1 Introduction

One of the most pursued strategies for fostering the successful market entry of a new product is to market it under the umbrella of an existing successful brand. Studies on the success of this strategy have found evidence that parent-brand strength and the fit between parent brand and transfer product are the main and most influential factors driving brand extension success (Aaker and Keller 1990; Hem et al. 2003; Völckner and Sattler 2006; Erdem and Chang 2012). However, the reciprocal effect of brand extensions, e.g., use of the extended product affecting a quality perception of the parent brand, remains under-researched, despite some widely cited studies (e.g., Roedder-John et al. 1998). Instead, brand loyalty is regarded as a consequence of the underlying assumption that customers transfer their quality perceptions, brand knowledge, and experience with the brand from one category to the other (Erdem and Swait 1998).

Most existing studies are focused on explaining factors which determine the success or failure of a brand extension using a logit or similar type of choice model (e.g., Swaminathan et al. 2001). Here we first aim to confirm that consumers who are loyal to the brand in the leading (parent) product category are more likely than other consumers to be loyal to that same brand in another (extension) category. From the theoretical side, signaling theory can contribute to explaining the formation and effects of this cross-category brand loyalty. Second, using the signaling approach and an alternative way of analyzing the relations in an umbrella brand context, the research should provide some insights into the direction of cross-category loyalty for brands operating in multiple product categories. We want to determine whether loyal consumers from one category tend to be loyal in other categories as well. Third, we contribute by developing a weighted measure to quantify the overall loyalty relations of any product under the umbrella brand with the categories against which the brand competes. Our proposed cross-category loyalty measure is able to quantify the role and strength of each product with respect to its integration within the umbrella brand’s product assortment in terms of brand loyalty leverage.

The article is structured as follows: we start with a brief overview of the conceptual and theoretical background of umbrella branding, followed by a derivation of our research hypotheses. The subsequent section focuses on the measurement of brand loyalty and introduces the share of category requirements approach as a basis for investigating individual brand loyalty. In an empirical study using purchase data from a household panel, we then investigate the existence of cross-category brand loyalty and discuss the associated effects in the context of a major national non-food brand. Using our cross category measure the strength and direction of the signaling effects of the underlying brands are estimated. We conclude with a summary and managerial implications as well as some limitations and ideas for further research.

2 Theoretical background

Especially in the area of consumer goods, brand manufacturers are leveraging their strong brands by cross-promoting and cross-selling different product categories under an umbrella (Kumar et al. 2008). Umbrella branding can be regarded as a form of economies of scope, as it economizes on the costs of creating a new brand (Cabral 2007). It owes its success to the fact that consumers make inferences from brand characteristics—most importantly, the quality of an established product—to the characteristics of others under the umbrella brand (Hakenes and Peitz 2004). The umbrella brand helps consumers make decisions about new products when quality information is missing or not fully employed. The transfer of parent-brand associations to the extension has been identified as one of the most important determinants of the brand extension’s success (Aaker and Keller 1990; Reddy et al. 1994). Research suggests that consumer evaluations of the parent brand have a signaling function for the brand-quality expectations of the extension (Aaker and Keller 1990; Loken and John 1993).

2.1 Umbrella branding and signaling theory

The managerial relevance of signaling by umbrella branding lies in the transfer of consumer quality perceptions across several product categories. Firms offering products in various categories can use the brand name of an established and successful product for a new one, assuming that they transfer a good reputation (Rasmusen 2016). The brand name then acts as quality cue informing consumers about the new product via their experience with the parent product (Wernerfelt 1988; Erdem and Swait 1998; Erdem et al. 2006). Moreover, Erdem and Sun (2002) provide evidence for the existence of marketing-mix spillover effects for umbrella-branded products, which enhance the effectiveness of marketing-mix activities. Sullivan (1990) was the first to present non-experimental evidence for spillovers in umbrella-branded products. Consumers even remain loyal to multi-product firms that fail to offer products matching their preferences better than those from competing firms (Anand and Shachar 2004).

The effect of signaling in umbrella branding is built on the premise of existing information uncertainty and consumers’ belief that the extension of a high-quality brand is likely to be of high quality as well (Wernerfelt 1988; Miklos-Thal 2012). Empirical work in the marketing literature demonstrates that the signaling argument of umbrella branding is broadly consistent with empirical data (e.g., Reddy et al. 1994; Erdem 1998; Balachander and Ghose 2003). Erdem et al. (2006) provide evidence that a need to transfer quality perceptions arises from uncertainty about the true product quality because of asymmetric and imperfect information. This uncertainty may persist even after product use, as some product attributes may not be fully revealed (Erdem and Swait 1998) or in cross country contexts (Erdem and Chang 2012) and online commerce (Mavlanova et al. 2012). Assuming that consumers dislike uncertainty, uncertainty about product quality may induce perceived risk in that consumers have to bear the risk of getting a low-quality product. As consumers tend to be risk averse in most contexts (Rao and Bergen 1992) and strong brands in general are associated with higher perceived quality, strong brands reduce perceived risk by becoming symbols of product quality (Montgomery and Wernerfelt 1992; Erdem et al. 2006).

Signaling theory also proposes that cross-category brand loyalty is a consequence of the perceived brand equity, defined as the expected added value a brand gives a product (Erdem and Swait 1998). Consumers offer their loyalty with the understanding that the new product will provide them utility via consistent product performance (Keller 1998). Any product under the same umbrella brand is associated with high perceived quality, whereby the perceived risk assigned to the product is decreasing. Hence, the expected utility increases and motivates consumers to buy the same brand repeatedly, as long as the new product confirms the quality expectations.

2.2 General research hypotheses

There is a theory explaining a phenomenon whose existence has not yet been empirically confirmed in the context of an umbrella brand’s complete product assortment. We examine cross-category brand loyalty in an empirical study with purchase data in a non-food FMCG market. Our first contribution therein is quantifying the cross-category brand loyalty relations between the products under the umbrella brand, thus offering empirical evidence for the theoretical argumentation of the underlying psychological process described above.

