that a long-run perspective is necessary for evaluatingthe efficacy of efforts to improve quality and customersatisfaction.

The long-run nature of the economic returns from im-proving customer satisfaction also has broad strategic impli-cations. If increasing customer satisfaction primarily affectsfuture cash flows, then resources allocated to improvingquality and customer satisfaction should be treated as invest-ments rather than expenses. Loyal and satisfied customersare a revenue-generating asset to the firm that is not withoutcost to acquire, retain, and develop. This is very differentfrom viewing sales as a set of more or less disjoint and mu-tually exclusive transactions. Implementing a customer-asset orientation means aligning the firm's processes, re-sources, performance measures, and organizational struc-ture for treating customers as an asset. Our findings providea rationale for firms to move in this direction. Once the po-tential ofa customer-asset orientation is acknowledged,there are two key procedural questions for management: (1)How do we measure the value of this asset? and (2) Howdo we increase its value? Answers to both these questionsare now being developed (e.g., Fomell 1991a, b; 1994).

Our findings also provide a preliminary indication oftrade-offs between customer satisfaction and market sharegoals. We find that customer satisfaction actually may fallas market share increases. For example, whereas a small mar-ket-share firm may serve a niche market quite well, a largemarket-share firm often must serve a more diverse and het-erogeneous set of customers. Gains in market share maycome from attracting customers with preferences more dis-

Customer Satisfaction, Marlcet Share, and Profitability / 63

tant from the target market. The firm may overextend its ser-vice capabilities as the number of customers and/or seg-ments grows. In such a situation, even though the overalllevel of customer satisfaction is falling, a firm's sales andprofits may be increasing. It is worth noting that this maybe a short run versus long run phenomenon. In the long run.it is possible that customer satisfaction and market share gotogether, but there is growing evidence that this is not al-ways the case in the short run or a cross-sectional analysis.

When quality and expectations increase, there is a posi-tive effect on customer satisfaction in the long run, but in-creased expectations may have a negative impact in theshort run. The large, positive impact of quality on customersatisfaction is intuitive. Expectations have a positive effecton customer satisfaction In the long run because they cap-ture the accumulated memory of the market concerning aUpast quality information and experience, as well as the mar-ket's forecast of the firm's ability to deliver quality in the fu-ture. This forward-looking component of expectations is im-portant because this, in part, is how a firm's reputation forproviding high or low quality influences the overall satisfac-

tion of Its customers. In the context of cumulative customersatisfaction, the long-run effects of increased (decreased) ex-pectations should outweigh the short-term effect of any tem-porary gaps and lead to a rise (fall) in overall customer sat-isfaction. This firm-level finding is consistent with individ-ual-level research showing that disconfirmation of expecta-tions has a weaker effect on cumulative customer satisfac-tion than the direct impact of perceived quality (Andersotiand Sullivan 1993).

Finally, our findings indicate that, in the aggregate, cus-tomers have adaptive but largely rational expectations.Changes in the level of quality provided by a firm enhanceor erode a firm's reputation for quality over time. This is animportant process to manage for the typical firm becausesubsequent changes in its reputation for providing qualitymay not be immediate. The implication for a finn trying tomake a quality "turnaround" or "comeback" is, therefore,not to expecl immediate retums but coordinate product/service improvements with efforts to accelerate the diffu-sion of information regarding such improvements throughthe marketplace.

REFERENCESAaker, David A. (1992). "The Value of Brand Equity," Journal

of Business Strategy. 13 (July/Augu.st), 27-32.Amemiya, Takeshi (1985). Advanced Econometrics. Cambridge.

MA: Harvard University Press.American Quality Foundation (1992). Best Practices Report. Cleve-

land: Ernst & Young,Anderson, Erin and Barton Weitz (1989), "Determinants of Con-

tinuity in Coiivenlinnal Industrial Channel Dyads." Marlceiing5ci>nctf. 8(Fall). 3HK23.

