capturing consumer

7/29/2019 Capturing Consumer

1/14

Capturing consumerheterogeneity in loyalty

evolution patternsKanghyun Yoon

College of Management, Long Island University,Brookville, New York, USA, and

Thanh V. TranCollege of Business Administration, University of Central Oklahoma,

Edmond, Oklahoma, USA

Abstract

Purpose This paper aims to propose a finite-mixture brand-choice model, which relaxes thecommon dichotomous assumption in market segmentation based on consumer loyalty. The model canbe used to estimate the optimal number of consumer segments and capture both across- andwithin-household heterogeneity in consumers evolutional patterns of loyalty behaviors.

Design/methodology/approach The proposed model is empirically calibrated on a scanner paneldata of two product categories, i.e. liquid detergent and toilet tissue obtained from the ERIM scannerpanel of AC Nielsen.

Findings Empirical results suggest that: the dichotomous classification of consumers into loyal andnon-loyal groups is sub-optimal in capturing the heterogeneity in consumer loyalty. Two evolutionalpatterns of loyalty behaviors exist, showing that past experience with a particular brand may eitherincrease or decrease the current level of loyalty, and consumers sensitivity to various marketingvariables differs between repeat and switching decisions (i.e. within-household heterogeneity).

Research limitations/implications The proposed model does not capture consumers

inertia/variety-seeking tendencies. Furthermore, it does not account for the role of scale parameterin capturing consumer heterogeneity.

Originality/value This study contributes to the stream of research on consumer loyalty andprovides important marketing implications for practitioners in designing effective targeting programsas well as attraction/retention strategies.

Keywords Consumer behaviour, Customer loyalty, Brand loyalty, Marketing strategy,Customer retention

Paper type Research paper

Introduction

Past research has devoted an immense amount of effort to help marketing practitionerssegment a target market. Among the many classification frameworks are two popularapproaches:

(1) the hard-core loyal (HCL; Colombo and Morrison, 1989); and

(2) proportion of purchase (POP; Krishnamurthi and Raj, 1991) approaches; both ofthem assume that there exist two distinct groups (or segments) of consumers inthe market, i.e. HCLs and potential switchers (PSs) under the HCL approach,and loyal and non-loyal customers under the POP approach.

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/2040-8269.htm

Capturingconsumer

heterogeneity

649

Management Research Review

Vol. 34 No. 6, 2011

pp. 649-662

q Emerald Group Publishing Limited

2040-8269

DOI 10.1108/01409171111136185


2/14

Though popular, these dichotomous classification approaches are subject to thefollowing limitations.

First, these dichotomous classification approaches seem to overlook (andconsequently,restrict the capability of the respective models in capturing) consumer heterogeneity in

loyalty. In other words, it is too restrictive to assume that a market only consists of twoconsumer segments loyal and non-loyal. In practice, consumers exhibit a variety ofpurchasing behaviors (McAlister and Pessemier, 1982; Lattin, 1987; Bawa, 1990; amongothers); accordingly, we extend this stream of research by proposing a new classificationframework that relaxes this dichotomous assumption and captures the heterogeneity inconsumer loyalty in a greater extent. Specifically, we address the question: what is theoptimal number of segments (i.e. the one that captures the heterogeneity in consumerloyalty most effectively) that marketer should divide the market into?

Next, the above-mentioned dichotomous classification approaches are based on thebehavioral loyalty perspective (Jacoby and Chestnut, 1978; Krishnamurthi and Raj, 1988,1991), under which loyal consumers are defined as those who exhibit high cumulativeshare (or repeat purchase rate) of a specific brand in a product category. Accordingly, theHCL approach defines hard-core loyal consumers as those who exhibit a loyalty level ofmore than 99.9 per cent (i.e.absoluteloyalty); in contrast,the POP classifies consumersintoloyals and non-loyals using the predetermined 50 per cent cut-off point of loyalty.In addition, implicit in these approaches is the assumption of repurchase exclusivity(i.e. loyal consumers, who show high level of loyalty to a specific brand up to time t2 1,are assumed to repeat purchase this brand at time t; Banasiewicz, 2005). This results in astatic classification that does not capture the dynamic nature of loyal behaviors, i.e. theimpact of past repeat-purchase behavior on current purchase decision and subsequently,the loyalty level of a consumer. Here, we address this limitation by capturing theheterogeneity in consumers evolutional patterns of loyal behaviors. Finally, past researchbased on these dichotomous classification frameworks also assumes that consumers in

each segment are homogeneous in their intrinsic preferences and responses to changes inmarketing variables such as price across purchase occasions. This within-householdhomogeneity assumption implies that consumers respond to marketing variables in thesame way regardless of whether they repeat purchase the same brand or switch to adifferent one. Accordingly, the differences in consumers motivations upon making thesetwo decisions have not been accounted for appropriately; this issue is addressed by ourproposed model which provides separate estimates of sensitivity to various marketingvariables when consumers make repeat-purchase and brand-switching decisions.

In this paper, we develop a finite-mixture brand-choice logit model and subsequentlycalibrate it on the scanner data of liquid detergent and toilet tissue. First, we relax thedichotomous assumption (i.e. consider the possibility that the market may consists ofmore than two segments of consumers) and investigate the optimal number of consumer

groups to segment the market empirically; this optimal number of segments isdetermined based on the measure of model fit, i.e. using the Akaike Information Criterion(AIC) and the Bayesian Information Criterion (BIC) indices. Our finding shows that thedichotomous classification is sub-optimal in capturing the heterogeneity in consumerloyalty; a four-segment classification turns out to be the most efficient in the two productcategories (i.e. liquid detergent and toilet tissue).