Following the signaling theory approach we derive hypotheses 1–3:

H1:

Consumers who are loyal to the brand in the parent product category are more likely than other consumers to also be loyal to that same brand in any extension product category.

H2:

Consumers who are loyal to the brand in an extension product category are more likely than other consumers to also be loyal to that same brand in the parent product category.

Parent brand experience and parent brand conviction have been identified as drivers of brand extension success (Völckner and Sattler 2006). In addition, following the argumentation of signaling theory, the signaling effect of the umbrella-branded product in the parent product category is highest. The core competence product is decisively responsible for the brand’s equity and therefore for the pure existence of the brand’s extension potential. On account of this, we hypothesize the following

H3:

The probability of being loyal to the brand in any extension product category, given loyalty to the brand in the parent product category, is higher than the probability of being loyal to the brand in the parent product category, given loyalty to the brand in any extension product category.

For a second research objective, investigating the size and direction of impacts we state hypotheses 4–6:

H4:

The parent product category has a higher signaling role within the umbrella brand’s product assortment than any of the extension products under the umbrella brand.

H5:

The overall reciprocal signaling effect (feedback effect) is highest on the parent product category.

The comparison of a branded product’s overall signaling effect on all other products under the same brand and the overall impact it receives, in terms of reciprocal signaling effects, from all other products under the same umbrella brand yields a net signaling balance. In line with H3, but focusing on the direction of the effects we propose the following hypothesis:

H6:

The parent product category has a positive net signaling balance.

3 Measuring brand and cross-category loyalty

The brand loyalty concept has been of enduring concern to both marketing practitioners and academics alike. Loyalty comes in many forms: contractual loyalty, transactional loyalty, functional loyalty, and emotional loyalty. The measures can be classified (amongst others) into proportion-of-purchase, sequence-of-purchase, and probability-of-purchase measures (Jacoby and Chestnut 1978). Mellens et al. (1996) also distinguish between behavioral and attitudinal measures as well as individual-oriented vs. brand-oriented measures. Most often, brand loyalty—neglecting its attitudinal component—is measured by the past purchasing patterns of customers (Chaudhuri and Holbrook 2001). These behavioral measures are easier and less costly to collect than attitudinal data (particularly relevant when studying the evolution of brand loyalty over an extended period of time). Several publications have introduced (Colombo and Morrison 1989) or investigated (e.g., Dekimpe et al. 1997) possible measures for behavioral loyalty, some of which are discussed below.

A well-established measure for cross-category loyalty is the share of category requirements. It has long been used as a metric of brand loyalty in the context of consumer packaged goods (Fader and Schmittlein 1993). According to Rundle-Thiele and Mackay (2001), the measure is strongly associated with the attitudinal brand preference, thus somehow combining attitudinal and behavioral aspects of brand loyalty.

The share of category requirements (SCR) captures the relative share of category purchases that individual households give to each brand they buy (Fader and Schmittlein 1993; Bhattacharya et al. 1996; Jung et al. 2009), defined to be each brand’s market share. Although the SCR measure is generally reported at an aggregate level, several studies employ it on an individual level (e.g., Du et al. 2007). It is defined by

$$SCR_{hicT} \; = \;\frac{{\mathop \sum \nolimits_{t\, \in \,T} q_{hict} }}{{\mathop \sum \nolimits_{k} \mathop \sum \nolimits_{t\, \in \,T} q_{hkct} }}$$
(1)

where SCR hicT is household h’s share of category requirements for brand i in category c during time period T, q hict is the quantity of brand i purchased in category c by household h on purchase occasion t (where t is an index of all purchase occasions during time period T), and k is an index for all brands in the category.

As we use the SCR as individual-oriented, behavioral, proportion-of-purchase measure, consumers can be classified as first choice buyers (FCB), second choice buyers (SCB), or competitive choice buyers (CCB) with respect to a specific brand within a product category. First choice buyers are those who bought the particular brand the most in terms of the amount purchased in that category. In case of two brands with equal purchase amounts, the monetary value spent is of relevance. Second choice buyers are those who purchased the brand within a certain time period but did not assign their highest preference to that brand in terms of the total amount purchased. For them, the brand investigated is just another choice alongside some other much-preferred brand. Competitive choice buyers are those who did not purchase this particular brand during the investigated time period at all; instead, they chose one or more competitive brands in that product category.

The share consumers assign to a particular brand is one important aspect in the context of brand loyalty measurement. A second issue, though, is accounting for different shopping types in terms of buying rates (e.g., heavy, average, and light buyers). Only the combination of category purchase frequencies and share of category requirements can bring out important insights regarding consumers’ brand-loyal purchase behavior and arising customer potential. In our succeeding empirical study, we suggest a median split of households into frequent and seldom buyers, according to category purchase frequencies.

4 Estimating cross-category brand loyalty leverage

In our empirical study the key interest is the relation between different forms of brand loyalty in related transfer categories. We do not investigate success factors of brand extensions in general (like, e.g., Völckner and Sattler 2006). We do also not investigate the impact of the product attributes in the choice relation using the MNL approach (Singh et al. 2005) or the MMP approach (Erdem and Chang 2012). We are looking at the probability of brand choices at the individual level, focusing on the degree of brand loyalty. This has been widely neglected in the empirical analysis of the determinants of successful brand extensions. We do highlight the existence of cross-category brand loyalty, which must be taken into account when considering extending the brand or evaluating the success of a brand extension.

