| With the rapid development of e-commerce, the problem of information overloading has arisen. Personalized recommendation systems can help consumers to reduce the degree of cognitive effects, and make shopping decisions, and then promote product sales. This is convenient for both consumers and online retailers, but unsuitable recommendations are annoying and distracting. This paper argues that in the different time, the number of similar products and related products provided by recommendation systems will affect the users to accept the recommended products. In addition, because of the difference in decision-making between utilitarian goods and hedonic goods, so this paper from the perspective of recommendation timing and recommendation structure to study the different effect of recommendation systems between utilitarian goods and hedonic goods.This paper based on the conceptual framework of two-stage decision-making to explore the appropriate proportions of similar products and related products provided by recommendation systems, so that it can be accepted more. We build a simulated shopping site, and invite subjects to do simulated shopping in our shopping scenario. We verify the hypothesis by collecting the shopping cart product data and the purchased product data.The empirical results found that in different stages of decision-making, consumer preferences are different:the first stage recommended products in more likely to be taken into consideration set of consumers than the second stage recommended products, and in the second stage of decision-making, compared to the similar products, related products in more likely to be taken into consideration set. The personalized recommendation structure in different stages of decision-making have a significant impact on the effect of accepting recommendation for utilitarian goods, but not for hedonic goods. The study results not only complement the theoretical study on the recommendation system, but also make recommendations for online retailers to make a more effective recommendation systems. |
