Measuring and managing brand loyalty

SIMON KNOXInstitute for Advanced Research on Marketing, Cran� eld School of Management,Cran� eld University, Cran� eld, Bedford, MK43 0AL, UK

DAVID WALKERSynectics Europe, 10 Wyndham Place, London W1H 1AS, UK

There has been much written about the role of brand loyalty as a primary measure ofeffective brands marketing and a partial measure of brand equity. To date, how-ever, progress in providing a practical measure of the construct has been very limited.In this empirical study of grocery brands, such a measure was developed in which bothbrand commitment and brand support were found to be necessary and suf� cient con-ditions for loyalty to exist. Based on this measure, four consumer purchasing styles wereidenti� ed and characterized as ‘loyals’, ‘habituals’, ‘variety seekers’ and ‘switchers’. Thestrategic implications of segmenting grocery markets on this basis is discussed in boththe context of the marketing of brands and managing brand equity.

KEYWORDS: Brand loyalty; brand commitment and support; repeat purchase rate; consumerpurchasing styles

INTRODUCTION

The concept of brand loyalty has had a long and inconsequent history. The very � rst mentionof the idea was attributed to Copeland (1923) and, since then, over 200 de� nitions haveappeared in the literature (Jacoby and Chestnut, 1978). This plethora of de� nitions provides aclue as to how important this concept is in marketing theory. The generation of ‘loyal’customers has been a primary objective of marketers for decades. The level of brand loyalty hasalso been used as a measure of the success of marketing strategy and also as a partial measure ofbrand equity. In fact, Aaker (1991) stated that ‘the brand loyalty of the customer base is oftenthe core of the brand’s equity’ (p. 5).

Despite all the interest in the general concept and the universal belief in the bene� ts ofloyalty, progress in measuring and clearly de� ning it has been very limited. This is particularlytrue for fast-moving consumer goods (FMCGs) where the arguments become further entwinedwith the debate about how involving their purchases actually are for consumers. Many authors(e.g. DeBruicker, 1979; Kassarjian and Kassarjian, 1979) believe that the process of purchasing inthese markets is treated with almost total apathy by many consumers. If this is the case thenconsumer devotion to a particular brand will be fragile and short-lived so the use of ‘brandloyalty’ as a partial measure of brand equity must be questioned. McWilliam (1992) went so

JOURNAL OF STRATEGIC MARKETING 9 111–128 (2001)

Journal of Strategic Marketing ISSN 0965–254X print/ISSN 1446–4488 online © 2001 Taylor & Francis Ltdhttp://www.tandf.co.uk/journals

DOI: 10.1080/09652540010029962

far as to question whether chief executive of� cers fully understand the nature of brand equityin these markets: ‘One has to ask if these chief executives are investing their time and moneywisely and on a correct understanding of the role of brands in the lives of their customers’(p. 131).

This paper reviews the progress made in understanding brand loyalty, from philosophicalconsiderations to de� nitional issues and the practical, managerial consequences of engenderingloyal behaviour. It then provides a new framework for understanding brand loyalty which isbased on a robust de� nition and derived from this analysis of panel data from grocery markets.Using this new framework, involved purchasing is identi� ed and separated from routinizedpurchasing. The key bene� t derived from this framework and other empirical observations onbrand loyalty is providing a practical tool for market segmentation and for marketers in order forthem to understand the performance of their brands better. It also provides a more robust modelfor understanding the dynamics of brand equity in FMCG markets.

The paper concludes by considering the implications of this empirical work for brandmarketers and managers interested in trying to estimate brand equity.

THE SIGNIFICANCE OF REPEAT PURCHASING IN PERSPECTIVE

With the exception of a very few product categories, a major objective of marketing strategyis to facilitate the process of consumers repurchasing a brand through preference. The reasonwhy repeat purchasing is important to marketing management is simple: retaining customersoften requires less marketing resources than recruiting new ones (see, for example, Reichheldand Sasser, 1990; Reichheld, 1996; Birgelen et al., 1997) and, thus, is economically desirable.Not surprisingly, the vagaries of repeat purchase behaviour have been the subject of extensiveresearch efforts throughout the history of the study of consumer behaviour. However, much ofthe work has suffered from de� nitional inconsistencies and inadequate operationalization(Muncy and Hunt, 1984). There are three salient factors which contribute to these dif� culties.First, understanding repeat purchase behaviour requires behavioural data. Collecting suchdata is costly in FMCG markets since it generally requires a panel. The substitute of self-report survey data is often unreliable since it is dependent on recall of purchasing events frommemory (see Wind and Lerner, 1979). Second, the terms repeat purchase, brand commitmentand brand loyalty have all been used interchangeably in the literature so empirical � ndingsbecome dependent upon particular and often imprecise reference points. Finally, since there aretwo opposing schools of thought as to the nature of repeat purchasing, assumptions aboutdependent variables will be very different. These differing perspectives are addressed next in thefollowing sections since they fundamentally in� uence any research design and measurementinstruments.

STOCHASTIC AND DETERMINISTIC PERSPECTIVES OF REPEATPURCHASE

The stochastic model of consumer purchasing anticipates the repeat purchase patterns of con-sumers from basic information about penetration and average purchase frequency. Consumersare considered to purchase brands in a random fashion which is predictable from knownprobability distributions. Thus, the levels of repeat purchase to be expected can be predictedfrom the basic variables. Models of this nature were extensively developed by Ehrenberg (1988).The philosophical assumptions underlying such models are that the levels of repeat purchase are

112 KNOX AND WALKER

stable for a given brand penetration and purchase frequency and are not easily altered by anyreadily identi� able causative variable. In defence of this seemingly unlikely scenario, Bass (1974)stated that ‘even if behaviour is caused but the bulk of the explanation lies in a multitude ofvariables which occur with unpredictable frequency, then, in practice, the process is stochastic’(p. 7).

Application of this view of repeat purchase is limited to modelling static market situations anddoes not provide any (indeed, may even deny the existence of) causative explanation. Both thetheory underlying the models and their application was discussed extensively by Ehrenberg(1988) and, more pragmatically, by East (1997).

