Category Information Transfer: Implications For Consumer Search

Nicholas Lurie, University of North Carolina
EXTENDED ABSTRACT - With over 13,000 mutual funds to choose from, it is easy for visitors to Quicken’s web site to be overwhelmed with information. Unlike brick-and-mortar retailers, virtual sellers face few constraints on the number of products they carry. At the same time, there are very real constraints on consumers’ cognitive capacities and time. To help consumers, information providers often organize information into categories. An important question for Quicken is: Which categorization is better for a given consumer goal? In particular, which categorization is most helpful to consumers in terms of increasing the efficiency of their search and ensuring that they choose the best alternative?
[ to cite ]:
Nicholas Lurie (2003) ,"Category Information Transfer: Implications For Consumer Search", in NA - Advances in Consumer Research Volume 30, eds. Punam Anand Keller and Dennis W. Rook, Valdosta, GA : Association for Consumer Research, Pages: 179.

Advances in Consumer Research Volume 30, 2003     Page 179

CATEGORY INFORMATION TRANSFER: IMPLICATIONS FOR CONSUMER SEARCH

Nicholas Lurie, University of North Carolina

EXTENDED ABSTRACT -

With over 13,000 mutual funds to choose from, it is easy for visitors to Quicken’s web site to be overwhelmed with information. Unlike brick-and-mortar retailers, virtual sellers face few constraints on the number of products they carry. At the same time, there are very real constraints on consumers’ cognitive capacities and time. To help consumers, information providers often organize information into categories. An important question for Quicken is: Which categorization is better for a given consumer goal? In particular, which categorization is most helpful to consumers in terms of increasing the efficiency of their search and ensuring that they choose the best alternative?

One way to think of categorization is as a set of messages about attributes of category members. For example, one way to categorize mutual funds is to organize them into stocks, bonds, and money market funds. The attributes of the funds (e.g., risk, return,) may be thought of as another message set. Given these two message sets, it is possible to determine the information transmission between categories and attributes. For example, the amount of information provided by a particular categorization about funds’ risk level can be calculated. By comparing multiple categorizations in this way, it is possible to determine which category structure transfers the most information about a given set of attributes.

At the same time, category structures that transfer more information may not always be preferable for consumers. Using a particular category structure may involve a tradeoff between search and thinking costs. Alternative category structures have implications for the extent to which search must occur across multiple categories (i.e., the number of categories that must be examined) and the number of alternatives that must be examined within a given category. The extent to which decision quality is improved by minimizing "between" versus "within" costs of category usage may, in turn, depend on the decision goal. If the cost of considering each alternative is low (for example, alternatives are being compared on a single attribute), then broader categories, that provide less information; yet contain more alternatives, may increase the chance of finding the "best" alternative in a given category. On the other hand, if the costs of considering each alternative are high, for example if tradeoffs need to be made between multiple attributes, then one would expect a preference for specific categories that allow one to eliminate alternatives from consideration and thus consider fewer alternatives.

This paper attempts to provide empirical evidence about the relationship between the amount of information provided by alternative category structures, information search and decision making. Three experiments use formal measures of information structure to assess the amount of information that alternative category structures provide. These measures are then used to predict search efficiency (search time and number of alternatives searched), perceived usefulness, attribute importance, and the probability of choosing a dominant alternative.

Experiment 1 used an agent task to examine how alternative categorizations of the same data, which are seen in pretests as "equivalent," and yet transfer different amounts of attribute information according to information measures lead to differences in search efficiency and perceived informativeness. In particular, when information transfer was higher, participants spent less time searching and searched fewer alternatives when information transfer was greater.

Experiment 2 extends these results by varying the amount of information transfer while keeping the correlation between category and attribute levels perfect and manipulating the distribution of alternatives across categories. In experiment 2, participants choose their own criteria for choosing a fund. In experiment 2, information transfer predicted search efficiency, perceived difficulty in using an information environment, perceived time and perceived usefulness of the category structure.

Experiment 3 looked at the tradeoff between information transfer and cognitive load offered by alternative levels of category specificity and decision tasks. Results from this study suggest that measures of information structure, such as those of information theory, can be used to both identify and predict the level of categorization that leads to better decision making as a function of decision processes. Specifically, experiment 3 found support for the hypothesis that when evaluating individual alternatives is cognitively difficult (for example, if tradeoffs are being made between multiple attributes), decision-making will be enhanced by providing decision makers with categories that transfer more information. At the same time, when evaluating individual alternatives is relatively easy (for example, when decisions are being made largely on the basis of a single attribute), decision making can, in fact, be enhanced by providing decision makers with categories that transfer less information. These results suggest that consumers’ decision processes should be accounted for when determining how to structure information environments.

This paper suggests that measures of "information transfer" can be used to organize information for individuals. If one knows the relative importance of different attributes to a given consumer, then one can create custom information structures (i.e., categories) that effectively transfer information on attributes relevant to that consumer and account for the extent to which that consumer is trading off multiple attributes in her choices. In this way, measures of information structure are useful tools for mass-customizing consumers’ information environments. Given that the organization of information in electronic environments can potentially be adapted to decision maker needs in real-time, determining which information structures are optimal for a particular decision maker will prove particularly fruitful.

----------------------------------------