Authors:
Toshiyuki Ono
1
;
Hirofumi Matsuo
2
and
Norihisa Komoda
3
Affiliations:
1
Systems Development Laboratory Hitachi, Ltd., Japan
;
2
Graduate School of Business Administration, Kobe University, Japan
;
3
Graduate School of Information Sceince and Technology, Osaka University, Japan
Keyword(s):
Product Preference, Online shopping, Multiattribute utility function, Conjoint analysis
Related
Ontology
Subjects/Areas/Topics:
B2B, B2C and C2C
;
B2C/B2B Considerations
;
Business and Social Applications
;
Communication and Software Technologies and Architectures
;
e-Business
;
Enterprise Information Systems
;
Society, e-Business and e-Government
;
Software Agents and Internet Computing
;
Web Information Systems and Technologies
Abstract:
Obtaining timely information on consumer preference is critical for the success of marketing and operations management. In a previous paper we proposed a method of estimating consumer preference by using their history of browsing among possible configurations of personal computer in an online shopping environment. It consisted of three steps: (1) collecting data on each consumer’s browsing history regarding quotations and purchase requests, (2) converting requests for quotations and purchase order data into ordinal preference data, and (3) estimating consumer preference for product attributes by applying a multiattribute utility function. The underlying assumption with this method was that a product configuration that was quoted later would be preferred to those quoted earlier. Another assumption was that how many times a product configuration was quoted would not affect estimates for product preference as long as this was quoted at least once. Although these assumptions are critical
in estimating consumer preference, their validity has not been examined. In this paper, we evaluate the validity of such hypotheses regarding the relationships between consumer preference and the sequence and frequency of quoted product configurations, and propose six methods of estimating consumer preference. We show through experiments that, for about 60% of examinees, all the proposed methods could approximate consumer preference obtained by conjoint analysis, and that the six methods have almost equal accuracy. We therefore concluded that any of the six methods could be used equally well for estimating consumer preference in a timely fashion.
(More)