In contrast to this is the deterministic view of repeat purchase behaviour in which a limitednumber of causes are considered to in� uence purchasing behaviour, that is a few independentvariables account for and can even predict the level of repeat purchasing for a given brand andset of consumers. However, causal research carried out by the deterministic school has met withonly partial success. Jacoby and Chestnut (1978) suggested that the reason for this could be dueto the fact that repeat purchase behaviour is in fact multicausal (pp. 4–5). However, the sameauthors went on to note that determinism has a place in identifying models of behaviour withina limited domain of repeat purchasing, speci� cally in the area of ‘brand loyalty’.

REPEAT PURCHASING AND BRAND LOYALTY

Repeat purchase behaviour is an axiomatic term which simply refers to the extent to whichconsumers repurchase the same brand after experiencing the brand. Since it is a purelybehavioural construct, it is simply measured as the number of times a given brand is repurchasedby a consumer in any given period of time (see Ehrenberg, 1988). In contrast, the term brandloyalty is a complex concept that may require both psychological (commitment) and behaviouralmeasurements. The de� nition and measurement of brand loyalty is discussed next.

In their extensive study of the brand loyalty literature, Jacoby and Chestnut (1978) reviewedover 200 studies in an attempt to determine the nature and characteristics of brand loyalty. The� rst point that they made was to underline the general lack of a coherent conceptual de� nition:‘the concept of brand loyalty has been de� ned by most researchers empirically instead oftheoretically, a few researchers have stated that the empirical de� nition is the theoreticalde� nition’ (Woodside and Clokey (1975) in Jacoby and Chestnut (1978), p. 70).

This lack of clarity has led to a great deal of dif� culty in interpreting many of the brandloyalty studies and, since there is no common theoretical de� nition, mutually exclusive researchstudies have generally been produced.

Jacoby and Chestnut (1978) also provided a classi� cation of the various approaches tomeasuring brand loyalty. There are three broad groups.

(1) Those that stress behaviour. Examples are ‘exclusive purchase’, i.e. to be brand loyalthe consumer must consistently purchase a single brand (Copeland, 1923; Churchill,1942; Brown, 1952), ‘two-thirds criterion’, i.e. out of a set of three brands offeredfour or more purchases of the same brand must occur in a 6 week period for brand loyaltyto exist (Charlton and Ehrenberg, 1976) and ‘three-in-a-row criterion’, i.e. a consumeris brand loyal when three or more purchases in a row occur (Tucker, 1964; McConnell,1968).

(2) Those that stress psychological commitment. Examples are ‘brand preference’, i.e. aconsumer is de� ned as loyal to the brand they name in response to the question which

MEASURING AND MANAGING BRAND LOYALTY 113

brand do you prefer? and ‘brand name loyalty’, i.e. loyalty is assessed on the basis of responsesto the statement ‘I make my purchase selection according to my favourite brand name,regardless of price’.

(3) Composite indices. Examples are ‘brand insistence’, which combines the behaviouralindex of exclusive purchase with an out-of-stock decision that another brand would onlybe purchased in the case of an emergency (Copeland, 1923) and ‘Li’, i.e. the ratio ofthe proportion of purchases devoted to brand L to the proportion of purchases devoted tobrand i (Day, 1969).

More recently, Dick and Basu (1994) identi� ed a conceptual framework for brand loyalty, whichis also a composite model combining ‘relative attitude’ towards the brand and ‘repeat patronage’of the brand as the central, independent variables of the ‘loyalty relationship’.

The number and diversity of these approaches makes the task of identifying an appropriatemeasure formidable. Jacoby and Chestnut (1978) also provided a review of the comparativereliability, validity and sensitivity of the various measures. Unfortunately, this too proved ratherinconclusive. However, in drawing together their discussion on the theory and measurementof brand loyalty, these authors identi� ed a conceptual de� nition (which was � rst proposed byJacoby and Olson (1970) for which they cited extensive theoretical substantiation. The con-ceptual de� nition (which will be adopted in this research) was expressed as a set of six necessaryand collectively suf� cient conditions. Thus, brand loyalty was seen as

(1) the biased (i.e. non random), (2) behavioural response (i.e. purchase), (3) expressed over time,(4) by some decision-making unit (5) with respect to one or more alternative brands out ofa set of such brands, and (6) is a function of psychological (decision making, evaluative)processes (Jacoby and Chestnut, 1978, p. 80).

It will be clear why this de� nition is appealing from the Introduction: it provides a methodof distinguishing between apparent devotion to a brand (or portfolio of brands) and devotionwhich is founded on a longer term, more deep-rooted psychological evaluation. It is also dis-tinct from the Dick and Basu (1994) framework which offered a narrower view of loyaltyde� ned by the attitudinal and behavioural responses consumers have towards a particularbrand rather than a set of brands in their purchasing portfolio. All the evidence suggests that,in FMCG markets, where involvement is relatively low, consumers purchase on a portfoliobasis as the norm rather than display single brand loyalty (Ehrenberg, 1988; Uncles et al.,1995; Kennedy and Ehrenberg, 2000). Consequently, the authors were keen to adopt theJacoby and Olson (1970) de� nition since it was thought likely to provide the richest data setsof the underlying patterns of repeat purchasing and switching motivations within productcategories.

Operationalizing this composite de� nition of brand loyalty still proved to be enormouslychallenging. However, since this is the only fully identi� ed de� nition that acknowledges thepossibility of multibrand loyalty, even if it does not produce the perfect measuring instrument,it did provide a sound conceptual framework for this research.

DEVELOPING MEASURES OF BRAND LOYALTY:THE BACKGROUND TO THIS RESEARCH

The authors’ original interest in brand loyalty stemmed from work they were conducting intothe relationship between consumer involvement and behaviour in FMCG markets (Knox and

114 KNOX AND WALKER

Walker, 1995). (Consumer involvement is ‘the degree of personal relevance which a stimulus orsituation is perceived to help achieve consequences and values of importance to the consumer’(Peter et al., 1999).) Initially, the authors were interested in determining whether or not therewas a signi� cant relationship between a consumer’s level of involvement with grocery brandsand their repeat purchase behaviour (see Assael, 1995). In developing this research protocol, amodel (Figure 1) was identi� ed that included a measure of the sources of involvement, involve-ment itself, commitment and brand support (an index of brand portfolio size and purchasefrequency) as the � nal dependent variable. Effectively, brand commitment, i.e. the consumer’spsychological attachment to a brand (Beatty and Kahle, 1988), was treated as an antecedent tobrand support on the grounds that any variance explained in brand support by brand com-mitment could legitimately be termed ‘brand loyalty’ (under the terms of Jacoby and Olson’s(1970) de� nition outlined above). The involvement part of the model (Figure 1) stems fromempirical work carried out by Mittal and Lee (1989) and Laurent and Kapferer (1988). UsingLISREL VII to identify the model through causal path analysis and panel data collected acrossthree product categories (see the data collection section), it was possible to establish that there isa signi� cant relationship between brand commitment and repeat purchasing (brand support),albeit a very weak one, explaining only approximately 16% of the variance.