Next, we capture the impact of past repeat-purchase behavior on the evolution ofconsumers loyal behavior by introducing a variable measuring consumer loyalty

MRR34,6

650


3/14

(which is operationalized as the share of a focal brand in a consumers purchase history)into the membership function of the finite-mixture model. This framework allowsmarketers to describe the various patterns of consumers loyal behavior in a dynamicmanner, based on the extent of prior loyalty level. We find that there exist two evolving

patterns of loyal behavior:

(1) past experience with a particular brand increases, i.e. reinforces the currentloyalty level; and

(2) past experience with a particular brand results in boredom (or other switchingmotivations) and, therefore, decreases the current level of loyalty.

Finally, we also investigate the differences in consumers motivations andconsequently, responses to various marketing variables for instance, prices between making repeat and switching behaviors within each of the identified segments.To do so, we relax the within-household homogeneity assumption in our model andcapture both across- and within-household heterogeneity (i.e. between repeat and

switching decisions) in consumers sensitivity to various marketing variables. Ourempirical findings provide the evidence for the existence of both types of heterogeneityand important managerial implications about how to segment the market as well as howto design effective loyalty programs.

The paper is organized as follows. In the next section, we briefly review the relatedliterature. The model and data description section describes the finite-mixturebrand choice logit model as well as the data used in our analysis. The results areprovided and discussed in the following section. The final section concludes the paper.

Literature reviewCurrent approaches of consumer classification

Past research on segmentation has proposed various approaches to classify consumersinto different segments. Typically, most of them rely on the assumption that there existtwo groups of consumers loyal and non-loyal (Jacoby and Chestnut, 1978; Lattin, 1987;Bawa, 1990) or inertial and variety-seeking (Jeuland, 1979; Givon, 1984; Seetharamanand Chintagunta, 1998) in the market. Among these dichotomous classificationframeworks, the HCL approach (Colombo and Morrison, 1989; Kamakura and Russell,1989; Grover and Srinivasan, 1992; Dillon and Gupta, 1996; Yim and Kannan, 1999) andthe POP approach ( Jacoby and Kyner, 1973; Jacoby and Chestnut, 1978; Krishnamurthiand Raj, 1991) have been popular in practice. Based on the perspective of behavioralloyalty, both approaches operationalize loyalty as the proportion of purchases devotedto each brand in a consumers purchase history; consequently, loyal (non-loyal)consumers are those who exhibit high (low) loyalty level.

Under this assumption, the HCL approach (which is a revised version of themover-stayer framework; Blumen, et al., 1955; Goodman, 1961) employs the 99.9 per centlevel of loyalty as the cut-off for identifying the two groups of consumers, i.e. HCLs andPSs. Stated differently, using this approach, consumers are classified into either the HCLsegment if they purchase only one brand exclusively during the entire purchasehistory or the PS segment, otherwise. In contrast, the POP approach uses thepredetermined 50 per cent loyalty level (as a median value) to classify consumers intoloyal and non-loyal segments for each brand.

Capturingconsumer

heterogeneity

651


4/14

The above-mentioned assumption seems to restrict the capability of both the HCLand POP approaches in capturing the heterogeneity in consumer loyalty. Accordingly, itremains unclear about the optimal number of segments that can capture the mostheterogeneity in consumer loyalty. Further, an important dimension of consumer

heterogeneity, which is the dynamic evolution patterns of loyalty, is not accounted forunder these approaches. Specifically, past research in this stream assumes thatconsumers who have shown high loyalty level in the past will repeat purchase the samebrand (i.e. repurchase exclusivity; Banasiewicz, 2005); as a result, the impact of past(repeat-) purchases on current purchase decision and loyalty (as well as the impact ofcurrent on future purchase decision) is not accounted for. In practice, consumerspreferences are changing over time. For instance, Heilman et al. (2000) demonstrate thatconsumers preferences may either strengthen or weaken (i.e. both the increasing anddecreasing patterns may exist simultaneously). In this study, we fill this gap in theliterature by investigating (and capturing) the heterogeneity in consumers evolutionpatterns of loyalty (i.e. the impact of past loyalty on current repeat/switching behaviors).

Across- and within-household heterogeneityAnother related stream of research is the one that investigates consumers within-household heterogeneity. In the frequently purchased product category, past researchhas shown that consumers tend to develop their own decision-making heuristic(s) whenmaking purchase decisions (Deshpande et al., 1982; Deshpande and Hoyer, 1983;Heilman et al., 2000). These heuristics (aka. choice tactics or choice strategies) arecreated and subsequently updated as consumers accumulate past purchaseexperience, and include:

. brand loyalty (or habit); and

. variety-seeking tactics (Hoyer and MacInnis, 2007).

In practice, when loyal consumers repeat purchase the same brand, their decision is (mostlikely) driven by their brand preference as a result of the unique value of this brand(Ailawadi et al., 2001). In contrast, when switching to a different brand, they may bemotivated by the attractiveness of new brands promotional incentives (such as price-cut).Analogously, non-loyal consumers may repeat purchasethe same brand in response to theaggressive economic incentives provided by that specific brand (Brown, 1974; Schneiderand Currim, 1991) or switch to another brand due to their desire for variety (McAlister andPessemier, 1982; Givon, 1984; Seetharaman and Chintagunta, 1998). This suggests thatboth loyal and non-loyal consumers may exhibit different levels of sensitivity to variousmarketing variables especially to prices when making repeat purchases andswitching brands (as a result of the different motivations underlying these decisions).

