On The Design of a Conjoint Analysis: Some Empirical Evidence
(A2025-125581)

AUTHORS

Pablo Marshall, Pontificia Universidad católica de Chile, PUC

ABSTRACT

Conjoint analysis is a methodology used in marketing to study consumer preferences. The design of these studies is crucial for researchers and decision-makers. Decisions related to the number of attributes and levels, the number of tasks and options each respondent evaluates, the use of a "no-choice" option, the sample size, and the use of training questions are key definitions in the design of conjoint analysis studies. This study uses 15 datasets from commercial applications with actual data from various industries to analyze the impact of different design decisions on the performance of a conjoint analysis study that predicts consumer preferences using the Choice-Based Conjoint methodology. The results of the study show that each additional attribute reduces the probability of a correct prediction by 7%, that more than ten tasks in the survey do not generate significant gains, that sample size is only relevant if the number of tasks is fewer than 10, that there is slight fatigue in respondent answers as the sequence of questions progresses, and that the use of training questions does not lead to significant improvements.