It was at this point that it was decided to review the de� nition of brand loyalty shown aboveand develop a way of classifying brand buying behaviour which is better able to retain informa-tion about consumers than the simple linear relationship speci� ed in the original model. Beforediscussing this approach, the paper now turns to reviewing the data collection procedures andproviding details of the measurements that were made.

DATA COLLECTION

In the exploration of progress made in the study of brand loyalty, it was concluded that a majorshortcoming of many research endeavours had been a failure to collect behavioural data.Self-reported measures of behaviour through surveys are notoriously unreliable because of poorrecall by consumers of what products had been purchased over long periods of time (see Wind

FIGURE 1. Schematic of involvement/support model (involvement sources and formsfollow Mittal and Lee (1989)).

MEASURING AND MANAGING BRAND LOYALTY 115

and Lerner, 1979). Therefore, in order to collect accurate data on actual purchasing behaviour,a consumer panel was recruited and operated for 16 weeks. A survey was administered to eachrespondent at the beginning of the panel recording period in order to measure involvement andcommitment states and elicit demographic information. Subsequently, participants were askedto record their purchases across three product categories on diary sheets which were suppliedevery two weeks. The product � elds investigated were paper kitchen towels, breakfast cerealsand national newspapers. These products were selected in order to re� ect the full range of levelsof consumer involvement to be found in grocery brands, from low (kitchen towels) to high(national newspapers), with breakfast cereals as a medium involving category. It was shown thatthere is a signi� cant difference in the levels of involvement between each of these product types(see the third subsection of Appendix 1 for mean scores).

SAMPLE

Since the primary interest here was in developing theory about the relationship betweenattitudes and behaviours towards brands, a sample that was representative of either the countryor of usage levels of speci� c brands was not speci� cally needed. The objective was to recruit asample that re� ected a cross-section of shoppers that spanned the main demographic variables.A clustered, random-sampling technique was employed and the new town of Milton Keynes wastargeted for recruitment. Out of the 113 Milton Keynes areas, eight were selected for recruit-ment and three to four streets in each area were used.

The aim was to recruit 200 respondents onto the panel and record their purchases in at leasttwo of the three product categories in order to produce an effective sample size of between 400and 600.

FIELDWORK

All obliging householders who had bought and used at least two of the product categoriesduring the previous month were recruited onto the panel. Upon recruitment, each member wasgiven a starter pack which consisted of the � rst questionnaire, � rst diary sheet and detailedinstructions about being a panel member. Details of how to complete the questionnaire anddiary sheets were explained to the respondents at the recruitment visit. Two hundred andninety-eight of the 540 households that were contacted successfully agreed to participate in thestudy. Two hundred and twenty-two actually returned their � rst questionnaire and 207 remainedat the end of the recording period. From this sample, 191 provided usable responses from eithertwo or three product categories, giving an effective sample size of 463 throughout the recordingperiod. Responses from the � rst two diary sheets were discarded (in order to allow behaviour tostabilize) which left data from a further eight diary sheets or 16 weeks of full panel data.

Information about current brand usage, stated preferred brands and background informationon shopping behaviour and demographic information was elicited from the self-completionquestionnaire, which was administered at the beginning of the panel recording period. Fourteenitems for assessing involvement and two further items speci� cally about brand commitmentwere also included in the questionnaire. One of the two commitment scales followed Traylor(1981) and was a simple � ve-point scale, while the other was a modi� cation of the scale used byCunningham (1967), which was used because it expresses the psychological construct of com-mitment set in a behavioural context. These involvement and brand commitment measures areincluded in the � rst subsection of Appendix 1, together with a reliability test of the involvement

116 KNOX AND WALKER

scales (second subsection of Appendix 1). Further details of the validity measures and test–retestreliability scores can be found in Knox et al. (1999).

In the 16 week period following the initial survey, participants were required to recordinformation about their purchases of all brands from the three product categories in diary sheetson the day of purchase. At the end of the two week diary period, respondents who had changedbrands within a category were requested to provide reasons for their switching behaviour.Due to the large choice of brands within each of the categories, a free response format was usedso that the diary sheets could be kept as simple as possible. Data from the diary sheets wasrecorded onto a PC database as the panel progressed. The database was also used to track diarysheet returns and signal the need for additional calls to be made on respondents who failed tomake a return.

ANALYSIS

Given the de� nition of brand loyalty identi� ed above, it was necessary to develop a measure ofbrand buying behaviour that re� ected the degree to which purchasing within a product categorywas devoted to a limited set of brands from the greater number that were available in the marketplace. Such an index was derived ‘using data on respondents’ ’ purchasing throughout the full 16week recording period, which is expressed mathematically as

Brand support index = ^brand 1

brand n5 (purchase of brand n)2

(total purchases of product)2 6 log (total purchases)

The main part of the equation is derived from the classical Hirschman–Her� ndahl index(Hirschman, 1987). The log total purchases multiplier was introduced in order to comply withthe � rst point of the conceptual de� nition of loyalty outlined above, i.e. the requirement fora non-random response. Its effect is to reduce the weight of the index for respondents whoonly made a small number of purchases in the category over the 16 week recording period.(This modi� cation was only employed in order to account for the limited recording period(16 weeks) and would be unnecessary if the recording period was longer.)