Anderson. Eugene W. (1994a). "Cross-Caiegory Variation in Cus-tomer Satisfaction and Repurchase Likelihood," Marketing Let-ters., forthcoming.

(1994b), "Post-Purchase Evaluations and Word of Mouth,"working paper. National Quality Research Center, Universityof Michigan Business School.

. . Claes Fomell. and Donald R. Lehmann (1993), "EconomicConsequences of Providing Quality and Customer Satisfac-tion," Working Paper #93-112. Cambridge, MA: Marketing Sci-ence Institute.

and Mar>' Sullivan (1993), "The Antecedents and Conse-quences of Customer Satisf^tion For Firms." Marketing Sci-ence, 12 (Spring), 125-43.

Anterasian. Cathy and Lynn W. Phillips (1989). "Discontinuitie.s.Value Delivery, and the Share-Ret." Working Paper #92-121.Cambridge. MA: Marketing Science Institute.

Assmus, Gert. John U. Farley, and Donald R. Lehmann (1984)."How Advertising Affects Sales: Meta-Analysis of Economet-ric Results," Journal of Marketing Research. 2! (February),65-74.

Bateson, John E. (1989), Managing Services Marketing. London:Dryden Press.

Bearden, William O. and Jesse E. Teel (1983), "'Selected Determi-nants of Consumer Satisfaction and Complaint Reptjrts." Jour-nal of Marketing Research, 20 (February). 21-28.

Becker, Gaiy S. (1965), "A Theory of the Allocation of Time," Ec-onomic Journal. 75 (September). 493-517.

Bolton, Ruth N. and James H. Drew (1991), "A Multistage Modelof Customers' Assessments of Service Quality and Value,"Journal of Consumer Research. 17 (Marcb), 375-84.

Boulding, William (1990), "Commeniaiy On 'Unobservable Ef-

fects and Business Performance: Do Fixed Effects Really Mat-ter?"' Marketing Science. 9 (Winter), 88-91.

, Richard Staelin, Ajay Kaira, and Valerie ZeithamI (1993)."A Dynamic Process Model of Service Quality: From Expec-tations lo Behavioral Intentions," Journal of Marketing Re-search, 30 (February), 7-27.

Buzzell, Robert D. (1990). "Commentary On 'Unobservable Ef-fects and Business Performance,"' Marketing Science, 9 (Win-ter), 86-87.

and Bradley T Gale (1987). The PIMS Principles. NewYork: The Free Pre.ss.

and Frederick Wiersema (i98la), "Modelling Changes InMarket Share: A Cross-Sectional Analysis." Strategic Manage-ment Journal. 2 (January-March), 27-42.

and Frederick Wiersema (1981b), "Successful Share-Build-ing Strategies," Harvard Business Review, 59 (January-Febru-ary), 135^t4.

Cadotte, Ernest R., Robert B. Woodruff, and Roger L. Jenkins(1987), "Expectations and Norms io Models of Consumer Sat-isfaction," Journal of Marketing Research, 24 (Augu.st), 305-14.

Capon, Noel. John U. Farley, and Scott Hoenig (1990), "Determi-nants of Financial Performance: A Meta-Analysis," Manage-ment Science. 36 (October). 1143-59,

Churchill, Gilbert A. and Carol Suprenant (1982), "An Investiga-tion Into The Determinants Of Customer Satisfaction," Jour-nal of Marketing Research. 19 (November), 491-504.

Cronin, J. Joseph, Jr. and Steven A. Taylor. "Measuring ServiceQuality: A Reexaminatiun and Extension." Journal of Market-ing. 56 (July). 55-68.

Crosby, P.B. (1979), Quality is Free. New York: McGraw-Hill.Czepiel. John A. and Robert Gilmore (1987), "Exploring die Con-

cept of Loyally in Services," in The Services Challenge: Inte-grating for Competitive Advantage. Jobn A. Czepiel, Carole A.Congram, and James Shanahan. eds. Chicago: American Mar-keting Association, 91-94.