Using the measure of the brand’s share of category requirements, we assign each panel household to the first, second, or competitive choice buyer segment for each product category separately. We propose that brand-loyal (first choice) buyers of a brand exhibit a higher probability of also being brand loyal to that same brand in another category. Taking the behavior of the second or competitive choice buyers as baseline, it is possible to calculate the differences in conditional probability of being a first choice buyer in any other product category, given being a first choice buyer in the product category under investigation. Significant differences are weighted depending on the probability level yielded and are summed up for all categories in which the brand competes.

4.1 The data set

Provided by GfK SE’s household panel, our data set consists of 20,000 representative panel households in Germany and includes two years of self-reported FMCG purchases by the household leader. To account for panel membership duration, the data are weighted with a continuous mass weight.Footnote 1 The data cover 1479 different brand names in 28 product groups. The sample studied measures all purchases in the categories of one major national brand that competes in the FMCG non-food sector. This brand’s core competence is in the area of body care. For a long time, the brand competed in only one category, but then it was extended as an umbrella brand for several other more or less related product categories.

4.2 Measuring aggregated category-specific brand loyalty

In the first step, each household’s number of different shopping days over the two-year examination period was counted. We identified 19,098 panel households with enough information for our investigation. According to the median value of 28 shopping trips, households were then grouped into ‘frequent’ or ‘seldom’ buyers. As this distinction holds true for any of the following analyses, we clustered all purchases into nine product categories (visage, beauty, hair, body, sun, hand, deo, clean, men).

To shed more light on households’ purchase and loyalty behavior within each of the nine product categories and to start investigating households’ loyalty behavior, we calculated the share of category requirements SCR hicT for the brand i for each household h for any category c over the observation period T according to Eq. (1). A household h was then assigned as first choice buyer (FCB), second choice buyer (SCB), or competitive choice buyer (CCB) for brand i in category c according to the following rules:

$$\begin{array}{l} FCB_{ic} \;{\text{if}}\;SCR_{hicT} \; \ne \;0\;{\text{and}}\;SCR_{hicT} \; > \; SCR_{hjcT} \;{\text{for any}}\;j \ne i \hfill \\ SCB_{ic} \;{\text{if}}\;SCR_{hicT} \; \ne \; 0{\text{ and}}\;SCR_{hicT} \; \le \; SCR_{hjcT} \;{\text{for any}}\;j \ne i \hfill \\ CCB_{ic} \;{\text{if}}\;SCR_{hicT} \; = \;0. \hfill \\ \end{array}$$

We left out households not exhibiting any category preference and calculated the shares of first, second, and competitive choice buyers among those households that made category purchases during the two-year observation period (see Table 1). Among the category of buyers, the highest shares of first choice buyers occur in the men (shaving equipment, men’s deodorant) and in the body (body lotion, body gel, after depilatory cream) category. The brand’s extension from body care products to various other categories over the past decades has not (yet) automatically led to high percentages of first choice buyers in all other categories. For example, the very low number of FCB in the beauty category leaves room for speculation as to whether this category is too far away from the brand’s core competence, leaving the brand name unable to attract the brand’s loyal customers in this area.

Table 1 Relative frequencies of frequent buyers with category preference; relative frequencies of seldom buyers are given in parentheses
Full size table

What the results in the Table 1 do not reveal is whether there are first choice buyers within a category that are also first choice buyers in any other category. The results displayed are only category specific and therefore do not allow any conclusions on cross-category brand-loyal behavior to be drawn.

4.3 Measuring aggregated cross-category brand loyalty

Thus far in the examination of brand loyalty, we have not crossed the product category borders. However, the existence of households that are loyal to products of the umbrella brand in multiple categories is a prerequisite for our study. On this account, we analyzed households’ first choice buying behavior across the nine product categories. About 20% of the frequent shoppers (n = 10,473 households with at least 28 shopping trips in the two-year observation period) and 13% of the seldom shoppers (n = 8626 households with 4–27 shopping trips in the two-year observation period) dedicated their largest share in terms of purchase volume to our investigated brand in at least two different product categories.

The about one-third of frequent buyer households made purchases in eight of the nine product categories, while many households made purchases in seven or even all nine categories. One-fifth to one-fourth of these households are first choice buyers in at least two different categories. Both the total number of households that are first choice buyers in at least four product categories and the relative share within the respective buyer segment are comparably low. A quarter of seldom buyer households made purchases in six of the nine product categories, with many households making purchases in five or seven categories. Again, both the total number of households that are first choice buyers in at least four product categories and the relative share within the respective buyer segment are comparably low. These initial results show that about 20% of the panel households exhibit first choice buying behavior in multiple categories. In general, we thus provide evidence that the share of cross-category brand-loyal customers cannot be neglected.

4.4 Measuring disaggregated cross-category brand loyalty

Subsequently, we focused our view on each of the nine product categories and their interrelations. The question arose of whether there are product categories in which customers exhibit a significantly higher share of loyalty to the umbrella branded product if they are also loyal to the umbrella brand in some other product category.

We examined our initially proposed research hypotheses H1, H2 and H3 by (1) investigating the occurrence and relevance of cross-category brand loyalty and (2) determining a household’s probability to be brand loyal—i.e., to be a first choice buyer—in a product category, given brand loyalty in another category. It was also of interest whether we would find differences among product categories. In this context, we specify the following hypotheses:

If a household is a first choice buyer of the brand in the parent category, its probability of also being a first choice buyer of the brand in any extension category is higher compared to a household that

H1a:

Is a competitive choice buyer of the brand in the parent category.

H1b:

Is a second choice buyer of the brand in the parent category.

If a household is a first choice buyer of the brand in any extension category, its probability of also being a first choice buyer of the brand in the parent category is higher compared to a household that

H2a:

Is a competitive choice buyer of the brand in the extension category.

H2b:

Is a second choice buyer of the brand in the extension category.

The probability of being a first choice buyer of the brand in any extension category, given being a first choice buyer of the brand in the parent category, is higher

H3a:

For the case of the comparison of first and competitive choice buyers of the brand.

H3b:

For the case of the comparison of first and second choice buyers of the brand.