Based on aggregate data across all product categories, it has already been noted that the linearrelationship between brand commitment and the index of brand support used here was relativelyweak. The amount of variance explained in each individual product category varied,being greatest for newspapers and almost negligible for kitchen towels. Intuitively, this aggregate� nding is surprising – if consumers express commitment to brands, it would be expected thatthey would purchase them (indeed, achieving positive discrimination to brands through com-mitment is the rationale behind most branding) (McWilliam, 1992; Keller, 1993, 2000). How-ever, this result suggests that there are reasons other than commitment that lead to the repeatpurchasing of a limited number of brands. In other words, brand purchasing behaviours in someinstances may not be a direct function of psychological (decision-making, evaluative) processes,particularly with low-involvement purchases.

If brand commitment and support for each respondent is plotted in matrix format, it can beseen why the linear relationship is so weak (see Figure 2). Although there are some respondentswho fall close to the straight line between low commitment/low support and high commitment/high support, there are a signi� cant number who lie outside this area. Using the aggregate dataacross all three product categories, the matrix approach was used to classify groupings of casesusing a simple K-means clustering procedure. Four clusters were speci� ed in order to identify

MEASURING AND MANAGING BRAND LOYALTY 117

patterns outside the straight line of the linear relationship. The clustering procedure was carriedout using SPSS quick-cluster. Conveniently, this provided one cluster in each of the four cornersof the commitment–support matrix.

The characteristics of each cluster and number of respondents in each are outlined below.

(1) Cluster 1, high commitment/high support, contained 72 cases, i.e. loyals.(2) Cluster 2, low commitment/high support, contained 34 cases, i.e. habituals.(3) Cluster 3, high commitment/low support, contained 180 cases, i.e. variety seekers.(4) Cluster 4, low commitment/low support, contained 177 cases, i.e. switchers.

The switching motivations recorded by individual respondents were then cross-tabulated foreach of the four clusters. A summary of the results of this analysis is shown in Table 1. Inaddition, we have also reported the mean levels of product involvement and brand risk for eachof these clusters (Table 2).

From the sample and recording period that was available, the number of switches that tookplace during the panel recording period was limited. However, there were substantial differencesbetween the recorded motivations for each of the clusters which provided enough informationfor naming the clusters and beginning to describe the behavioural biases of respondents withineach. Interpretation of the behavioural motivations for each cluster follows next.

INTERPRETATION

The proposed cluster names (which were drawn from these data and their logical position onthe dimensions) are designated in Figure 2. There were some very interesting differences in thesaliency of reasons for switching offered by consumers in each of these clusters (highlightedin Table 1). For both the ‘loyal’ and ‘habitual’ classi� cations, the ‘out-of-stock’ situation wasmentioned most often as a reason for switching. Peter et al. (1999) also observed the samephenomenon for ‘loyals’; they may switch away from their normal brand when out of stock.

FIGURE 2. Commitment–support matrix.

118 KNOX AND WALKER

TA

BL

E 1

.R

epo

rted

sw

itch

ing

mo

tiva

tio

ns

by c

lust

er (

n=

463)

Clu

ster

nam

e

Loya

lsH

abitu

als

Vari

ety

seek

ers

Switc

hers

Switc

hing

mot

ivat

ions

Num

ber

of switc

hes

% o

fsw

itche

s

Num

ber

of switc

hes

% o

fsw

itche

s

Num

ber

of switc

hes

% o

fsw

itche

s

Num

ber

of switc

hes

% o

fsw

itche

s

Pack

agin

gN

ew p

rodu

ct t

rial

Pric

eSt

ore

loca

tion

Prod

uct

qual

ity/fe

atur

esVo

uche

rsFr

ee g

iftVa

riet

yFi

rst

choi

ce o

ut o

f sto

ckC

hild

ren’

s in

�uen

ceTe

levi

sion

adv

ertis

emen

tC

hang

ed s

tore

Tota

l rep

orte

d sw

itche

s

— 2 — — 13 — — 1 20 — — — 36

—5.

6— — 36

.1— —

2.8

55.6

— — — 100.

0

— — — 6 9 4 — 2 27 — — — 48

— — — 12.5

18.8 8.3

—4.

256

.3— — — 10

0.0

2 6 1 16 30 2 2 51 26 50 6 719

9

1.0

3.0

0.5

8.0

15.1 1.0

1.0

25.6

13.1

25.1 3.0

3.5

100.

0

2 4 30 2 10 8 7 28 7 39 6 19 162

1.2

2.5

18.5 1.2

6.2

4.9

4.3

17.3 4.3

24.1 3.7

11.7

100.

0

MEASURING AND MANAGING BRAND LOYALTY 119

Habitual buying behaviour was characterized by Kotler et al. (1996) as passive information pro-cessing and brand familiarity rather than brand conviction, which can easily be broken if avail-ability is reduced and routine purchasing patterns are disrupted. The only other major switchingmotivation for the ‘loyal’ classi� cation was ‘product quality or features’. In a separate study ofgrocery store loyalty, McGoldrick and Andre (1997) found that ‘loyal’ shoppers cited ‘productsof better quality’ as a secondary cause for switching, which tends to support the � ndings here.

There were many different reasons for switching for those classed as ‘switchers’, the mostsalient being children’s in� uence, price and variety. Bloemer and Kasper (1995) identi� edpurchasers with low brand commitment as being ‘immediately captured by another brand thatoffers a better deal, coupon, or enhanced point of purchase visibility’. Reichheld (1996) describedswitching customers as ‘gliding into your arms for a minimal price discount who then dance awaywith someone else at the slightest enticement. Coupons and price discounts � nd these customers likeheat-seeking missiles!’ (p. 82).

In the event of a brand being repeat purchased by a ‘switcher’, it is not out of loyalty butindifference; category ‘switchers’ do not perceive a signi� cant difference in quality betweenbrands that are within their purchasing portfolio.

Apart from seeking variety itself, two other important reasons for switching were found forvariety seekers – children’s in� uence and product-related features. Menon and Kahn (1995)argued that ‘variety seekers’ are motivated by a desire to maintain an optimal level of stimulationwhich they may attempt to satisfy by choosing different or novel brands. They suggested thatvariety-seeking behaviour can occur even when preferences for brands remain constant aschoice is contextual and in� uenced by usage, point of purchase displays and feelings ofmonotony and boredom. Kotler et al. (1996) also made the point that variety-seeking behaviourcan occur in situations where there are perceived brand differences so that, when switchingoccurs, it may not be because of dissatisfactions.

