Special Session Summary Integrating the Consumer Into Category Assortment Design

Barbara Kahn, University of Pennsylvania
[ to cite ]:
Barbara Kahn (2002) ,"Special Session Summary Integrating the Consumer Into Category Assortment Design", in NA - Advances in Consumer Research Volume 29, eds. Susan M. Broniarczyk and Kent Nakamoto, Valdosta, GA : Association for Consumer Research, Pages: 281-283.

Advances in Consumer Research Volume 29, 2002     Pages 281-283



Barbara Kahn, University of Pennsylvania


The category assortment decision is probably one of the most important decisions that retailers face. Most of the retailing research investigating this question has focused on optimal space allocation issues since shelf space is one of the scarcest resources in a retail environment. Accordingly, previous research has examined how much space to allocate to a category, how much space to allocate to each brand within the category, and where that brand should be placed on the shelf within the category from a profit maximization perspective. In this session we extend that research to examine three additional issues that a retailer should consider as a part of the overall assortment decision. The first paper considers the impact of the category organization (sorting by attribute or by alternative) on consumers’ perceptions of the variety offered in a category. The second paper considers the impact of the intensity of variety seeking in a particular category on consumers’ desire for broad assortment in a category. The third paper allows the retailer, in the process of selecting items to include in the assortment, to look beyond unit sales and consider the assortment of consumers that a particular brand will bring to the category

The first paper (Morales and Kahn) investigates how various different organizational structures of an assortment or menu can change consumers’ notion of how much variety is present and how positively the assortment will be evaluated. Morales and Kahn show that whether options in an assortment are described by attributes (e.g., 7 different types of shirts, 5 different types of pants; create your own outfit) 7 different sauces, 5 different pastas, create your own entree) or by alternatives (e.g., clothing ensembles or the typical Chinese menu), or by a combination of the two formats will affect the perceived variety of the assortment, satisfaction and, ultimately, the decision to return to the store. In addition if features or recommendations are included, consumers may value these as positive (e.g, helpful, enlightening), or negative (e.g., restrictive or manipulative) depending upon how they are presented.

The second paper (Inman, Maier and Vickers) investigates how understanding the types of attributes that consumers are likely to satiate on quickly can affect how much variety should be offered in an assortment. These authors show that the sensory-specific satiety of a food can drive consumers’ need to switch among flavors within a product category. The higher the need to switch the more valued variety in an assortment is likely to be. The results offer implications to research in attribute-level effects and are helpful to retailers trying to manage category assortment vis-a-vis produt attributes. Further, the approach can be used to determine which flavors of a category offer the greatest potential (e.g., for retailer’s private label strategy).

Finally, the third paper (McAlister, George and Chien) suggests that retailers need to look beyond a brand’s unit sales when considering whether to include that brand in a category assortment. The authors develop a logit-based methodology that allows the retailer to characterize the assortment of customers drawn to a particular brand and to also characterize the way the brand’s customer assortment changes in response to managerial action (e.g., promotion). Using this methodology, a retailer can consider the match between a brand’s customer base and the retailer’s strategy when deciding whether to include that brand in the category assortment.



Andrea Morales, University of Pennsylvania

Barbara E. Kahn, University of Pennsylvania

Previous work in consumer behavior and psychology has shown that evaluations are largely dependent on the frame through which individuals view a good. Even when the good itself is held constant, if the frame through which an individual views a good is changed, the perception of the good is changed as well. In this research we examine how consumers’ evaluations of assortments change depending on how they are presented to consumers. More specifically, we examine the differences in the way consumers evaluate the same assortment when it is presented in varying manners by attribute, alternative or a combination of the two.

In thinking about how consumers evaluate assortments several different factors come into play. Prior research suggests that consumers desire assortments that provide a large number of options (variety), allow them to find exactly what they want (customization), and are not too overwhelming or difficult to process (complexity). Ideally retailers should try and offer assortments with all of these qualities, but the reality is that it is difficult to meet all this criteria at once. Assortments that are perceived as providing a lot of options or a great deal of variety often overwhelm or confuse consumers. Similarly, when the assortment is too small or simplistic, consumers feel that they cannot find exactly what they want. As a result, retailers must often make tradeoffs as to which aspects of their assortments to emphasize.