We start with a discussion of cross-tabulation of segment membership (FCB, SCB, or CCB) conditioned probabilities for any possible combination of two categories based on Table 2. To capture the cross-category impact of brand loyalty, we must calculate and compare probability values. The conditional probability of being a first choice buying household in category c, given being a first choice buying household in category c ∗, must be related to a reference conditional probability value (‘baseline’ value), e.g., the conditional probability of being a first choice buying household in category c, given being a competitive buying household in category c ∗. If the difference between those two values is very small or not significant, the loyalty behavior in category c is independent of the loyalty behavior in category c ∗, regardless of the absolute value of the conditional probability. The differences between the two respective conditional probabilities are displayed in Table 3.

Table 2 Conditional probabilities of first choice buying behavior of frequent buyers
Full size table
Table 3 Differences between the conditional probabilities of first choice buyers and competitive choice buyers [Δ(FCB − CCB)]; differences between first choice buyers and second choice buyers given in parentheses [Δ(FCB − SCB)]
Full size table

The differences were tested on their statistical significance under the null hypothesis that the shares of first choice buyers in the respective category are equal, i.e., that there is no difference between the conditional probabilities. The t test assesses whether the means of two groups are statistically different from one another. Although we do not test the difference between means but rather the difference between (conditional) probabilities here, the t-test is still appropriate because the (conditional) probabilities are the share of first choice buyers (see Simonson and Tversky (1992) for a similar approach). First choice buyers are coded as ‘1’; all others are coded as ‘0’. Hence, calculating the mean of this variable returns the share of first choice buyers.

A group test statistic for the equality of conditional probabilities is reported for equal and unequal variances. Therefore, before deciding which test would be appropriate, a test for equality of variances was conducted (α = 0.05) for each of the cases above. Depending on the results of these tests, the adequate t-test statistic was used, i.e., either the one for equal variances or the one for unequal variances. Table 3 displays the significant absolute differences in conditional probabilities. These differences are a valuable measure for quantifying the relation between two categories. As the absolute differences may differ from both directions, the matrices are asymmetric.

According to Table 3, hypothesis H1a [line Body (f)] holds true for frequent buyers in all cases but one: in the beauty category is there no difference in conditional probabilities. The picture does not change notably when investigating the seldom shoppers [line body(s)]. However, the differences are smaller and we find one difference (hand category) that lacks significance. The brand’s parent category of body care products underlines its important position. The first choice buyers in the body category exhibit a significantly higher probability than competitive choice buyers in the body category of also being a first choice buyer in any of the extension categories.

H1a:

Cannot be rejected for frequent shoppers in all but the beauty category.

H1a:

Cannot be rejected for seldom shoppers in all but the beauty and hand categories.

The results displayed in the body column of Table 3 also offer empirical evidence for hypothesis H2a. Both frequent and seldom shoppers exhibit a significantly higher probability than competitive choice buyers in any extension category of also being brand loyal in the parent body category, if they are already loyal in the respective extension category. Again, the differences in conditional probabilities are greater for the frequent than the seldom shoppers.

H2a:

Cannot be rejected for frequent shoppers in all extension categories.

H2a:

Cannot be rejected for seldom shoppers in all extension categories.

Before examining the differences between first and second choice buyers, we shortly address the other results displayed in Table 3. The insignificant results for frequent buyers all occur when the beauty or hand product categories are involved. Taking the beauty category as the basis, the changes in conditional probabilities for the sun and hand categories are not significant, while when taking the hand category as the basis, the changes for the beauty and deo categories are not significant. However, the conditional probabilities for being a first choice buyer in the beauty category either exhibit significant but small changes when comparing competitive and first choice buyers in the basis category or do not change significantly at all. A similar picture is revealed for the conditional probabilities in the hand product category. The exceptional positions of the beauty and hand categories may be due to the relatively small number of first choice buyer households in those categories (n = 53 for beauty, and n = 292 for hand). Moreover, the hand category additionally suffers from a high share of households that do not buy at all in the category.

Similar to the comparison of FCB and CCB, the comparison of FCB and SCB reveals that frequent buyers who are first choice buyers in the parent body category exhibit a significantly higher probability than second choice buyers of also being first choice buyers in any extension category (H1b). As with the beauty category, the men category also lacks a difference in conditional probabilities. For the seldom shoppers, we only find three extension categories with significant differences: visage, clean, and men. In the clean category the difference is even larger in value than for frequent shoppers, and in the men category the significance of the difference is appearing.

H1b:

Cannot be rejected for frequent shoppers in all but the beauty and men categories.

H1b:

Cannot be rejected for seldom shoppers in the visage, clean, and men categories.

The results displayed in the body column of Table 3 provide empirical evidence for hypothesis H2b. Frequent shoppers exhibit a significantly higher probability of also being brand loyal in the parent body category if they are already loyal in any extension category, compared to second choice buyers in the respective extension category. For seldom shoppers, we do not find significant differences in the beauty, hand, deo, or men categories.

H2b:

Cannot be rejected for frequent shoppers in all extension categories.

H2b:

Cannot be rejected for seldom shoppers in the visage, hair, sun, and clean categories.

Following the argumentation of signaling theory, we hypothesized that the matrix of conditional probabilities is asymmetric in that the probability of being loyal to the brand in any extension product category, given loyalty to the brand in the parent product category, is higher than vice versa. If we compare the values in the body category line with those in the body category column of Table 3 for the difference between first and competitive choice buyers, and based on the parenthetical values for first and second choice buyers, we must reject H3 for any case.

H3a:

Must be rejected for frequent and seldom shoppers in all extension categories.

H3b:

Must be rejected for frequent and seldom shoppers in all extension categories.

