Variety for sale: mass customization or mass confusion?

A common strategy in retailing has been to compete by offering a wide variety of items within a category, designed to appeal to every consumer taste. Consider the growth of "category killers" such as Toys "R" Us, Barnes and Noble (books), or Circuit City (electronics) that feature huge assortments within specific product categories. Large assortment strategies such as these can backfire, however, if the complexity causes information overload such that a customer feels overwhelmed and dissatisfied, or chooses not to make a choice at all (see Jacoby, Speller, and Berning, 1974).

The frustration and information overload may become even more of a concern when retailers look toward customization as way to provide customers with exactly what they want. For example, Choice Seating Gallery, a customized sofa shop, makes the following offer: "Choose from 500 Styles - Choose from 3000 Fabrics - Choose from 350 Leathers." The problem is that each customer ultimately only wants one sofa. To design that ideal sofa, the customer needs to know what the attributes of sofas are, his/her preferences within those attributes, and which attributes are more or less important. Non-experts, or consumers new to the category, may not have that knowledge and thus may have difficulty finding what they want. Accordingly, the huge number of potential options (150,000 different fabric sofas; 17,500 leather sofas) may be confusing and overwhelming rather than beneficial.

If the customers become frustrated or dissatisfied with the complexity, a large variety or customization strategy obviously would not be a competitive advantage: research shows that dissatisfaction with the shopping process is attributed largely to the retailer, which can ultimately impact store traffic and the percentage of customers who make a purchase (Fitzsimons, Greenleaf, and Lehmann, 1997). The key to customer satisfaction with the entire shopping interaction is to ensure that the customer is equipped to handle the variety. To do so, we propose that retailers who offer a large variety of items in each category should ask their non-expert consumers to explicitly indicate within-attribute preferences as a way to help them sort through the variety and figure out which option best fits their needs. Accordingly, category killer retailers such as Circuit City may see fit to provide more information and sales support than general merchandise retailers who carry smaller assortments in any given category.

We postulate that to maximize customer satisfaction with the shopping experience, a retailer needs to control both the way the information is presented and the input the consumer provides in the process of learning about the available attributes and alternatives. We examine two ways of presenting information about the category: (1) attribute-based and (2) alternative-based. In the first method, the customer is asked which level s/he prefers for each attribute of the product or service and then a product is chosen from a large assortment (or a customized product is developed) based on those preferences. In contrast, when information about options is presented by alternative, customers are shown several options and asked to formulate within-attribute preferences by comparing the alternatives. For example, in the customized sofa store, customers are often led through a showroom and salespeople encourage customers to indicate what they like and don't like about the exhibited sofas. Our results show that for high variety assortments, the attribute-based format reduces perceived complexity, increases satisfaction with the process, and facilitates consumers' willingness to make a choice.

We also examine three levels of consumer input to the preference-learning process. The first level is minimal, i.e., consumers are merely shown the product attributes and the various levels of those attributes. In a custom sofa shop, for example, consumers could be shown that they can choose between loveseats and sofas, leather and fabric coverings, and various types of construction. The second level of input is more effortful, i.e., consumers are asked to indicate their favorite level within each attribute. For example, rather than just being shown the various options, consumers could be asked whether they prefer leather or fabric, or which of the different sofa back shapes they prefer. The third level of input is the most demanding. Here, consumers are asked which attributes are most important. Then they are asked to indicate their preferred levels within each of the attributes that they consider important.

The results of two experiments indicate that the optimal level of customer input is the second one, in which consumers are asked to indicate within-attribute preferences. Both too little input (where consumers merely learn what the available attribute levels are) and too much input (where consumers indicate relative importance across various attributes) result in lower levels of customer satisfaction with the shopping experience.

The contribution of this research is to better understand how retailers should expose customers to product information, particularly in a high variety situation. We show that by presenting information in a simplified, manageable way and by asking consumers to explicitly indicate within-attribute preferences, a retailer can increase customer satisfaction with the total shopping experience. This will help ensure that a high variety or customization strategy is indeed an advantage rather than a liability. Others (e.g., Green and Srinivasan, 1990; Huber et al., 1993) have examined the effectiveness of various methods of preference elicitation (e.g., full profile conjoint) on the accuracy of predicting choice, but they have not addressed how these different procedures affect customer satisfaction with the shopping process.

THEORETICAL DEVELOPMENT AND HYPOTHESES

Consumers experience confusion or complexity in high variety categories because there are numerous options to consider (see Malhotra, 1982) and because the task of categorizing items based on desired or disliked attributes is difficult (see Hutchinson and Alba, 1991). The confusion a consumer experiences with a wide assortment of options, however, is due to the perceived complexity, not necessarily the actual complexity or variety. Actual variety may be small, but the consumer may perceive that a large set of options is being offered (e.g., the Chinese menu which seems to offer an enormous number of choices but can actually be reduced to 4 kinds of meat, 4 sauces, etc.). On the other hand, actual variety may indeed be large, but if consumers can selectively attend to only those alternatives that are acceptable based on their goals or preferences (e.g., the consumer who knows she wants 2-inch square white tiles for her new bathroom), then the relevant perceived complexity may be much smaller and more manageable.

Experts, for example, can process more options and are less subject to information overload (e.g., Chase and Simon, 1973), because they know their preferences for different levels within attributes. They can accordingly selectively attend to some alternatives and not others and organize the alternatives based on overall liking (Alba and Hutchinson, 1987; Huffman and Houston, 1993). We therefore propose that perceived complexity is reduced when consumers learn their preferences within product attributes. Learning permits subjects to better understand the environment, reducing complexity. One way that a retailer might facilitate within-attribute preference learning is to manage the way information about the attributes is presented to the consumer.

Information Presentation Format

Two common means of presenting information about products are by attribute and by alternative. When information is presented by attribute, a customer is asked what level s/he prefers within each attribute of the product or service. An example of this is the method used by Dell Computer on the web. After gathering preferences for each attribute (e.g., which size hard drive is preferred, which RAM level is preferred), the customized computer is developed and shown to the consumer for approval prior to ordering.

In contrast, when information about the options is presented by alternative, customers are shown several alternatives and asked to formulate preferences among them. For example, in customized sofa stores or in customized kitchen design shops, customers are often led through showrooms that exhibit alternatives, and salespeople encourage the customers to express their likes and dislikes of the options on display. These preferences can then be used to help select an appropriate alternative or to create a customized one.

Prior research shows that learning within-attribute preferences from alternatives is difficult (see Brehmer, 1980; Hoch and Deighton, 1989; Meyer, 1986, 1987). The reason is that such learning requires consumers to engage in what might be called "intuitive regression" where the consumer must decompose the alternative into its attributes and infer how each contributes to overall evaluative reactions (Meyer, 1987). This reasoning is consistent with research showing that subjects have a difficult time using more than two dimensions to evaluate an alternative (Malhotra, 1982; Siegler, 1981). In addition, when alternatives are at least moderately complex and can be described on many attributes, presentation of information by attribute is easier simply because the information is presented in smaller chunks that are easier for non-expert consumers to process (Chase and Simon, 1973). When the alternatives are at least moderately complex, preference learning is therefore likely to be easier and faster when the information is presented by attribute rather than by alternative. The relative ease of learning preferences in this manner is predicted to increase satisfaction with the sales interaction, which we term satisfaction with the process.

As argued earlier, the perceived complexity of a choice set is reduced when a person can direct attention only to relevant information and thus disregard irrelevant information (Bettman and Park, 1980; Urbany, Dickson, and Wilkie, 1989). Since presentation of information by attribute is predicted to …