Darby. Michael R, and Edi Karni (1973), "Free Competition andibe Optimal Amount of Fraud." Journal of Law and Econom-ics, 16 (April), 67-88.

Deighton, John (1984), "The Interaction of Advertising and Evi-

64 / Journal of Marketing, July 1994

dence," Journal of Consumer Research, II (December). 763-70.

Dodds, William B., Kent B. Monroe, and Dhruv Grewal (1991),"Efrecis of Price, Brand, and Store Information on Buyers'Product Evaluations.'" Journal of Marketing Research, 28 (No-vember), 307-19.

The Economist (1992). "The Cracks in Quality" (April 18), 67-68.

Ettlie, John and Michael D. Johnson (1994), "Impediments to Suc-cessful Implementation of Quality Programs." Marketing Ut-ters, forthcoming.

Fomell, Claes (1991a). "National and Corporate Customer Satis-faction Indices," paper presented at the World Quality Day,World Trade Center, Amsterdam. Tlie Netherlands.

{1991b), "TheChallengesof a Customer Orientation," inM-novation. Integration A Strategic Processes: A Managerial Ap-proach. S. Barlinn. K, Haanes, P.S. Bronn, and P. Lorange,cds. Oslo: Bedriftsokonomens Forlag. 33-44.

(1992), "A National Customer Satisfaction Barometer; The

search Center. University of Michigan Business School._and Claes Fomell (1991). "A Framework For Comparing Cus-

Swedish Experience." Journal of Marketing, 55 (January). 1-21.

(1994), "Productivity. Quality and Customer Satisfaction asStrategic Success Indicators ai Firm and National Levels." inIntegra! Strategy, H.B. ThoreiU. eil. forthcoming.

and Birger Wemerfelt (1987). "Defensive Marketing Strat-egy by Customer Complaint Management: A Theoretical Anal-ysis," Journal of Marketing Research. 24 (November), 337-46.

and (1988). "Model For Customer Complaint Manage-ment." Marketing Science. 1 (Summer). 271-86.

Garvin. David A. (1988), Managing Quality: The Strategic andCompetitive Edge. New York: The Free Press.

Griffin, Abbie and John R. Hauser (1993). ' "The Voice of the Cus-tomer." Marketing Science. 12 (Winter). 1-27,

Gronroos. Christian (1990), Service Management and Marketing:Managing the Moments of Truth in Service Competition. Lex-ington, MA: Lexington Books.

Hauser. John R. and Steven M. Shugan (1983). "Defensive Mar-keting Strategies." Marketing Science, 2 (Fall). 319-60.

Duncan I. Simester. and Birger Wernerlelt (1993), "Cus-tomer-Saiisfaction Based Incentive Systems," working paper.International Center for Research on the Management of Tech-nology. Sloan School of Management. MassachussetLs Instituteof Ifechnology.

Helsen. Harry (1964), Adaptation-Level Theory. New York:Harper & Row.

Hoch, Stephen J. and Young-Won Ha (1984). "Consumer Learn-ing; Advertising and the Ambiguity of Product Experience."Journal of Consumer Research, 13 (September), 221-33.

Holbrook. Morris B. (1994). "The Nature of Customer Value: AnAxiology of Services in the Consumption Experience." in Ser-vice Quality: New Directions in Theory and Practice, RolandT Rust and Richard L. Oliver, eds. Thousand Oaks. CA: SagePublications. Inc.. 21-71.

Howard, John A. and Jagdesh N. Sheth (1969), The Theory ofBuyer Behavior. New York: John Wiley & Sons, Inc.

Hunt, H. Keith (1977). "CS/D—Overview and Future Research Di-rections." in Conceptualization and Measurement of CustomerSatisfaction and Dissatisfaction, H. Keith Hunt, ed. Cambr-idge. MA: Marketing Science Institute, 455-88.

Jacobson. Robert D. (1990a). "Unobservable Effects and BusinessPerformance," Marketing Science, 9 (Winter), 74-85.