There is an underlying consistency in the interpretation of the data here that suggests that theclusters do indeed have a meaningful interpretation and this mapping approach provides a moreuseful method of interpreting the relationship between commitment and support than the linearanalysis.

The authors conclude that both brand commitment and brand support were found to benecessary and suf� cient conditions for loyalty to exist.

In their paper on brand loyalty and consumer satisfaction, Bloemer and Kasper (1995) alsosupported the view that brand commitment is a necessary condition for brand loyalty tooccur. They too made a clear distinction between repeat purchasing behaviour and brandloyalty, adopting the terms ‘true’ loyalty (selective brand purchasing based on commitment) and‘spurious’ loyalty (selective brand purchasing without commitment). These respective de� nitionsproposed by Bloemer and Kasper (1995) broadly correspond to the behaviours manifested by‘loyals’ and ‘habituals’ in this paper.

TABLE 2. Mean levels of brand risk and product involvement by cluster (n = 463)

Cluster name

Loyals Habituals Variety seekers Switchers

Brand riskProduct involvement

3.70*4.14*

2.47*3.24*

3.70*3.63*

2.05*2.46

* Signi� cant difference at p = 0.05 with all items recorded on seven-point agree/disagree scale.

120 KNOX AND WALKER

Referring to an earlier analysis (Knox et al., 1999) it was determined that the only ante-cedents of brand commitment that were signi� cant among these grocery brands were productinvolvement and brand risk (in the model here these antecedents operated through the inter-vening variable of brand involvement). Product involvement refers to the extent to which theproduct class is motivating for the consumer (Mitchell, 1998), whilst brand risk refers to therisks associated with making a poor brand choice (i.e. dissonance post-purchase). These con-structs were measured with a total of four items in the initial self-completion questionnaire(Appendix 1). The mean levels of these two key antecedents are reported for each cluster inTable 2.

In summary the table suggests the following.

(1) Loyals: high product involvement and medium risk.(2) Habituals: low product involvement and low risk.(3) Variety seekers: medium product involvement and medium risk.(4) Switchers: low product involvement and low risk.

This implies that the risks associated with moving out of a brand and into an alternative arekey to developing brands, the customers of which are loyal in any enduring sense. In otherwords, ‘loyals’ use repeat purchasing of a brand as a means of reducing risk (Assael, 1995).Product involvement is also important in this respect, but it is more dif� cult to see how thisconstruct could be manipulated to the advantage of a speci� c brand. Laaksonen (1999) providedan example of how this may be possible in detergents marketing. He cited the case of consumerswho have a high sense of involvement with the product category, which is normally regarded asa low-involvement purchase, due to environmental (or health) reasons. The marketing of anappropriate detergent brand that promotes green values (or non-enzyme health bene� ts) is likelyto lift sales of the subcategory in general and the brand in particular. Referring back to Table 2,this is most likely to occur among ‘loyals’ and ‘variety seekers’.

These results clearly provide clues for marketers interested in ensuring consumer preferencesand for developing appropriate advertising for grocery brands. The implications of this aredeveloped further below.

IMPLICATIONS FOR STRATEGIC MARKETERS

Like many academic researchers in the past, this research was designed with a view to pursuingthe determinant causes of brand loyalty. This proved to be a rather optimistic goal. Consumerbehaviour is a complex phenomenon and people behave differently when purchasing differentproducts. A limited relationship between commitment and repeat purchasing, which mayprovide some clues as to how to increase loyalty, was found in earlier work. However weak therelationship between involvement and loyalty might be, it is still an important one to understandbecause these loyal customers account for a disproportionately large volume of sales. Forexample, this research has shown that the 47% of newspaper purchasers in the loyal clusteraccount for 60% of the sales. However, the contribution of the exploratory work reported hereis not derived from the causal analysis but from the identi� cation of differing responses along thedimensions of brand commitment and purchasing behaviour. Four classi� cations of consumerbehaviour which marketers can begin to approach in different ways have been described here.

For instance, it has been seen that the most likely cause of a change in brand purchasing in thehabitual classi� cation is due to the out-of-stock situation. This is consistent with the instinctive

MEASURING AND MANAGING BRAND LOYALTY 121

assertions about extinguishing habitual responses (i.e. the removal of habitual choice) (see Assael,1995). It will be clear to the marketer that, for this segment of consumers, maximizing dis-tribution will be critical to the success of marketing strategy. Aaker (1991) stated that ‘A set ofhabitual buyers has considerable value because they represent a value stream that can go forwardfor a long time’ (p. 72). This is only true if steps are taken to guard against the loss of suchcustomers. This study implies that the key to achieving this is very different to building brandloyalty per se and rests with removing the opportunity of switching out of the brand by closingdistribution gaps. The assertion here is that marketers of grocery brands need to understand thathigh repeat purchase rates do not necessarily re� ect what this paper calls high levels of brandloyalty. Therefore, using national television advertising or other brand building techniquesshould not be the primary focus of maintaining this type of brand equity – unless (as is oftenthe case) it is needed to justify distribution gains with the retail trade.

In contrast, high distribution may play a less signi� cant role in retaining variety seekers.The results here imply that sampling an alternative product may not lead to anything more thantemporary defection. A more important cue for variety seekers is the provision of choice withinthe umbrella brand. Hence, the role of the brand is to provide assurances about quality whilestill allowing the consumer choice – factor which companies such as Kellogg’s certainly seem tohave capitalized on. It appears that more conscious mental processing is still important for theloyal segment and modi� cations in the brand which trigger this will be more relevant to them.The levels of product involvement are relatively high for both loyal and variety-seekingbehaviours, which implies that rational advertising messages may be appropriate, even in grocerymarkets. Advertising certainly has a role to play for the loyals in increasing the risks associatedwith moving away from the brand. The risks in question appeared to be entirely associated withproduct performance for the grocery products studied here. Other authors have reportedon FMCG products where emotional bene� ts are important. For example, Beatty and Kahle(1988) reported that connections between consumers’ life experiences and their use of softdrinks are important in establishing commitment to speci� c brands. This will be no surprise tothose familiar with the often-quoted case of Coca-Cola’s attempt to introduce a new variant ofCoke in the USA.