The question that now arises is whether this positive spillover effect is only true for the bilateral relation between the parent category and any one extension category or it also appears within the complete product assortment. To address this issue, in the subsequent section we take all the bilateral relations a category can have (in our case, one category has bilateral relations with eight other categories) and generate an overall general measure for the brand’s category-specific power in terms of cross-category loyalty leverage.

5 Quantifying the category-specific brand loyalty leverage force

In this section, we examine our initially proposed research hypotheses H4, H5, and H6 by quantifying the integration of the brand within the umbrella brand’s product assortment. We achieve this by investigating the brand’s ability to leverage brand-loyal customers between product categories. We no longer focus on bilateral, non-causative relations, instead assuming causal, multi-lateral relations between choice behaviors in the categories investigated. Our goal is to derive directions of brand loyalty leverage between product categories (e.g., Balachander and Ghose 2003). However, as correlations do not prove causation, we must first discuss the relation between conditional probabilities and causal inferences.

5.1 Conditioning and causation

A simple form of the frequency interpretation states that the conditional probability of an event A in a finite reference class B is the relative frequency of the actual occurrence of A within B. An individual assessing the conditional probability P(A/B) may perceive different types of relationships between A and B depending on the context (Tversky and Kahneman 1982). If B is perceived to be a cause of A, P(A/B) is viewed as a causal relation, and if A is perceived to be a possible cause of B, P(A/B) is viewed as a diagnostic relation (Falk 1989; Diaz and de la Fuente 2007).

In our case, even though our data cover a period of two years, we did not carry out a dynamic analysis and therefore do not have a temporal order of choice behavior. Based on Einhorn and Hogarth (1998) and Falk (1989), a choice order is not necessary for further analysis. Instead, we compared households’ behavior in two different loyalty segments by balancing the two respective conditional probabilities.

We use the following example to illustrate our rationale based on probabilistic causality (Feller 1968; Krämer and Gigerenzer 2005): Assume that there are 5000 households, 1000 of which are brand loyal in category A. Two hundred of these are also brand loyal in category B, which yields a conditional frequency of 20%. From the 4000 households that are not brand loyal in category A, 400 are also brand loyal in category B. This means that even though they are not brand loyal in category A, they do exhibit brand loyalty in category B, which can be interpreted as category B brand loyalty that is not caused by brand loyalty in category A. Hence, 10% of the 1000 category A brand-loyal households are brand loyal in category B not due to their brand loyalty in category A. However, category A brand loyalty is causal for category B brand loyalty in the remaining 100 cases.

Taking this rationale, we state that a brand’s cross-category loyalty leverage force in category c ∗ comes from two directions: tractive and attractive force. To what extent do first choice buyers in category c ∗ have a larger propensity to also be first choice buyers in category c, in comparison to second or competitive choice buyers in category c ∗ (tractive force of category c ∗)? Likewise, to what extent do first choice buyers in category c have a larger propensity to also be first choice buyers in category c ∗, in comparison to second or competitive choice buyers in category c (attractive force of category c ∗)?

Therefore the following hypotheses are stated:

H4a:

Comparing first and competitive choice buyers of the brand, the body product category has a higher loyalty tractive force than any extension product category under the umbrella brand.

H4b:

Comparing first and second choice buyers of the brand, the body product category has a higher loyalty tractive force than any extension product category under the umbrella brand.

H5a:

Comparing first and competitive choice buyers of the brand, the body product category has a higher loyalty attractive force than any extension product category under the umbrella brand.

H5b:

Comparing first and second choice buyers of the brand, the body product category has a higher loyalty attractive force than any extension product category under the umbrella brand.

H6a:

Comparing first and competitive choice buyers of the brand, the body product category has a positive net loyalty leverage force.

H6b:

Comparing first and second choice buyers of the brand, the body product category has a positive net loyalty leverage force.

5.2 Tractive force

We start by developing a measure of the tractive force by accounting for two different tractive levels: the difference in conditional probabilities between FCB (in the following referred to as group g1) and CCB (in the following referred to as group g3) as well as the difference between FCB and SCB (in the following referred to as group g2).

With the first measure (FCB vs. CCB), we can capture the total cross-category effect, consisting of a brand experience and a brand loyalty effect. For each product category c ∗, the two buyer segments of first and competitive choice buyers are compared in terms of their buying behavior in any other category c. The competitive buyers are not loyal to the brand in category c ∗ and do not purchase the brand in category c ∗ during the two-year observation period, i.e., they neither exhibit brand loyalty nor have any brand experience. On the other hand, the second measure (FCB vs. SCB) disentangles the two effects, only capturing the brand loyalty effect. In this case, the two buyer segments of first and second choice buyers are compared. The second choice buyers do have brand experience, i.e., they make purchases of the brand in category c ∗, but do not assign the largest share in volume to the brand.

The cross-category loyalty leverage measure \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) for the differences in conditional probabilities between first choice buyers (g1) and second (g j  = g2) or competitive choice buyers (g j  = g3) in the product category c ∗ is composed of two components.

$$\begin{array}{l} LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }} = \left( {\frac{1}{C - 1}\mathop \sum \limits_{c = 1, c \ne c*}^{C} d_{{c^{*} c}}^{{g_{1} - g_{j} }} w_{{c^{*} c}}^{{g_{1} - g_{j} }} I_{{c^{*} c}}^{{g_{1} - g_{j} }} } \right)\star \hfill \\ \quad \quad \quad \quad \quad \quad \left( {\frac{1}{C - 1}\mathop \sum \limits_{c = 1, c \ne c*}^{C} I_{{c^{*} c}}^{{g_{1} - g_{j} }} } \right) \hfill \\ \quad \quad \quad \quad \quad \quad j \in \left( { 2, 3} \right) \hfill \\ \end{array}$$
(2)

In the first component, the differences d in conditional probabilities [see Table 3 (FCB − CCB) and (FCB − SCB) for Eq. (2)] are weighted by a factor w and a dummy variable I, indicating the significance of the difference d, and are then summed up over all product categories c ≠ c ∗. This sum is averaged over the (C −1) product categories under examination.