(1990b). "Unohservable Effect.s and Business Performance,Reply to Comments of Boulding and Buzzell," Marketing Sci-ence, 9 (Winter). 92-95.

Johnson. Michael D., Eugene W. Anderson, and Clae.s Forncll(1994), "Rational and Adaptive Expectations in a Customer Sat-isfaction Framework," working paper. National Quality Re-

tomer Satisfaction Across Individuals and Product Catego-ries." Journal of Economic Psychology, 12 (2), 267-86.

Juran. J.M. (1988). Jurans Quality Control Handbook, 4th ed.New York: McGraw Hill.

Katona, George (1979), "Toward a Macmpsychology." AmericanPsychologist, 34, 118-26.

Keller. Kevin L. (1993), "Conceptualixing. Measuring, and Man-aging Brand Equity," Journal of Marketing, 57 (January), 1-22.

Kotler. Philip and Sidney J. Levy (1969). "Broadening the Con-cept of Marketing," Journal of Marketing, 33 (February), 10-15.

LaBarbera. Priscilla A. and David Mazursky (1983), "A Longitu-dinal Assessment of Consumer Satisfaction/Dissatisfaction:The Dynamic Aspect of the Cognitive Process." Journal ofMarketing Research, 20 (November), 393-404.

Lancaster, Kelvin (1979). Variety. Equity, and Efficiency. NewYork: Columbia University ft^ss.

Lovelock, Christopher H. (1984). Services Marketing. EnglcwoodCliffs, NJ: Prentice-Hall. Inc.

Lucas. Roben E.. Jr. (1973), "Some International Evidence of Out-put-Inflation Tradeoffs," American Economic Review, 63(July). 326-34.

Maddala. George S. (1977), Econometrics. New York: Mc-Graw-Hill.

Montgomery. Cynthia A. and Birger Wemerfelt (1988), "Diversi-fication. Ricardian Rents, and Tobin's q." RAND Journal of Ec-onomics, 19 (Winter). 623-32.

Montgomery, David B, (1975). "New Product Distribution: AnAnalysis of Su|>emiarket Buyer Decisions." Journal of Market-ing Research, 12 (July), 255-64.

Morrison. Donald G. (1979), "Purchase Intentions and PurchaseBehavior," Journal of Marketing. 43 (Spring). 65-74.

Muth. John F. (1961), "Rational Expectations and the Theory ofPrice Movements," Econometrica, 29 (July). 315—35.

Nelson. Phillip (1970). "Information and Consumer Behavior,"Journal of Political Economy, 82 (July), 311-29.

Newsweek (1992), "The Cost of Quality" (September 7), 48.Oliver. Richard L. (1977). "Effects of Expectations and Disconfir-

mation on Postexposure Prtxluct Evaluations," Journal of Ap-plied P.sychology, 62 (April). 246-50.

(1980), "A Cognitive Mtxlcl of ihe Antecedcnt.s and Conse-quences of Satisfaction Decisions," Journal of Marketing Re-search, 17 (November). 460-69.

(1989). "Priicessing of the Satisfaction Response in Consump-tion: A Suggested Framework and Research Propositions,"Journal of Consumer Satisfaction, Dissatisfaction, and Com-plaining Behavior, 2, 1-16.

(1993). "A Conceptual Model of Service Quality and Ser-vice Satisfaction.'' in Advances in Services Marketing and Man-agement, Ifcresa A. Swanz. David E. Bowen. and Stephen W.Brown, eds- Greenwich, CT: JAI Press, 65-86.

and Wayne DeSarbo (1988), "Response Determinants InSatisfaction Judgments," Journal of Consumer Research, 14(March). 495-507.

and John E. Swan(I989). "Consumer Perceptions of Interper-sonal Equity and Satisfaction in Transactions: A Field SurveyApproach." Journal of Marketing, 53 (April), 21-35.

and Russel S. Winer (1987). "A Framework for the Forma-tion and Structure of Castomer Expectations: Review and Prop-ositions," Journal of Economic Psychology, 8. 469-99.

