Page All:
Page 1
Improving Internet Recommender Systems
CHICAGO, Aug. 9, 2012 /PRNewswire-USNewswire/ -- Customers find it hard to evaluate before purchase experience products because many of the attributes are yet to be experienced or significant information on non-quantifiable attributes is missing. Examples of such products are movies, music albums, e-books and automobiles.
In a new analysis appearing in the June 2012 issue of the American Marketing Association's Journal of Marketing Research Professors Jaihak Chung and Vithala R. Rao develop a general consumer preference model for experience products. The model overcomes the limitations of consumer choice models, especially, when it is not easy to consider some qualitative attributes of a product or too many attributes for online recommendation services.
They rely on the experience of those who purchased such products and rated them in developing a prediction for future customers of the products. Their three-step approach integrates the non-quantifiable information with consumer preference similarity via the concept of virtual experts. They first identify virtual experts by Bayesian clustering of consumers' preference data and then assign a virtual expert for each customer, using the similarity of the customer to the virtual experts and then model the utility of a product to the customer in terms of product attributes observable before purchase and the matched virtual expert's opinions.
The authors' approach is a combination of well-established collaborative filtering systems used in recommender engines developed in the information systems area and attribute-based choice models quite common in marketing research. It embeds automated word-of-mouth information, which is the primary fuel of the predictive power of collaborative filtering. It combines the standard random utility model with collaborative systems in a complementary way. With their approach, firms can take full advantage of customer databases for predictive improvement of future customer preferences. Their model is flexible enough to use all different types of information, such as product attributes, consumer characteristics, and consumer preference similarity, that are available in e-retailers' customer databases.
According to the authors, "We apply our model to data on movie ratings for online movie recommendation and show that our approach yields superior fits than either the collaborative approach or the attribute-based choice models. The improvement is significant both statistically and managerially."
The primary take-away from this article is a firm's ability to utilize its vast customer data to improve the recommendations made to its customers to increase revenues.
About the American MarketingAssociation:
The American Marketing Association (AMA) is the professional association for individuals and organizations who are leading the practice, teaching, and development of marketing worldwide. Learn more at marketingpower.com.
Contact: Christopher Bartone - 312.542.9029 - cbartone@ama.org
SOURCE American Marketing Association
American Marketing Association
Web Site: http://www.marketingpower.com
Improving Internet Recommender Systems
CHICAGO, Aug. 9, 2012 /PRNewswire-USNewswire/ -- Customers find it hard to evaluate before purchase experience products because many of the attributes are yet to be experienced or significant information on non-quantifiable attributes is missing. Examples of such products are movies, music albums, e-books and automobiles.
In a new analysis appearing in the June 2012 issue of the American Marketing Association's Journal of Marketing Research Professors Jaihak Chung and Vithala R. Rao develop a general consumer preference model for experience products. The model overcomes the limitations of consumer choice models, especially, when it is not easy to consider some qualitative attributes of a product or too many attributes for online recommendation services.
They rely on the experience of those who purchased such products and rated them in developing a prediction for future customers of the products. Their three-step approach integrates the non-quantifiable information with consumer preference similarity via the concept of virtual experts. They first identify virtual experts by Bayesian clustering of consumers' preference data and then assign a virtual expert for each customer, using the similarity of the customer to the virtual experts and then model the utility of a product to the customer in terms of product attributes observable before purchase and the matched virtual expert's opinions.
The authors' approach is a combination of well-established collaborative filtering systems used in recommender engines developed in the information systems area and attribute-based choice models quite common in marketing research. It embeds automated word-of-mouth information, which is the primary fuel of the predictive power of collaborative filtering. It combines the standard random utility model with collaborative systems in a complementary way. With their approach, firms can take full advantage of customer databases for predictive improvement of future customer preferences. Their model is flexible enough to use all different types of information, such as product attributes, consumer characteristics, and consumer preference similarity, that are available in e-retailers' customer databases.
According to the authors, "We apply our model to data on movie ratings for online movie recommendation and show that our approach yields superior fits than either the collaborative approach or the attribute-based choice models. The improvement is significant both statistically and managerially."
The primary take-away from this article is a firm's ability to utilize its vast customer data to improve the recommendations made to its customers to increase revenues.
About the American MarketingAssociation:
The American Marketing Association (AMA) is the professional association for individuals and organizations who are leading the practice, teaching, and development of marketing worldwide. Learn more at marketingpower.com.
Contact: Christopher Bartone - 312.542.9029 - cbartone@ama.org
SOURCE American Marketing Association
American Marketing Association
Web Site: http://www.marketingpower.com