Recently, Yoon (2008) and Yoon and Kwak (2009) demonstrate that consumers are

more likely to use the habit tactics when making repeat purchases. These studies suggestthat consumers may exhibit different sensitivities to various marketing variables whenmaking repeat and switching behaviors. We add to this stream of research byinvestigating the magnitude of consumers within-household heterogeneity and how itaffects their sensitivities to prices upon making repeat purchases versus switchingbrands. Overall, our paper provides more insights into consumers loyal behavior andsuggests a more efficient framework to segment the consumer market. In the next section,we outline the model and describe the data used in our empirical analysis.

MRR34,6

652


5/14

Model and data descriptionBased on the framework proposed by Kamakura and Russell (1989), we develop afinite-mixture brand choice logit model which assumes that the market could be(optimally) divided into Ssegments based on consumers heterogeneity regarding their

loyalty levels; S is estimated subsequently. Here, by relaxing the dichotomousassumption, which is used under the HCL and POP approaches, our proposed modelallows us to capture optimally the across-household heterogeneity. Given thathousehold h belongs to segment s s 1; 2; . . . ; S, the utility that this householdderives upon purchasing brand j j 1; 2; . . . ; J at purchase occasion tis given by:

Vhtjjs IRphjtXb

Rps I

Sws Xb

Sws

IRphjt

bRpjs

bRpPs pricehjt bRpFs featurehjt b

RpDs displayhjt

ISwhjt bSwjs b

SwPs pricehjt b

SwFs featurehjt b

SwDs displayhjt 1hjt;

1

where IRphjt 1 and ISwhjt 0 if the purchase of brand j in occasion tis a repeat purchase,

andIRphjt 0 andISwhjt 1 otherwise;bRps and bSws are the repeat and switching parametervectors (respectively) associated with the vector of marketing variables X, whichincludes price, feature and display. Here, consumers within-household heterogeneity(i.e. between repeat-purchasing and switching behaviors) is captured by the two sets ofparameters, i.e. bRps and b

Sws . This framework eventually allows us to understand the

different types of purchase motivations used by consumers when making repeatpurchases and switching brands. Assuming that the error term, 1hjt, follows the extremevalue distribution (Guadagni and Little, 1983; Ben-Akiva and Lerman, 1985), we derivethe probability that household h chooses brandj on purchase occasion t, conditional on hbelonging to segment s at time purchase occasion tas follows:

Phtjjs

exp VhtjjsPSk1 exp Vhtjjk

:

2

Next, the probability that household h is a member of segment s depends on the rate ofrepeat purchase, mshrht this variable represents consumer hs level of loyalty atpurchase occasion t and is operationalized as the maximum share of all the brandspurchased by this household up to time t; mshrht is updated at each purchase occasion and is given by:

Phts exp lstPS

k1 exp lkt; 3

where

lst as bs mshrht;

mshrht max{sharehtj}wherej 1; 2; . . . ;J;

andsharehtj is the share? of brandj in the purchase history up to t:

8>>>:

Note thatsharehtj and consequently, mshrht is updated in each purchase occasion(Krishnamurthi and Raj, 1988, 1991); this variable allows us to capture the evolution ofloyal behavior of consumers. This is similar to using a concomitant variable in themembership function of the finite-mixture model (Gupta and Chintagunta, 1994).

Capturingconsumer

heterogeneity

653


6/14

Given the data consisting ofNobservations and Jbrands, the likelihood function ofthe sample is given by:

L YNn1

YHh1

XSs1

PhtsYJj1

Phtjjsdhjt" #( );

4

where dhjt 1 if household h buys brand j on occasion t and dhjt 0 otherwise. Themodel is estimated using the maximum likelihood method and the optimal number ofsegments (i.e. S) is determined by carrying out the estimation for S 1,2,3,4,5, . . . ,andthen comparing the resulting model fit using the AIC and AIC (i.e. the AIC and BIC;Bucklin and Gupta, 1992) as follows:

AIC 22 LL2kN

; and

BIC LL2 k2

logN where k is the number of parameters:

Here, it is worthwhile to compare our model to two popular ones suggested byKamakura and Russell (1989) and Guadagni and Little (1983). Unlike Guadagni andLittles (1983) approach of using the loyalty variable in the utility function to capture theeffect of brand loyalty on current brand choice decisions, our model includes the loyaltyvariable in the membership function to segment the market (based on consumersevolutional patterns of loyalty). Note that Guadagni and Little (1983) operationalize thebrand loyalty variable using the popular exponential smoothing method, while ourstudy uses the maximum of brand-specific shares, which varies over time in thepurchase history of consumers, to measure their loyalty levels (Krishnamurthi and Raj,1988, 1991). As a result, Guadagni and Littles (1983) model captures the heterogeneity inconsumers brand loyalty in a cross-sectional perspective (Fader and Lattin, 1993), whileour model captures the heterogeneity in consumers loyalty behaviors in a dynamic way

and, consequently, identifies the various evolutional patterns of loyalty behaviors.Next, our study is distinct from the study of Kamakura and Russell (1989) in the

following sense. Their study uses a finite-mixture model to capture the heterogeneityin consumers brand preferences and responses to marketing variables in a staticperspective (i.e. consumers are classified into segments at the household level).In contrast, our model classifies consumers into segments using the loyalty variable which are measured at the purchase-occasion level in the membership function of thefinite-mixture model. This allows us to understand how consumers loyal behaviorsevolve over time and how these evolutional patterns are affected by other determinants,such as brand-intrinsic preferences and promotions (as well as other marketing mixvariables).