$$d_{{c^{*} c}}^{{g_{1} - g_{j} }} = Pr\left( {g_{1c} |g_{{1c^{*} }} } \right) - Pr\left( {g_{1c} |g_{{jc^{*} }} } \right)$$
(3)
$$I_{{c^{*} c}}^{{g_{1} - g_{j} }} = \;\,1 \;{\text{if}}\; d_{{c^{*} c}}^{{g_{1} - g_{j} }} \,\,{\text{significant}}, 0 \,\,{\text{else}}$$
(4)

The weight w is introduced to capture the level of change in conditional probabilities, i.e., the same difference is evaluated differently dependent on the baseline conditional probability. For example, a rise from 0% to 5%, a rise from 20% to 25%, and a rise from 80% to 85% do all have the same difference of 5%. But do they all have the same value to our cross-category loyalty leverage measure? Given the little relative market shares, we suggest giving more value to changes in the lower regions of conditional probabilities. Comparable to Gossen’s first law of decreasing marginal utility of a good we argue that the higher the baseline conditional probability already is (and thus the larger the share of loyal buyers of the brand among the reference group of second or competitive choice buyers), the lower is the additional gain. In contrast, starting with a very low or even zero share of loyal customers, an increase and thus a move into appearance or perception is valued comparably higher. With this weight factor, we accommodate for the fact that gaining the first percentage point in market share is harder than expanding the market share when already competing in the market. The weight factor w considers \(Pr\left( {g_{1c} |g_{{jc^{*} }} } \right)\), the basis level of conditional probability. The reciprocal of the exponential function accounts for the aimed effect of decreasing weight with increasing basis level of conditional probability.

$$w_{{c^{*} c}}^{{g_{1} - g_{j} }} = \left( {\frac{1}{{{ \exp }{\kern 1pt} (Pr\left( {g_{1c} |g_{{jc^{*} }} } \right)}}} \right)$$
(5)

In the second factor of Eq. (2), the values of the dummy variable, indicating significance of a difference in conditional probabilities on a 5% level, are summed up over all other categories c ≠ c ∗. The sum value represents the number of categories with significant differences in conditional probabilities and is used to account for the cross-category leverage effect. The more categories c ≠ c ∗ with significant differences, the larger the tractive force of the category c ∗.

In the first two columns of Table 4 the results for the different category-specific \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) indices for frequent buyers are displayed. We now concentrate only on frequent buyers, as they are responsible for the main sales in a product category. The brand’s tractive force in category c ∗ comes from brand loyalty (FCB − SCB) or from a total brand effect (FCB − CCB).

Table 4 Cross-category loyalty leverage force \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\), \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) and the cross-category net loyalty leverage force \(LoyL_{{c^{*} }}^{{g_{1} - g_{j} }}\)
Full size table

For frequent buyers, the brand’s highest tractive force occurs in the clean category, where soap, bath additives, and shower gel are combined. Brand-loyal customers in this category have the highest propensity to also be brand loyal in any of the other categories in which the brand competes. Beauty and hair build the mid-range of index values, while hand, deo, body, sun, and men constitute the group of product categories with low values. Visage has a high value in FCB − CCB but a low value in FCB − SCB.

These results are in contrast to our findings presented in Table 2. The \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) index only reaches a medium to small size for the brand’s parent category (body). The cross-category tractive force of the brand in the body category falls off compared to other categories such as clean, beauty, visage, and hair. Although there is a high share of brand-loyal customers in the body category (see Table 1), those customers are obviously less likely to exhibit brand-loyal behavior in any other category, whereas in the beauty category, for example, the almost negligibly low share of brand-loyal customers demonstrate a high propensity to also be brand loyal in other categories. Therefore,

H4a:

Must be rejected for frequent shoppers,

H4b:

Must be rejected for frequent shoppers.

The highest signaling role within the umbrella brand’s product portfolio comes from the umbrella branded product in the clean product category.

5.3 Attractive force

The process and the argumentation of developing a measure for the attractive force of the brand in each category c ∗ takes the equivalent course as for the tractive force in Sect. 5.2. Accordingly, we account for two different attractive levels: the difference in conditional probabilities between FCB and CCB as well as the difference between FCB and SCB. In the first measure (FCB vs. CCB), for each product category c the two buyer segments of first and competitive choice buyers are compared in terms of their first choice buying propensity in the category c ∗. In the second case (FCB vs. SCB), the two buyer segments of first and second choice buyers are compared.

The cross-category loyalty leverage measure \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) for the differences in conditional probabilities between first choice buyers (g1) and second (g j  = g2) or competitive choice buyers (g j  = g3) is composed of three components.

$$\begin{aligned} LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }} &= \;\left( {\frac{1}{C - 1}\mathop \sum \limits_{c = 1,\; c \ne c*}^{C} d_{{cc^{*} }}^{{g_{1} - g_{j} }} w_{{cc^{*} }}^{{g_{1} - g_{j} }} I_{{cc^{*} }}^{{g_{1} - g_{j} }} } \right)\star \hfill \\ &\quad \left( {\frac{1}{C - 1}\mathop \sum \limits_{c = 1, \;c \ne c*}^{C} I_{{cc^{*} }}^{{g_{1} - g_{j} }} } \right) \hfill \\& \quad j \in \left( { 2, 3} \right) \hfill \\ \end{aligned}$$
(6)

The two components of \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) are similar to those of \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\). The essential difference is the direction of examination and calculation. In Eq. (3), the differences between conditional probabilities are calculated between c * and any other category c, with category c ∗ as anchor. In Eq. (7), the differences between conditional probabilities are calculated between c ∗ and any other category c, with any category c being the anchor. The same applies for the weight factor w (Eqs. (5) and (9)) and the indicator variable I (Eqs. (4) and (8)).