Para.suraman. A.. Valerie A. Zeithaml. and Leonard L. Berry(1985). "SERVQUAL: A Multiple-Item Scale for MeasuringConsumer Perceptions of Service." Journal of Retailing, 64(Spring). 12-^0.

Phillips. Lynn W.. Dae Chang, and Robert D. Buzzell (1983).

Customer Satisfaction, Market Share, and Profitability / 65

"Product Quality. Cost Position, and Business Performance: ATest of Some Key Hypotbeses." Journal of Marketing, 47(Spring), 26-43.

Porter, Michael E. (1980), Competitive Strategy. New York:MacMillan/Free Pres.<!.

Ramaswamy. Venkatram. Eugene W. Anderson, and Wayne S. De-Sarbo (1994). "A Disaggregate Negative Binomial RegressionProcedure for Count Data Analysis." Management Science,forthcoming.

Reicbheld, Fredrick F. and W, Earl Sasser (1990). "Zero Defec-tion.s; Quality Comes to Services." Harvard Business Review.68 (September/October). 105-11.

Ricardo, David (1817). Principles of Political Economy, 3rd ed.London: John Murray.

Robertson. Thomas S. and Hubert Gatignon (1986), "CompetitiveEffects on Tfechnology Diffusion." Journal of Marketing, 50(July). 233-5S.

Rosen, Sherwin (1974). "Hedonie Prices and Implicit Markets:Product Differentiation in Pure Competition." Journal of Polit-ical Economy. 82 (January/February). 34—55.

Rust. Roland and Anthony Zahorik (1993). "Customer Satisfac-tion, Customer Retention, and Market Share." Journal of Retail-ing. 69 (Summer). 145-56.

Sawyer, Alan G. and Peter Dickson (1984). "Psychological Per-spectives on Consumer Response to Sales Promotion." in Re-search on Sales Promotion: Collected Papers. Katherine Jocz.ed. Cambridge, MA: Marketing Science Institute.

Schmalansee, Richard (1978). "A Model of Adveiiising and Prod-uct Quality," Journal of Political Economy, 86 (September),485-503.

Shostack, G. Lynn (1977). "Breaking Free From Product Market-ing." Journal of Marketing, 41 (April). 73-80.

Shugan, Steven M. (1989). "TTie Haagen-Dazs Dilemma: ProductAssortment in a Triopoly," Management Science, 35 (March),304-20.

Steenkamp. Jan-Benedict E.M. (1989). Product Quality. Assen/Maastricht, The Netherlands: Van Gorcum.

Szymanski. David M.. Sundar G. Bharadwaj. and P. Rajan Vara-darajan. "An Analysis of the Market Share-Profitability Rela-tionship." Journal of Marketing, 57 (July), 1-18.

TARP/Technical As.sistance Research Program (1979). ConsumerComplaint Handling In America: An Update Study. Washing-ton, DC: White House Office of Consumer Affairs.

(1981). Measuring the Grapevine: Consumer Response andWord-of Mouth. Atlanta. GA: The Coca-Cola Co.

Taylor. John (1979). "Staggered Wage Setting in A MacroModel." American Economic Review. 63 (May). 108-18.

Tse. David K. and PeterC. Wilion (1988). "Models of ConsumerSatisfaction Formation: An Extension." Journal of Marketing,25 (May), 204-12.

Yi. Youjae (1991), "A Critical Review of Customer Satisfac-tion," in Review of Marketing }989, Valarie A. Zeithmal. ed.Chicago; American Mariceting Association.

Zeithaml. Valarie A. (1981). "How Consumer Evaluation Pro-ces.ses Differ Between Goods and Services." in Marketing ofServices, James H. Donnelly and William R. George, eds. Chi-cago: AM A. 186-89.

(1988). "Consumer Perceptions of price. Quality, and Value:A Mean.s-End Model and Synthesis of Evidence," Journal ofMarketing, 52 (July). 2-22.

66 / Journal of Marketing, July 1994