The proposed model is calibrated on the scanner data of liquid detergent and toilettissue, obtained from the ERIM scanner panel of A.C. Nielsen. The data preparationprocess is similar to that of Krishnamurthi and Papatla (2002). Table I provides adescription of several key features including brand shares, average price andpromotional frequencies.

Liquid detergentWe consider the top five brands, i.e. Surf, Wisk, Era, Tide and Bold, which account formore than two-thirds (67.1 per cent) of the purchases in this category. We qualify

MRR34,6

654


7/14

households for inclusion into the analysis using two criteria. First, each householdshould have devoted at least 50 per cent of its purchases of the two types of detergents,i.e. powder andliquid, to liquids. Second, each household should have made at least threepurchases of the selected brands over the 138 weeks of the panel (from week 1 of 1986 toweek 34 of 1988). This effort results in a pool of 1,397 households and a total of14,082 observations. After stabilizing the loyalty level of each consumer and deletingthe first three observations of each household, we eventually obtain a sample of5,357 observations for our analysis with 1,202 households.

Toilet tissuesWe consider the top five brands, including Northern, Charmin, White Cloud, Cottonelle

and Scott, which, together, account for more than 80 per cent of the purchases in thiscategory. There are a total of 207,663 purchases of these five brands by 10,043households. The above-mentioned selection procedure is used to obtain a pool of 8,105households and 203,124 purchase records. Again, after stabilizing the loyalty level ofeach consumer and deleting the first three observations of each household, weeventually obtain a sample of 5,650 observations for our analysis with 402 households.

FindingsIn this section, we discuss the results of our empirical analysis, including:

. the optimal number of consumer groups;

. the parameter estimates that demonstrate consumers heterogeneity in the

evolution patterns of loyalty (i.e. across-household heterogeneity); and. the evidence of within-household heterogeneity across purchase occasions

(i.e. between repeat-purchase and switching behavior).

The optimal number of consumer groupsFollowing Gupta and Chintagunta (1994), we determine the optimal number ofconsumer groups (i.e. S) by calibrating our model with 1, 2, 3, 4, 5, . . . , segments untilthere is no significant improvement in the BIC and log-likelihood upon adding

Average unitshelf price

Average unitpaid price

Percentage ofpurchase on

Product category Market share Display Feature

Liquid detergentSurf 10.4 5.22 3.83 1.45 1.71Wisk 33.1 4.90 3.71 4.33 4.53Era 24.7 5.98 5.21 1.49 1.79Tide 24.1 6.09 5.19 4.25 2.85Bold 7.6 6.29 5.09 0.23 0.43Toilet tissueNorthern 30.7 0.32 0.27 4.31 10.48Charmin 27.6 0.32 0.25 5.57 10.36White cloud 6.8 0.42 0.36 2.15 2.30Cottonelle 13.5 0.33 0.25 2.39 4.54Scott 21.5 0.21 0.22 4.91 7.05

Table I.Descriptive statistics of

each product category

Capturingconsumer

heterogeneity

655


8/14

an additional segment. The results, as summarized in Table II, show that thefour-segment specification (i.e. S 4) provides the best model fit in both the liquiddetergent and toilet tissue categories.

Heterogeneity in the evolution pattern of consumer loyaltyGiven the optimalnumber of consumer segments (i.e. S 4), we estimate the parametersand identify the characteristics of these segments with respect to the evolutionalpatterns of consumer loyalty (i.e. across-household heterogeneity). The results aresummarized in Tables III and IV.

Product category Model spec. Number of parameters LL AICb BICc

Liquid detergent One segment 14 23,660.46 1.3718 23,720.56(n 5,357, H 1,202) Two segments 30 23,372.30 1.2702 23,501.09

Three segments 46 23,251.92 1.2312 23,449.40Four segments

a62 23,071.92 1.1700 23,338.09

Five segments 782

3,068.23 1.17462

3,403.09Toilet tissue One segment 14 24,968.93 1.7638 25,029.41(n 5,650, H 402) Two segments 30 24,421.63 1.5757 24,551.22

Three segments 46 24,307.56 1.5410 24,506.26Four segmentsa 62 23,842.89 1.3822 24,110.71Five segments 78 23,834.09 1.3848 24,171.03

Notes: aModel selected for each productcategory;bAIC 22 (LL 2 k)/N; cBIC LL2 (k/2) log(N),where k is number of parameters