Finally, it should be recognized that, within the loyal classi� cation where purchasing port-folios are smaller, distribution again seems to play an important role. Out-of-stock occurrencescan cause even loyals to switch brands. In an early study of stock-out situations, Schary andChristopher (1979) observed that more brand loyal consumers chose to switch brands ratherthan to delay purchase until their next visit. Whilst the other factors this paper has describedare also important, marketers should not underestimate the impact of Loyals experiencing analternative brand.

The switchers category is perhaps the most dif� cult to target; price, variety, store choice andchildren’s in� uence all seem to play a role. A combination of various tactical promotions maybe the best suggestion to make. However, given the nature of this consumer group, these wouldhave to be viable on the basis of a single sale or provide some additional advantage to customersin the other clusters where a longer run return could be expected.

The position in the matrix of the majority of consumers at either product category or brandlevel provides a valuable contribution to the understanding of consumer behaviour in eachmarket. For example, in this research it was found that newspaper purchasers were over-represented in the loyal cluster, breakfast cereals in the variety seeker’s cluster and kitchentowels in the habitual cluster. It may also be possible to target the remaining customers in theother groups separately. The success of the latter approach is, of course, dependent on � nding

122 KNOX AND WALKER

secondary variables (eg. geodemographic or life-style variables) which are related to the clustersand can be used for targeting marketing effort, e.g. advertising and database marketing.

IMPLICATIONS FOR BRAND EQUITY

The role that brand loyalty plays in measuring brand equity is considered to vary con-siderably. Farquhar (1989) referred only to ‘favourable brand attitudes’, whereas other authors(e.g. Aaker, 1991; Kapferer, 1992; Uncles and Laurent, 1997) have attached a much more impor-tant role to the behaviour. It is, of course, only ever a partial measure; a brand can have equityfor many reasons other than its ability to achieve high repeat purchase rates, for examplebecause of a brand’s ability to achieve a price premium over competing products (Yasin,1995). However, brand loyalty certainly seems to be a key variable for management interestedin the value of brand equity when measured from a consumer perspective rather than forthe balance sheet (Keller, 2000). The research � ndings here must therefore raise questionsin the minds of those interested in the value of their brands. For example, which is themore valuable: a brand with a high level of habitual buyers or one with a high core of loyalcustomers? Aaker (1991) considered the committed, loyal buyers to be the most highly valued;this probably stems from the desire to treat direct marketing investments (e.g. advertising) asassets rather than costs which is fundamental in the management of brand equity (seeKapferer, 1992). However, if one applies the principles outlined above, there is no reasonwhy habitual consumers should not be just as valued and long-lived as the loyals. In this case,the marketing investments are just as valuable but operate through the ability of the brandmanagement to gain and hold distribution. This is a subtle but signi� cant difference since theability to distinguish the reasons for brand strength that this approach offers must surelyprovide the would-be brand acquirer with improved information over traditional measuresof brand strength such as share and sales volumes. In his discussion of brand equity andrelationship management, Ambler (1995) recognized the need for taking a more holistic view ofthe value of the brand which is grounded within a marketing rather than � nance framework.The approach used here may provide the roots of just such a framework within FMCGmarkets.

LIMITATIONS, RESEARCH DIRECTIONS AND CONCLUDING REMARKS

This paper has suggested that it is more useful to examine brand loyalty in the context of brandcommitment and brand support, with both as necessary and suf� cient conditions for loyaltyto exist. This approach allows marketing management to target marketing activity according tothe behavioural style of their customers rather than simply focus on increasing ‘loyalty’. Thispaper has shown, through empirical evidence, that the four segments de� ned in the clusteranalysis have different switching behaviours and has suggested alternative strategies for reachingeach of these.

Clearly, there is signi� cant work still to be done in clarifying the impact of these � ndings atthe individual brand level and in the measurement of brand equity. The sample resourcesavailable in the study reported here were limited in three respects. First, only data in threeproduct categories were collected, second, the sample was not large enough to isolate thespeci� c differences between brands within the product � elds and, third, the study did not dis-tinguish between the decision makers, buyers and users within a family. In particular, this paper’sexplication of how consumers purchase brands does not furnish any real understanding of how

MEASURING AND MANAGING BRAND LOYALTY 123

they use them which, in turn, may in� uence purchasing behaviours outside the domain ofrational problem solving. Further work also needs to be done in order to determine whether itis consumers who behave differently per se (i.e. whether there is a tendency for ‘loyal’ consumersfrom one category to be loyal in other categories) or whether behaviour is solely dependent onthe product category.

The current work provides most guidance to those managing mature brands, which is asigni� cant grouping, considering that over half the current top 50 brands in the UK werelaunched in the 1950s (Brady and Davis, 1993). However, the matrix can also be used foranalysing purchasing behaviours during the brand-building process. Research derived from thematrix would be a long-term longitudinal gains/loss analysis for identifying whether brandmarketers can effectively in� uence the position of consumers in their approach to purchasing(i.e. whether customers can be moved from one cluster to another) or whether they must acceptthat purchasing style is enduring and should target their marketing efforts accordingly.

Finally, marketers might also like to consider using the mapping approach outlined here inorder to help focus their marketing research efforts by purchaser type. The matrix used herehas been developed as a means of guiding decision making on where certain strategic analysistechniques should be directed and applied. For example, for product categories where theconsumers fall mainly into the loyal classi� cation, cohort analysis can identify how to targetmarketing effort towards the most important age groups. This type of analysis would be lessappropriate to the switcher classi� cation because these consumers never develop loyalty and areunlikely to carry their allegiances to a brand throughout their lives. By analysing distributionpro� les, it is possible to help make improvements in targeting distribution within a portfolioof brands in order to optimize sales in product categories where many of the consumers arehabitual purchasers. Similarly, the authors are developing models for revealing the responses ofconsumers to promotions, which is particularly relevant to consumers in the switcher andvariety seekers categories. These are only three examples of research techniques which couldbe applied to the different segments. Equally, research into other marketing mix variables couldbe developed using the loyalty framework outlined in this paper.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the � nancial assistance of the Isle of Man GovernmentDepartment of Education and the Institute for Advanced Research in Marketing, Cran� eldSchool of Management. We would also like to acknowledge the contribution of the two blindreviewers for their constructive feedback on an earlier version of this paper.