While some retailers may explicitly choose to emphasize certain qualities of their assortment, we believe that others may be highlighting certain aspects unknowingly. In this research, we demonstrate how the manner in which options are presented to consumers emphasizes different qualities of the assortment. When assortments are presented by attributes, where the consumer is responsible for choosing the attributes that form the final product, we hypothesize that the assortment is viewed as providing less total options but increased customization. In addition, for an attribute-based presentation the degree of perceived complexity is highly dependent on expertise in the category. Since consumers are responsible for creating the final product in an attribute presentation, subjects with little expertise in the category will find the choice process difficult and effortful while subjects with expertise will find making a choice fairly simple. On the other hand when assortments are presented by alternative, where the final products are pre-configured groupings of attributes, we hypothesize the reverse effects. The assortment will be perceived as providing more total optins but less customization, and since every combination of attributes is presented, perceived complexity will increase due to the overwhelming number of options consumers must process.

Although the predictions for the attribute and alternative-based presentations are fairly clear-cut, it is less obvious how combinations of these two different presentation styles will be perceived. In various retail settings, combination presentation styles are being used where the assortment is presented by attribute along with a select number of alternatives or only a select number of alternatives are presented but consumers are told that they can modify the alternatives to include any of the different attribute combinations. The question is whether these combination methods of presentation lead to more favorable evaluations than the attribute or alternative-based presentations. We hypothesize that using a combination presentation may result in the following positive and negative effects. Consumers may view the select alternatives as recommendations because of the implied expertise underlying their creation, and as a result may feel more confident in their choice of these select alternatives. The pre-configured alternatives may also help consumers to see new possibilities in the assortmentCthings they would not have seen on their own if presented only by attribute. However, the select alternatives may also lead to the reverse effect where the select alternatives actually limit or restrict the possibilities that consumers see in the assortment. In addition, if consumers view the select alternatives as duplications of the attribute presentation and not as helpful suggestions, they may cause consumers to feel manipulated by and irritated with the retailer.

In a series of several laboratory experiments, we test the above hypotheses and show how the same assortment (i.e., the same options in the potential choice set) is evaluated differently depending on which of the four presentation styles is used. The results show that the different presentation styles do indeed lead to varying levels of perceived variety, customization and complexity, which in turn affects overall satisfaction and store choice. The implications of these findings to store organization and web design are discussed.



J. Jeffrey Inman, University of Pittsburgh

Andrea S. Maier, University of Minnesota

Zata Vickers, University of Minnesota

Retailers and manufacturers struggle to design their assortment of attributes within a particular category. For example, should the available yogurt flavors include lemon, lime, orange, grapefruit and mandarin orange or would lemon, orange, hazelnut and chocolate chip be a better mix? In the absence of established methodologies to address this issue, they largely rely on ad hoc approaches with little consideration of which attributes are more important to consumers in terms of assortment. If consumers desire more assortment on some attributes than others, it is important for retailers to understand dynamic changes in attribute preference and respond accordingly. Improved understanding of the attributes that contribute to switching within a product category will better enable food retailers to design their category assortment as part of their ECR initiative and category management objectives.

We propose that sensory-specific satiety is an attribute of foods that can drive consumers to switch among products within a category. Sensory-specific satiety is a temporary drop in the pleasantness of a food produced by eating that food. This drop in pleasantness is relatie to changes in pleasantness of other foods that have not been recently eaten (e.g., Lyman 1989; Rolls et al. 1981a). For example if one eats a bowl of strawberry yogurt, liking for strawberry yogurt drops but liking for other foods (e.g. chocolate, bananas, bread, chicken, etc.) remains unchanged. This drop in liking has been shown to be primarily dependent on the sensory attributes of a foodBnot its macronutrient content. Importantly, when two foods have similar sensory attributes the liking for one can be diminished by consuming the other (e.g. eating a bowl of strawberry yogurt might also diminish the liking for a strawberry jelly bean). Sensory-specific satiety differs across foods (Johnson and Vickers 1992), and it may partly explain why consumers seek variety on some attributes more intensively than others. Inman (2001) reported that consumers switched more intensively among sensory attributes (flavors) than among non-sensory attributes (brands) in nearly all the categories examined. Other factors such as preference heterogeneity and perceived risk failed to explain that difference in relative switching intensity.

The goal of our research is to determine the attributes of food products that encourage repeat consumption or that drive consumers to switch among them. That is, we seek to determine whether the sensory-specific satiety of a food drives consumers’ need to switch among flavors within a product category. Our objectives are twofold: (a) to compare measures of sensory-specific satiety with measures of consumption and with measures of switching among flavor choices in a product category in both an experimental and a field context and (b) to examine the effects of liking, flavor intensity and sensory similarity among flavors on sensory-specific satiety and on switching among flavors in both experimental and field contexts.

We draw upon the sensory-specific satiety literature to generate four hypotheses. First, we hypothesize that greater levels of sensory-specific satiety will correspond to increased switching away from a flavor and lower overall consumption of that flavor in both experimental and field contexts. Second, we hypothesize that both sensory-specific satiety and switching among flavors are inversely related to the flavors’ similarity. Third, we hypothesize that more intensely flavored products will produce more sensory-specific satiety and thus be less frequently selected from a category assortment. Finally, we hypothesize that sensory-specific satiety and switching among flavors will be moderated by preference. The effects will be stronger for less preferred flavors.