Table II.The model fit result

Variables Segment 1 Segment 2 Segment 3 Segment 4

RepeatSurf 21.39 (22.15) 22.09 (23.73) 5.02 (1.72) 26.60 (21.98)Wisk 20.84 (21.48) 23.96 (24.67) 3.48 (4.12) 29.83 (21.56)Era 5.58 (6.45) 2.02 (1.96) 4.21 (1.91) 22.09 (21.29)Tide 4.77 (6.51) 2.18 (2.55) 3.99 (1.91) 22.04 (21.20)Bold 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)Price 23.51 (27.83) 27.32 (27.85) 20.38 (22.66) 24.35 (22.71)Feature 20.13 (20.12) 5.78 (2.82) 0.34 (0.22) 20.47 (20.48)Display 20.93 (21.51) 20.19 (20.10) 21.31 (20.80) 1.19 (1.27)SwitchSurf 23.82 (25.01) 22.79 (24.63) 217.46 (23.75) 1.26 (3.14)Wisk 215.18 (23.02) 25.87 (24.15) 218.63 (23.15) 0.72 (1.46)Era 22.08 (22.37) 1.45 (2.34) 210.17 (23.36) 1.31 (3.75)Tide 21.20 (21.29) 1.50 (2.45) 29.20 (22.71) 1.27 (3.81)

Bold 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)Price 23.50 (27.88) 27.33 (29.99) 211.33 (24.83) 20.54 (24.16)Feature 0.81 (0.84) 20.19 (20.36) 1.17 (0.93) 0.95 (2.96)Display 1.06 (0.68) 1.04 (2.06) 20.19 (20.06) 0.33 (1.22)Segment parameters a 24.78 (27.37) 1.09 (2.82) 20.95 (22.97) 0.00 (0.00)

b 7.17 (11.05) 20.92 (21.23) 2.11 (3.96) 0.00 (0.00)Segment size 0.393 0.401 0.130 0.076

Note: t-ratios are provided in parentheses

Table III.The estimatedparameters of the liquiddetergent category

MRR34,6

656


9/14

The estimates of the parameters of the membership function (i.e. a and b) in bothproduct categories show two evolution patterns of consumer loyalty. The first patternis characterized by the increase in consumers loyalty level upon making additionalpurchases of the same brand as reflected by the positive estimates of b; this pattern can

be observed in segments 1 and 3 of the liquid detergent and segments 2 and 3 of thetoilet tissue (Tables III and IV). In contrast, consumers in segments 2 of the liquiddetergent and 1 of the toilet tissue show a different pattern of loyalty evolution; thissecond patterns is characterized by a decrease in loyalty level as consumers makeadditional purchases of the same brand. (Note that segment 4 is the benchmark, whosemembership parameters are normalized for the purpose of identification.)

Interestingly, another aspect of across-household heterogeneity can be identified whencomparing the two segments in each product category that exhibit the increasing patternsof consumer loyalty (i.e. segments 1 and 3 of the liquid detergent and segments 2 and 3 ofthe toilet tissue); one segment (i.e. segment 1 of the liquid detergent and segment 3 of thetoilet tissue) show a relatively high-price sensitivity compared to that of the other segment(i.e. segment 3 of the liquid detergent and segment 2 of the toilet tissue; see the repeat

estimates in Tables III and IV). This suggests that the reinforcing effect of repeatpurchases resulting in the above-mentioned increasing patterns of loyalty amongconsumers belonging to these segments is driven by different factors. Specifically, those insegment 1 of the liquid detergent and segment 3 of the toilet tissue are quite sensitive inprices upon making repeat purchases and, therefore, are probably motivated to reinforcetheir loyalty following the intensive price promotions offered by the same brand(s).In contrast, consumers in segment 3 of the liquid detergent and segment 2 of the toilettissue focus less on prices (as shown by the lowestimates of price sensitivity); instead, they

Variables Segment 1 Segment 2 Segment 3 Segment 4

RepeatNorthern 5.51 (3.25) 5.81 (0.53) 20.55 (20.34) 13.12 (14.20)

Charmin 7.64 (9.03) 7.42 (0.28)2

3.18 (2

1.55) 13.19 (15.34)White cloud 11.84 (12.95) 19.76 (0.23) 1.81 (1.24) 15.11 (16.77)Cottonelle 9.08 (9.81) 7.52 (0.68) 0.91 (0.73) 14.49 (16.68)Scott 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)Price 27.08 (28.17) 20.24 (20.28) 22.48 (23.11) 27.09 (210.19)Feature 8.91 (4.95) 6.27 (0.44) 1.88 (1.37) 0.45 (0.76)Display 4.48 (5.90) 21.69 (20.10) 1.39 (2.19) 1.08 (2.99)SwitchNorthern 4.96 (7.08) 0.78 (0.54) 21.09 (25.14) 12.55 (13.65)Charmin 4.12 (4.03) 212.47 (20.24) 21.30 (25.68) 12.29 (12.93)White cloud 7.83 (6.58) 18.64 (0.20) 21.43 (28.41) 14.29 (15.95)Cottonelle 6.07 (10.11) 16.52 (0.18) 20.69 (23.43) 12.83 (15.11)Scott 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)Price 27.15 (216.37) 231.92 (20.19) 20.04 (213.35) 27.03 (217.23)Feature 4.76 (7.70)

22.84 (

20.05) 1.42 (3.14) 0.72 (2.49)

Display 3.99 (4.93) 9.53 (0.33) 1.40 (6.01) 1.23 (3.04)Segment parameters a 1.11 (1.03) 24.21 (27.21) 21.67 (24.24) 0.00 (0.00)

b 23.96 (21.22) 5.71 (7.50) 0.85 (0.83) 0.00 (0.00)Segment size 0.137 0.067 0.189 0.607

Note: t-ratios are provided in parentheses

Table IV.The estimated

parameters of the toilettissue category

Capturingconsumer

heterogeneity

657


10/14

seem to be attracted by the unique characteristics of their favourite brand(s) as reflected inthe significant estimates of the intrinsic brand-specific constants (i.e. 5.02 for Surf insegment 3 of liquid detergent and 5.81 for Northern in segment 2 of toilet tissue). Thisindicates that some consumers probably develop their increasing (evolutional) patterns of

loyalty due to the unique characteristics of these brands.