REFERENCES

Aaker, D.A. (1991) Managing Brand Equity: Capitalising on the Value of a Brand Name. New York: The FreePress.

Ambler, T. (1995) Brand equity as a relational concept. In Proceedings of the second International Colloquium inRelationship Marketing. Cran� eld School of Management, Cran� eld, November, pp. 13–20.

Assael, H. (1995) Consumer Behaviour and Marketing Action. Cincinnati: South Western College Publishing.Bass, F.M. (1974) The theory of stochastic preference and brand switching. Journal of Marketing Research,

11, 1–20.Beatty, S.E. and Kahle, L.R. (1988) Alternative hierarchies of the attitude behaviour relationship: the impact

of brand commitment and habit. Journal of the Academy of Marketing Science 16(2), 1–10.

124 KNOX AND WALKER

Birgelen, M., Wetzels, M. and De Ruyter, K. (1997). Commitment in service relationships: an empiricaltest of its antecedents and consequences. In EMAC Conference Proceedings, University of Warwick,Warwick, 20–23 May, pp. 1255–71.

Bloemer, J.M.M. and Kasper, H.D.P. (1995) The complex relationship between consumer satisfaction andloyalty. Journal of Economic Psychology 16, 311–29.

Brady, J. and Davis, I. (1993) Marketing’s mid life crisis. McKinsey Quarterly 2, 17–28.Brown, J.D. (1952) Consumer Loyalty for Private Food Brands. Muncie, IN: Bureau of Business Research,

Report No. 3, Ball State University.Charlton, P. and Ehrenberg, A.S.C. (1976) An experiment in brand choice. Journal of Marketing Research 13,

152–60.Churchill, H. (1942) How to measure brand loyalty, Advertising and Selling 35(24), 11–16.Copeland, M.T. (1923) Relation of consumer’s buying habits to marketing method. Harvard Business Review 1,

282–9.Cunningham, S.M. (1967) Perceived risk and brand loyalty. In D. Cox (ed.) Risk Taking and Information

Handling in Consumer Behaviour. Boston: Harvard University Press, pp. 132–8.Day, G.S. (1969) A two dimensional concept of brand loyalty. Journal of Advertising Research 9, 29–35.DeBruicker, F.S. (1979) An appraisal of low-involvement consumer information processing. In J.C. Maloney

and B. Silverman (eds) Attitude Research Plays for High Stakes. Chicago: American Marketing Association,pp. 112–33.

Dick, A.S. and Basu, K. (1994) Customer loyalty: towards an integrated conceptual framework. Journal of theAcademy of Marketing Science 22(2), 99–113.

East, R. (1997) Consumer Behaviour. Hemel Hempstead: Prentice Hall.Ehrenberg, A.S.C. (1988) Repeat Buying – Facts, Theory and Applications. London: Aske.Farquhar, P.H. (1989) Managing brand equity. Marketing Research 3 September, 7–12.Hirschman, E.C. (1987) Marketing research: to serve what purpose? In R.W. Belk and K. Clarke (eds)

Marketing Theory, Proceedings of the AMA Winter Educators’ Conference. Chicago: American MarketingAssociation, pp. 204–8.

Jacoby, J. and Chestnut, R.W. (1978). Brand Loyalty. New York: Wiley, pp. 67–90.Jacoby, J. and Olson, J.C. (1970) An attitudinal model of brand loyalty: conceptual underpinnings and

instrumentation research. Purdue Papers in Consumer Psychology 159, 14–20.Kapferer, J.-N. (1992) Strategic Brand Management: New Approaches to Creating & Evaluating Brand Equity.

London: Kogan Page.Kassarjian, H.H. and Kassarjian, W. (1979) Attitudes under low commitment conditions. In J.C. Maloney

and B. Silverman (eds) Attitude Research Plays for High Stakes. Chicago: American Marketing Association,pp. 3–15.

Keller, K.L. (1993) Conceptualizing, measuring, and managing customer-based brand equity. Journal ofMarketing 57, 1–22.

Keller, K.L. (2000) The brand report card. Harvard Business Review 1 January–February, 147–57.Kennedy, R. and Ehrenberg, A.S.C. (2000) Competitive brands’ user-pro� les hardly differ. In

S. Broadbent (ed) Proceedings of the MRS Conference 2000. London: The Market Research Society,pp. 42–51.

Knox, S.D. and Walker, D. (1995) Empirical Developments in the Measurement of Involvement,Brand Loyalty and their Structural Relationships in Grocery Markets. Cran� eld: Cran� eld School ofManagement Working Paper.

Knox, S.D., Walker, D. and Marshall, C. (1999) Measuring consumer involvement with grocery brands: modelvalidation and scale-reliability test procedures. In G.J. Hooley and M.K. Hussey (eds) QuantitativeMethods in Marketing. London: Academic Press, pp. 263–80.

Kotler, P., Armstrong, G., Saunders, J. and Wong, V. (1996) Principles of Marketing: The European Edition. HemelHempstead: Prentice Hall Europe.

Laaksonen, P. (1999) Involvement. In P. Earl and S. Kemp (eds) The Elgar Companion to Consumer Research andEconomic Psychology. Cheltenham: Edward Elgar, pp. 341–6.

MEASURING AND MANAGING BRAND LOYALTY 125

Laurent, G. and Kapferer, J.-N. (1985) Measuring consumer involvement pro� les. Journal of Marketing Research22, 41–53.

McConnell, J.D. (1968) The development of brand loyalty: an experimental study. Journal of MarketingResearch 5, 13–19.

McGoldrick, P.J. and Andre, E. (1997) Consumer misbehaviour: promiscuity or loyalty in grocery shopping.Journal of Retailing and Consumer Services 4(2), 73–81.

McWilliam, G. (1992) Consumers’ involvement in brands and product categories. In M.J. Baker (ed)Perspectives on Marketing Management, Vol. 2. Chichester: John Wiley & Sons, p. 131.

Menon, S. and Kahn, B.E. (1995) The impact of context on variety seeking in product choices. Journalof Consumer Research, 22, 285–95.