$$d_{{cc^{*} }}^{{g_{1} - g_{j} }} = Pr\left( {g_{{1c^{*} }} |g_{1c} } \right) - Pr\left( {g_{{1c^{*} }} |g_{jc} } \right)$$
(7)
$$I_{{cc^{*} }}^{{g_{1} - g_{j} }} = 1 \;{\text{if}}\; d_{{cc^{*} }}^{{g_{1} - g_{j} }}\,\, {\text{significant}}, \;0\;\,{\text{else}}$$
(8)
$$w_{{cc^{*} }}^{{g_{1} - g_{j} }} = \left( {\frac{1}{{{ \exp }(Pr\left( {g_{{1c^{*} }} |g_{jc} } \right)}}} \right)$$
(9)

The basic level of conditional probability \(Pr\left( {g_{{1c^{*} }} |g_{jc} } \right)\), where the probability to be a first choice buying household in the investigated category c ∗ is conditioned on the behavior in category c, is now considered in the weight factor. In the second component of Eq. (6), the values of the dummy variable, indicating significance of a difference in conditional probabilities, are summed up over all other categories c ≠ c ∗. The more categories c ≠ c ∗ with significant differences that exist, the larger the attractive force affecting the category c ∗ will be. In the two middle columns of Table 4 the results for the different category-specific \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) indices for frequent buyers are displayed.

In the frequent buyers case, the highest index values for the FCB − CCB case appear in the body product, visage, and men categories. Clean, hair, and deo build the mid-range of index values, while sun, hand, and beauty constitute the group of product categories with low values. The relatively high difference between the FCB − CCB and the FCB − SCB case for the men product category is surprising. Obviously, the brand’s overall ability in all other categories c together to stimulate first choice buying behavior in the men category gains much of its impact from the difference between competitive and second choice buyers in the respective categories c, whereas when comparing first and second choice buyers in the respective categories c, there is very little attractive force towards brand-loyal behavior in the men category. The lowest attractive force comes from the beauty category. The households’ first choice buying behavior in any other category c is nearly independent of the households’ behavior in the beauty category, i.e., the probability of being a first choice buying household in any category c is about the same for competitive, second, and first choice buyers in the beauty category.

The highest attractive force appears for the parent body category. The high \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) index in the body category indicates that brand-loyal customers in any of the extension categories exhibit a higher propensity to also purchase the brand in the parent category. Purchases in the parent body category are indeed less likely to lead to first choice purchases in an extension category than vice versa. This result highlights the brand’s strength in the parent category. Customers that are loyal to the brand in any of the extension categories are also more likely to be brand loyal in the parent category. Altogether, the results displayed in the two middle columns of Table 4 by the majority support H5 (highest reciprocal signaling effect on body category).

H5a:

Cannot be rejected for frequent shoppers.

H5b:

Cannot be rejected for frequent shoppers.

The overall reciprocal signaling effect (what we call attractive force) is highest on the parent product category of body care.

5.4 Overall cross-category leverage force

In Sects. 5.2 and 5.3 we investigated each category’s tractive force, i.e., its ability to stimulate brand-loyal purchase behavior in any other category in which the brand competes, as well as the attractive force each category develops in all the other categories. The results are now combined by subtracting the \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) and \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) index values. This net effect allows for the assessment of each category with regard to its role and importance within the brand manufacturer’s product range.

$$LoyL_{{c^{*} }}^{{g_{1} - g_{j} }} = LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }} \; - \;\;LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}$$
(10)

A category with a positive \(LoyL_{{c^{*} }}^{{g_{1} - g_{j} }}\) value evolves a stronger tractive force towards the other product categories in comparison to the overall attractive force in the other categories. Accordingly, a negative \(LoyL_{{c^{*} }}^{{g_{1} - g_{j} }}\) value denotes stronger attractive forces. In the last two columns of Table 4, the results for frequent buyers, distinguishing between the FCB − CCB and the FCB − SCB are displayed.

For the frequent buyers, the core competence body category exhibits the highest negative \(LoyL_{{c^{*} }}^{{g_{1} - g_{j} }}\) index values in both cases, meaning that this category is strongly affected by its attractive force towards all the other product categories in which the brand competes. Brand-loyal customers in any other category are more likely than second or competitive choice buyers in those categories to also be brand-loyal customers in this core competence category. The same holds true, though in a diminished manner, for the deo and visage categories. On the other hand, there are categories such as hand, clean, and beauty, whose tractive force towards the other product categories exceeds the attractive force. Brand-loyal customers in these categories are more likely to also be brand-loyal customers in any other category than second or competitive choice buyers are. The results for the men, sun, and hair categories differ between the two cases of FCB − CCB and FCB − SCB customer groups in that the hair and especially the men category are dominated by attractive forces when comparing first and competitive choice buyers, whereas the sun category in this case develops stronger tractive force.

For the comparison of first and second choice buyers, the results are vice versa. Due to the negative net effects for the parent product category displayed in Table 4, H6 cannot be confirmed.

H6a:

Must be rejected for frequent shoppers.

H6b:

Must be rejected for frequent shoppers.

Altogether, we find evidence for stronger and weaker product categories in view of the brand’s ability to leverage brand loyalty to other product categories within the product offering. We can identify product categories with a strong ‘feedback’ role within the brand’s product offering. These categories exhibit a larger attractive force towards other product categories than their tractive force on the other categories. Our main interest category of body care products is the leading category when it comes to attractive force. The fact that the brand’s parent product category does not take the leading role when it comes to pulling other categories in which the brand competes is a surprising result that demands managerial interest.

6 Summary and managerial implications

The purpose of this research was to examine customers’ brand-loyal purchase behavior in the context of a multi-category analysis, which is of special interest to brand manufacturers of brands competing in multiple categories. From the GfK SE German household panel data, we selected a major national non-food brand for our investigation. According to households’ total purchase frequencies, we made a median split with our data. We calculated each household’s share of category requirements for that brand and grouped households into first choice (FCB), second choice (SCB), and competitive choice (CCB) buyers of that brand for nine different product categories.