Within-household heterogeneityThe results, as summarized in Table V, show that within-household heterogeneity inconsumers responses to market variable(i.e. upon making repeat purchases and switchingto different brands) does exist. Here, we report the characteristics of the four identifiedsegments, including the size and magnitude of price elasticity of each segment, amongothers. Accordingly, we observe threedifferent patterns of within-household heterogeneity.

Specifically, the first pattern as observed in segments 1 and 3 of the liquid detergentand segment 2 of the toilet tissue shows significantly higher price elasticity whenconsumers switch brands than when they repeat purchase the same one. This impliesthat these consumers develop their brand loyalty (i.e. repeat purchase the same brand)

based on the unique features of the brand rather than on its attractive prices andswitch to other brands only when these competing brands provide significant economicincentives (via low prices). In other words, within-household heterogeneity does exist inthese consumer segments.

Next, segments 4 of the liquid detergent and 3 of the toilet tissue exhibit a reversepattern of within-household heterogeneity. Consumers in these segments aresignificantly more price sensitive when making repeat purchases than when switchingto other brands. This behavior can be driven by the variety-seeking tendency of theseconsumers; they seem to get bored with the previously purchased brands and enjoythe variety offered by different brands. These consumers will repeat purchase the samebrand only when it has significantly attractive prices.

Finally, and interestingly, the third pattern is characterized by relatively similarmagnitudes of price elasticity under repeat-purchase and switching decisions; thispattern is observed in segment 2 of the liquid detergent and segments 1 and 4 of the toilettissues. This suggests that for consumers in these segments, attractive prices do notseem to affect their choice among brands they respond to prices in the same way no

Product category Segment 1 Segment 2 Segment 3 Segment 4

Liquid detergentSegment size 0.393 0.401 0.130 0.076Repeat 20.86 (4.62) 21.03 (3.77) 20.14 (5.33) 24.16 (5.27)Switch 24.26 (4.11) 23.24 (3.92) 29.98 (4.07) 22.33 (5.83)Average of maximum sharea 0.943 (0.095) 0.518 (0.127) 0.698 (0.175) 0.593 (0.194)

95 per cent CI (0.939, 0.947) (0.512, 0.523) (0.574, 0.612) (0.685, 0.711)Toilet tissueSegment size 0.137 0.067 0.189 0.607Repeat 22.52 (0.254) 20.08 (0.304) 22.69 (0.229) 210.08 (0.268)Switch 22.54 (0.246) 214.59 (0.297) 20.04 (0.219) 211.92 (0.256)Average of maximum share 0.381 (0.069) 0.803 (0.149) 0.567 (0.195) 0.549 (0.175)95 per cent CI (0.376, 0.386) (0.788, 0.818) (0.555, 0.579) (0.543, 0.555)

Note: aThe value in parentheses are the standard deviation of maximum value

Table V.The estimates of priceelasticity of eachconsumer segments

MRR34,6

658


11/14

matter of repeat purchasing the same brand or switching to a new one. In other words,the above-mentioned within-household heterogeneity does not seem to exist in theseconsumer segments.

In summary, various patterns of within-household heterogeneity do exist among

consumers belonging to different segments, suggesting that consumers follow avariety of motivations when making different purchase decisions.

Conclusion and future researchThis study addresses three issues related to the popular dichotomous segmentationframeworks:

(1) the number of consumer groups that marketer should segment the market inorder to optimally capture the heterogeneity in consumers loyalty behavior;

(2) the evolutional patterns of consumer loyalty that characterize these consumersegments (i.e. the across-household heterogeneity); and

(3) the within-household heterogeneity reflected in different responses to marketingvariables specifically, prices when making repeat purchases and switching todifferent brands, as well as the possible explanation of these behaviors.

First, we find that the traditional dichotomous classification framework is sub-optimalin capturing the heterogeneity in consumers loyalty. Instead, the four-segmentclassification turns out to be the most efficient in both product categories used in ourstudy. Most importantly, these four segments are characterized by two differentevolutional patterns of loyalty:

(1) an increasing (i.e. reinforcing) pattern of loyalty upon each additional purchaseof the same brand; and

(2) a decreasing pattern of loyalty when consumers happen to purchase a specific

brand they do so probably due to the various economic incentives offered bythis brand.

Finally, we also find the evidence showing the existence of within-householdheterogeneity in price elasticity upon repeat purchasing and switching brands.

Given the two types of evolutional patterns of consumer loyalty as identified above,marketing practitioners can increase the effectiveness of their marketing strategies byfocusing on the different types of incentives/motivations used by consumers in differentsegments. For those consumers who exhibit an increasing (i.e. reinforcing) loyaltypattern, marketing efforts should focus on the attractiveness of the brand viahighlighting the quality features or offering various promotional programs, such as thepopular loyalty programs. In contrast, when targeting at consumers who show a

decreasing pattern of loyalty, marketing managers should be more cautious with theoveruse of promotions since these activities may adversely affect loyalty and, in the longrun, result in significantly lower prices that these consumers are willing to pay.