Mitchell, A. (1998) Loyal yes, staying, no. Management Today 5 May, 104–5.Mittal, B. and Lee, M.S. (1989) Separating brand-choice involvement from product involvement via

consumer involvement pro� les. Advances in Consumer Research 15, 43–9.Muncy, J.A. and Hunt, S.D. (1984) Consumer involvement: de� nitional issues and research directions.

In A. Mitchell Advances in Consumer Research, Vol. XI. Ann Arbor, MI: Association for ConsumerResearch, pp. 193–6.

Peter, J.P., Olson, J.C. and Grunert, K.G. (1999) Consumer Behaviour and Marketing Strategy, European edn.Maidenhead: McGraw-Hill.

Reichheld, F.F. (1996) The Loyalty Effect. Boston: Harvard Business School Press.Reichheld, F.F. and Sasser, W.E. (1990) Zero defections: quality comes to services. Harvard Business Review

5 September–October, 105–11.Schary, P.B. and Christopher, M. (1979). The anatomy of a stock-out. Journal of Retailing 55(2), 59–70.Traylor, M.B. (1981) Product involvement and brand commitment. Journal of Advertising Research 21, 51–6.Tucker, W.T. (1964) The development of brand loyalty. Journal of Marketing Research 1, 32–5.Uncles, M. and Laurent, G. (1997) Editorial. International Journal of Research in Marketing 14(5), 399–404.Uncles, M., Ehrenberg, A.S.C. and Hammond, K. (1995) Patterns of buyer behavior: regularities, models and

extensions. Marketing Science 14(3), 671–8.Wind, Y. and Lerner, D. (1979) On the measurement of purchase data: surveys vs. purchase diaries. Journal of

Marketing Research 16, 39–47.Woodside, A.G. and Clokey, J.D. (1975) A general model of brand switching behaviour. In E. Mazze (ed.)

Combined Proceedings. Chicago: American Marketing Association, pp. 175–80.Yasin, J. (1995) The effects of advertising on fast moving consumer goods markets. International Journal of

Advertising 14(2), pp. 133–47.

BIOGRAPHIES

Simon Knox is a professor of brand marketing at the Cran� eld School of Management andis a consultant to a number of multinational companies, including McDonald’s, BT, Diversey-Lever and the Ocean Group. Prior to joining Cran� eld, Simon worked for Unilever in anumber of senior marketing roles in both detergents and foods. He publishes extensively onbrand equity issues and customer purchasing styles and is the co-author of Competing on Value:Bridging the Gap between Brand and Customer Value, which was published by Financial TimesPitman Publishing in the UK and USA. Currently, he is leading research into customer relation-ship management on behalf of Computer Sciences Corporation.

David Walker is a � rst class honours graduate and has a PhD in consumer behaviour fromCran� eld School of Management. He began his career at The Planning Business where hequickly progressed to senior consultant helping companies such as United Distillers, Pedigreeand McCann Erikson with brand planning, modelling and new product development initiatives.Prior to joining Synectics he headed the brand planning group at Coca-Cola Great Britain

126 KNOX AND WALKER

where he was responsible for developing future focused brand strategies and identifying theopportunities needed to grow some of the world’s most successful brands further. During histime at Coca-Cola he was instrumental in implementing signi� cant changes in the way planningoperated which resulted in greater creativity and improved integration of consumer insightswithin the organization.

APPENDIX 1: MEASURING INVOLVEMENT AND BRAND COMMITMENT

The measurement device

(1) Product involvementI have a strong interest in . . . . .

(2) Brand decision involvementI would choose my brand of . . . . . very carefully.

(3) Product sign

(3.1) Using . . . . . helps me express my personality.(3.2) Knowing whether or not someone uses . . . . . tells a lot about that person.

(4) Product hedonic

(4.1) I would give myself great pleasure by purchasing . . . . . (4.2) To buy . . . . . would be like giving myself a present or treat.

(5) Product utility

Using . . . . . would be bene� cial.

(6) Brand sign

You can tell a lot about a person from the brand of . . . . . s/he buys.

(7) Brand hedonic

(7.1) I believe differing brands of . . . . . would give different amounts of pleasure.(7.2) All brands of . . . . . would not be equally enjoyable.(7.3) No matter what brand of . . . . . you buy, you get the same pleasure.

(8) Brand risk

(8.1) When you buy . . . . ., it is not a big deal if you buy the wrong brand by mistake.(8.2) It is very annoying to buy a . . . . . which is not right.(8.3) A bad buy of . . . . . could bring you trouble.

All 14 items used seven-point strongly agree/disagree scales.

(9) When buying the products, how committed are you to buying your favourite brands,rather than an alternative brand? (please circle the number on the scale which best re� ectsyour opinion):

I am . . . . .

| 5 | 4 | 3 | 2 | 1 |highly not at allcommitted committed

MEASURING AND MANAGING BRAND LOYALTY 127

. . . . . to buying my favourite brand(s) of . . . paper kitchen towel,. . . breakfast cereal,. . . national newspaper.

(10) If you could not get your favourite brand(s) of . . . . . paper kitchen towel, etc., . . . . . at thestore you had gone to for them, would you:

Please tick oneof the options

Happily buy a different brand ` 1

Reluctantly buy a different brand ` 2

Not buy the product until the next time you shopped ` 3

Try a different shop ` 4

Keep trying different shops until you got the brand you wanted ` 5

TABLE A1. The reliability test (involvement)

Measurement item Reliability measures –Cronbach’s a coef�cient

Product involvementBrand decision involvementProduct utilityProduct signBrand signProduct hedonicBrand hedonicBrand risk

0.940.940.890.870.870.840.720.72

TABLE A2. The mean scores (involvement)

Product category means

Kitchen towels Cereals Newspapers

Product involvementBrand decision involvementBrand hedonicBrand riskBrand signProduct hedonicProduct signProduct utility

4.39*4.02*4.55*4.85*5.175.86*5.342.35*

3.73*2.633.013.754.87*5.485.37*2.79*

3.012.432.803.552.974.613.663.29

* Signi� cant difference at p = 0.05Source: Knox et al. (1999).

128 KNOX AND WALKER