We test our hypotheses in two studies. The first is based on a consumption diary panel and the second utilizes an experiment. In the first dataset, which is already in hand, 850 participants completed a diary of their consumption occasions of ten product categories over a six-week period. The database consists of approximately 6000 eating occasions for potato chips, our focal category, and includes measures of number of times consumed, amount consumed, and switching among flavors and brands of potato chips (Inman 2001). We integrate this data with primary data that we are collecting on the same flavors in the Food Science laboratories at the University of Minnesota (data collection underway). Specifically, we measure flavor intensity, preference, degree of similarity among the flavors, sensory-specific satiety, and flavor choice immediately after consuming each flavor.

The experiment will enable us to test the hypotheses in a more controlled setting and to perform a mediation analysis. Subjects participating in the sensory-specific satiety tests will have previously participated in the liking, intensity and similarity testing session. Each subject will participate in a separate sensory-specific satiety test for each of the flavors they indicated they would be willing to purchase. On arrival at the test site subjects will rate their hunger and rate their liking for small pieces of several different foods. One of these will be the test food (e.g. one flavor from the product category). Next they will be given a serving of the test food and asked to eat the entire aount. Immediately after consuming the test food the subjects will re-taste and re-rate the set of food items. Sensory-specific satiety is measured as the change in liking for the eaten food (test food) compared with the mean change in liking of the uneaten foods in the rating set. After re-tasting and re-rating the foods the subjects will be given an opportunity to choose any of the product category flavors to eat. We will use mediation regression analysis (Baron and Kenny 1986) to examine the impact of sensory-specific satiety on flavor switching. We expect the influence of similarity, intensity, and preference to be mediated by sensory-specific satiety. This laboratory choice will provide the first experimental evidence of flavor-driven switching. Data collection for this study is underway and will be complete by late April. All analyses will be complete well before ACR.



Leigh McAlister, University of Texas, Austin

Ed George, University of Texas, Austin

Alex Chien, University of Texas, Austin

The logit model has been widely used to develop an understanding of the affinity between individuals and choice alternatives. Typical logit models investigate this affinity from the perspective of the individual doing the choosing. These models consider the probability that an individual will choose a particular choice alternative. Considering the affinity relationship from this perspective provides insight into the characteristics of choice alternatives that attract customers. The traditional perspective has been shown to provide a parsimonious model that fits and predicts brands’ market shares. The model’s parameters provide managerial insight and allow the estimation of market response to retailers’ actions. In this work we reverse the traditional perspective and examine the affinity between individuals and choice alternatives from the perspective of the choice alternative. That is, we consider the probability that a choice alternative will be selected by a particular individual. Considering the affinity relationship from this perspective allows us to gain insight into the characteristics of the individuals drawn to a particular brand. We will show that the proposed formulation provides insights that are analogous to those provided by the traditional formulation. We will fit and predict brands’ customer profiles. Our model’s parameters will provide insight into the assortment of customers drawn to a brand and the way that customer assortment changes in response to retailers’ actions like promotions.

In this illustrative example we focus on brands’ customer assortments in a single supermarket over a 7-week period. Our customer descriptors are the number and value of items in the customer’s shopping cart when he/she checks out ("basket size" and "basket value"). We find that the customer assortment for snack brands (e.g., Coke, Pepsi, Dr. Pepper, Snickers) tends to include a disproportionate number of shoppers buying only a few items while the customer assortment for brands that might be used to prepare a meal (e.g., Mazola, Ragu, Kraft Mac and Cheese) tends to include a disproportionate number of shoppers buying an especially large number of items on their shopping trip.

Some brands’ customer assortments change when the brand is promoted. Kraft Mac and Cheese, Mazola and Snickers tend to draw disproportionately more small basket shoppers when on promotion. Clorox, on the other hand, tends to draw disproportionately more large basket shoppers when on promotion.

By examining the shift in a brand’s customer assortmentfrom weekday to weekend, we are able to identify Gold Medal Flour, Kellogg’s Cereals, Minute Maid Orange Juice and Ragu Spaghetti Sauce as "trip generator" brands. These brands are more likely to be bought by small basket shoppers on weekdays than on weekends, perhaps a reflection of at-home stock-out driven "quick trips" on weekdays.

This reformulation of the logit model to explore customer assortment should prove useful to brand managers who need a deep understanding of the target markets for their brands. Similarly, retailers can benefit from understanding the target audiences for different brands as they blend brands into the overall strategy for their stores.