This study is subject to the following limitations. First, it does not capture consumersvariety-seeking or inertia tendencies (e.g. using a dummy variable for last-purchaseindicator as suggested by past research; Lattin, 1987; Chintagunta, 1998, 1999).Nevertheless, the brand-specific constants in the model can actually capture (part of)the inertia tendencies in repeat purchase cases and variety-seeking tendencies in switching

Capturingconsumer

heterogeneity

659


12/14

behavior cases. Next, this study does not account for the effect of the scale parameter,which has an important impact on the recovery of (unbiased) parameter estimates(Swait and Louviere, 1993; Yoon and Kwak, 2009). Accordingly, future extensions may:

. include a specific variable that captures consumers inertia and variety-seekingtendencies; and

. incorporate the scale parameter in the model to capture the heterogeneity in thescale of parameters and improve the empirical estimates.

Finally, the effects of marketing variables on brand choice decisions which are capturedvia the bparameters in our proposed model are assumed to be time-invariant in thisstudy. In practice, these effects may change over time due to a variety of reasons, such assudden changes in consumers taste and disposable income[1]. For instance, Kim et al.(2005) explore the choicedynamics with time-varying parameters by modeling thedynamicevolution of parameters under a random utility framework. Accordingly, future researchmay investigate how these effects change along with consumers loyalty behaviors over

time. We hope that our study will spark further work in this important research area.

Note

1. We appreciate the reviewer for suggesting this issue.

References

Ailawadi, K.L., Neslin, S.A. and Gedenk, K. (2001), Pursuing the value-conscious consumer: storebrands versus national brand promotions, Journal of Marketing, Vol. 65, January, pp. 71-89.

Banasiewicz, A. (2005), Loyalty program planning and analytics, Journal of ConsumerMarketing, Vol. 22 No. 6, pp. 332-9.

Bawa, K. (1990), Modeling inertia and variety seeking tendencies in brand choice behavior,Marketing Science, Vol. 9 No. 3, pp. 263-78.

Ben-Akiva, M. and Lerman, S.R. (1985), Discrete Choice Analysis: Theory and Application toTravel Demand, The MIT Press, Cambridge, MA.

Blumen, I., Kogan, M. and McCarthy, P.J. (1955), The Industrial Mobility of Labor as a ProbabilityProcess, Cornell Studies of Industrial and Labor Relations, Vol. 6, Cornell University Press,Ithaca, NY.

Brown, R.G. (1974), Sales response to promotions and advertising, Journal of AdvertisingResearch, Vol. 14 No. 4, pp. 33-9.

Bucklin, R.E. and Gupta, S. (1992), Brand choice, purchase incidence, and segmentation:an integrated approach, Journal of Marketing Research, Vol. 29, May, pp. 201-15.

Chintagunta, P.K. (1998), Inertia and variety seeking in a model of brand-purchase timing,

Marketing Science, Vol. 17 No. 3, pp. 253-70.

Chintagunta, P.K. (1999), Variety seeking, purchase timing, and the Lightning Bolt brandchoice model, Management Science, Vol. 45 No. 4, pp. 486-98.

Colombo, R.A. and Morrison, D.G. (1989), A brand switching model with implications formarketing strategies, Marketing Science, Vol. 8, Winter, pp. 89-99.

Deshpande, R. and Hoyer, W.D. (1983), Consumer decision making strategies, cognitive effortand perceived risk, in Murphy, P.E. (Ed.), AMA Educators Proceedings, AmericanMarketing Association, Chicago, IL, pp. 88-91.

MRR34,6

660


13/14

Deshpande, R., Hoyer, W.D. and Jeffries, S. (1982), Low Involvement Decision processes:the importance of choice tactics, in Bush, R.F. and Hunt, S.D. (Eds), Marketing Theory:

Philosophy of Science Perspectives, American Marketing Association, Chicago, IL, pp. 155-8.

Dillon, W.R. and Gupta, S. (1996), A segment-level model of category volume and brand choice,

Marketing Science, Vol. 15, Winter, pp. 38-59.Fader, S.P. and Lattin, J.M. (1993), Accounting for heterogeneity and nonstationarity in a

cross-sectional model of consumer purchase behavior, Marketing Science, Vol. 12 No. 3,pp. 304-17.

Givon,M. (1984),Variety seekingthrough brand switching,Marketing Science,Vol.3No.1,pp.1-22.

Goodman, L.A. (1961), Statistical methods for the Mover-Stayer model, Journal of the AmericanStatistical Association, Vol. 56, pp. 841-68.

Grover, R. and Srinivasan, V. (1992), Evaluating the multiple effects of retail promotions onbrand loyal and brand switching segments, Journal of Marketing Research, Vol. 29,February, pp. 76-89.

Guadagni, P. and Little, J. (1983), A logit model of brand choice calibrated on scanner data,Marketing Science, Vol. 2, Summer, pp. 203-38.

Gupta, S. and Chintagunta, P.K. (1994), On using demographic variables to determine segmentmembership in logit mixture models, Journal Marketing Research, Vol. 31, February,pp. 128-236.

Heilman, G.M., Bowman, D. and Wright, G.P. (2000), The evolution of brand preference andchoice behaviors of consumers new to a market, Journal Marketing Research, Vol. 37,May, pp. 139-55.

Hoyer, W.D. and MacInnis, D.J. (2007), Consumer Behavior, 4thed., Houghton Mifflin, NewYork, NY.

Jacoby, J. and Chestnut, R.W. (1978), Brand Loyalty: Measurement and Management, Wiley,New York, NY.

