An investigation of product search engines in Web-based electronic commerce.

摘要

In business-to-consumer (B2C) electronic commerce, the conversion rate of lookers-to-buyers averages 2%, 2 buyers for every 100 lookers. We believe that this rate is due, in part, to decision aids that are not designed to fit the large search space faced by online shoppers. A decision aid (DA) is a software tool designed to help decision makers, e.g., online shoppers. We investigate whether some sequences of DAs help shoppers more than other sequences.; We develop a shopping model, which combines Effort, Accuracy and the Technology Acceptance Model (TAM) (Davis 1989) to support the shopper with different electronic store (e-store) designs, which are sequences of two or three DAs. The shopper's goal is consistent with the Effort-Accuracy Model (EAM): maximize accuracy and minimizing effort. Our integrated shopping model shows that TAM extends EAM to better predict decision confidence (DC). In order to measure accuracy, we develop metrics that identify choices of (1) products that dominate others solely on the basis of attribute values, and (2) products that outrank others on the basis of attribute values and subjective utility weights. The first metric is objective, whereas the second metric combines objective and subjective elements.; We use a controlled experiment on 116 subjects and treatments that are four different e-store designs. We test a number of hypotheses using the collected data. To test the hypotheses, we use three analytical techniques, ANOVA, Multiple Regression, and Structural Equation Modeling, on two levels of analysis, Macro and Micro. The Macro-Level considers the sequence of DAs as a whole, whereas the Micro-Level considers each DA, in its own stage, separately. The analysis helps us determine the best experimental treatment, i.e., e-store design. There are two key findings: (1) some e-store designs minimize effort, maximize accuracy, or maximize DC significantly more than others, and (2) several TAM-related variables are important predictors of DC.; This research could have direct implications for electronic commerce decision aid designers who are trying to increase revenues. The designers could have their decision aids dynamically detect the current task complexity and either recommend or impose a particular decision aid. The dynamic detection could be tailored to the individual's customer profile or real-time behavior. This research is the initial step towards a General Purpose Shopping Simulator for decision strategy research.