The lowest shares of category buyers occur in the sun and hand product categories. Considering only category buyers, the categories men and body reveal the highest shares of first choice buyers. Taking the men category as a special case, conditional upon the special target market, the results reflect the brand’s historical development. The basic positioning—‘natural care’—originates from the body product category. The body category is the brand’s core competence category with the highest share of brand-loyal customers. We obtain a clear overall picture for all product categories in which the brand investigated competes. Approximately 20% of the frequent buyer panel households exhibit first choice buying behavior for the brand in at least two different product categories. Thus, in general, we find evidence for cross-category brand loyalty.

Given uncertainty about product quality, signaling theory proposes that consumers believe that the extension of a high-quality brand is likely to be of high quality as well. Taking this as fact for the products under an umbrella brand, our aim was to provide empirical evidence for consumers’ tendency to be cross-category brand loyal. Our accordingly stated propositions hold true for frequent buyers in the most bilateral category relations. The probability of being a first choice frequent buyer in the respective other product category decreases with the decreasing share of category requirements in the core competence product category. Especially in the parent category of body care, both propositions can be verified for frequent buyers, with the exception of two categories: beauty (in both the FCB − CCB and FCB − SCB cases) and men (in the FCB − SCB case).

The brand’s tractive force in the parent body category is lower than in other categories like clean or hair. The fact that the brand’s highest first choice buyers share occurs in the body category does not imply that this loyal customer base also involves brand loyalty in the extension categories. On the contrary, in the body category the brand develops a higher attractive force than in any other category, given the existence of brand-loyal customers in the extension categories. Altogether, we find evidence for medium force going from and comparably high force coming to the parent body category.

Comparing the brand’s \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) and \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) index values, in the clean, beauty, and hand categories the tractive force of the brand is higher in absolute value than its attractive force. Within this group of categories, the clean category occupies an exposed position because its \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) index value is the highest among the categories and its \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) index value is among the highest. In contrast, the categories body, deo, men, and visage have a larger \(LoyL_{{c^{*} . attractive}}^{{g_{1} - g_{j} }}\) index in absolute value than their \(LoyL_{{c^{*} . tractive}}^{{g_{1} - g_{j} }}\) index. The index values for the hair category are quite similar.

Although there is a very small or even non-existent share of first choice buyers in the beauty category, the category has a comparably high tractive force when it comes to stimulating loyal choice behavior in any of the other categories in which the brand competes. A similar, albeit less extreme, picture is drawn in the clean category. These are starting points for the brand’s management, i.e., the increase of the share of first and second choice buyers in these categories should lie in the focus of marketing strategies. Once these shares are increased, there is a positive feedback effect in other product categories.

In the other direction (attractive force), we found that if there is a loyal customer base in any extension category, or if the brand management creates such a loyal customer base by promoting the brand accordingly, the probability of keeping those customers loyal to the brand in the parent body category is further increased. Thus, in general, the loyal customers in the brand extension categories introduced develop a shearing force for the brand in the parent body category. Only for the beauty and the hand categories must we cut back in this respect.

Overall, the brand’s extensions to several, more or less related product categories proved to be successful in terms of leveraging brand-loyal customers back and forth. We found evidence for various relations between the different categories in which the brand is offered. Our results offer reference for the implementation of promotional activities and the allocation of advertising budgets across product categories. Against our expectations, promotional activities in the parent category are not recommended, as there are other extension categories with a higher net tractive force to involve positive spillover effects. Moreover, as we found empirical evidence for significant differences in brand loyalty already between second and competitive choice buyers of the brand in another category, implementing, e.g., free product trials could be a relevant marketing tool for an initial product or brand contact. This is given the assumption that category buyers who do not purchase the brand (CCB) would behave like category buyers who buy the brand as one of several brands in the category (SCB).

7 Limitations and further research

In the second part of our empirical study (Sect. 5), we followed the argumentation of Falk (1989), who states that temporal order has no role in formal probability theory and probabilistic reasoning. We derived causality by balancing the conditional probabilities for brand loyalty in two different loyalty segments. The question here remains, though, of whether the resulting approach for the calculation of the loyalty leverage index really is pure causal reasoning. We are of course aware that this may be a potential target for criticism.

Our results present challenging opportunities for future research. First, our empirical analysis is ex post, i.e., after the brand investigated was extended from the core product category to various related product categories. We can only contribute to the question of, concerning the leverage of brand-loyal customers, whether the umbrella branding strategy has been thus far successful and the matter of relative strength within the brand’s product assortment. However, it would be of enormous interest for the brand management to look ahead and examine further extension potential.

Second, it would be of immense managerial interest to find out about the households’ characteristics. Therefore, we would like to urge further analyses to go beyond purely behavioral customer segmentation and investigate the drivers (e.g., demographics, attitudes, and marketing-mix sensitivities) that may lie behind the purchase behavior exhibited. Who are those cross-category loyal customers that are valuable for any brand extension strategy?

Third, as we only investigated non-food product categories, the results cannot necessarily be generalized to other markets. Further research should also include food categories for comparison; we expect differences due to involvement levels. Moreover, as we focused on only one major national brand, it might be fruitful to extend our model to other brands.

Fourth, we segmented the panel households based on category-specific share of category requirements in first, second, and competitive choice buyers of the brand. Hence, our measure of brand loyalty is based on revealed brand preferences. The integration of an attitudinal component would probably be a more realistic approach to brand-loyal behavior. Furthermore, the use of conditional probabilities as measures of brand loyalty leverage might be too narrowly defined. We hope that our research stimulates further effort in developing more comprehensive measures of cross-category brand loyalty.