Jacoby, J. and Kyner, D.B. (1973), Brand loyalty vs repeat purchasing behavior, Journal ofMarketing Research, Vol. 10, Feburary, pp. 1-9.

Jeuland, A.P. (1979), Brand choice inertia as one aspect of the notion of brand loyalty,Management Science, Vol. 25 No. 7, pp. 671-82.

Kahn, B.E. and Raju, J.S. (1991), Effects of price promotions on variety-seeking andreinforcement behavior, Marketing Science, Vol. 10 No. 4, pp. 316-37.

Kamakura, W.A. and Russell, G.J. (1989), A probabilistic choice model for market segmentationand elasticity structure, Journal Marketing Research, Vol. 26, November, pp. 379-90.

Kim, J.G., Menzefricke, U. and Feinberg, F.M. (2005), Modeling parametric evolution in a randomutility framework, Journal of Business and Economic Statistics, Vol. 23 No. 3, pp. 282-94.

Krishnamurthi, L. and Papatla, P. (2002), A probit model of choice dynamics, MarketingScience, Vol. 11 No. 2, pp. 189-206.

Krishnamurthi, L. and Raj, S.P. (1988), A model of brand choice and purchase quantity price

sensitivities, Marketing Science, Vol. 7 No. 1, pp. 1-20.Krishnamurthi, L. and Raj, S.P. (1991), An empirical analysis of the relationship between brand

loyalty and consumer price elasticity, Marketing Science, Vol. 10, Spring, pp. 172-83.

Lattin, J.M. (1987), A model of balanced choice behavior, Marketing Science, Vol. 6 No. 1, pp. 48-65.

McAlister, L. and Pessemier, E. (1982), Variety seeking behavior: an interdisciplinary review,Journal of Consumer Research, Vol. 9 No. 3, pp. 311-22.

Schneider, L.G. and Currim, I.S. (1991), Consumer purchase behaviors associated with active andpassive deal-proneness, International Journal of Research in Marketing, Vol. 8, pp. 205-22.

Capturingconsumer

heterogeneity

661


14/14

Seetharaman, P.B. and Chintagunta, P.K. (1998), A model of inertia and variety-seeking withmarketing variables, International Journal of Research in Marketing, Vol. 15 No. 1, pp. 1-17.

Swait, J. and Louviere, J. (1993), The role of the scale parameter in the estimation and comparison ofmultinomial logit models, Journal of Marketing Research, Vol. 30, August, pp. 305-14.

Yim, C.K. and Kannan, P.K. (1999), Consumer behavioral loyalty: a segmentation model andanalysis, Journal of Business Research, Vol. 44, pp. 75-92.

Yoon, K. (2008), Identifying the truly and the deal-prone hard-core loyals, working paper, LongIsland University, New York, NY.

Yoon, K. and Kwak, K. (2009), Role of purchase-occasion heterogeneity in the implementationof acquisition and retention programs, working paper, Long Island University,New York, NY.

Further reading

Einhorn, H.J. and Hogarth, R.M. (1981), Behavioral decision theory: processes of judgment andchoice, Annual Review of Psychology, Vol. 32, pp. 53-8.

Hoyer, W.D. (1984), An examination of consumer decision making for a common repeatpurchase product, Journal of Consumer Research, Vol. 11, pp. 822-9.

McCann, J.M. (1974), Market segment response to the marketing decision variables, Journal ofMarketing Research, Vol. 11, November, pp. 399-412.

Massy, W.F. and Frank, R.E. (1965), Short term price and dealing effects in selected marketsegments, Journal of Marketing Research, Vol. 2, May, pp. 171-85.

Sivakumar, K. (1995), The role of quality-tier loyalty on consumer price sensitivity for frequentlypurchased products, Journal of Marketing Theory and Practice, Vol. 3 No. 4, pp. 84-96.

Starr, M.K. and Rubinson, J.R. (1978), A loyalty group segmentation model for brand purchasingsimulation, Journal of Marketing Research, Vol. 15, August, pp. 378-83.

van Trijp, H.C.M., Hoyer, W.D. and Inman, J.J. (1996), Why switch? Product category-levelexplanations for true variety-seeking behavior, Journal of Marketing Research, Vol. 33No. 3, pp. 281-92.

Webster, F.E. Jr (1965), The Deal-Prone consumer, Journal of Marketing Research, Vol. 2, May,pp. 186-9.

About the authorsKanghyun Yoon is currently an Assistant Professor of Marketing at Long Island University,C.W. Post Campus. He earned his BS and MS degrees from Hankuk University of Foreign Studies(Seoul, Korea), his MBA degree from Illinois Institute Technology (Chicago, USA), and his PhDdegree from University of Wisconsin (Milwaukee, USA). His main research interests are inmodeling of consumer decision makings, brand loyalty, customer satisfaction, online auctionsand shopping value. Kanghyun Yoon has published in journals such as Information Systems

Research among others. Kanghyun Yoon is the corresponding author and can be contacted at:

[email protected] V. Tran is an Assistant Professor of Marketing at the University of Central Oklahoma.

He earned his PhD in Marketing at the University of Central Florida. His research focuses onconsumer decision-making models, distribution channels and emerging online mechanisms.

To purchase reprints of this article please e-mail: [email protected] visit our web site for further details: www.emeraldinsight.com/reprints

MRR34,6

662

capturing consumer

Documents