CONSUMER PREFERENCE MODELING
Emotion and experience are two concepts that have been brought to the forefront of design. Both have been demonstrated to have a significant impact on consumer preference, product sales, and company success. The goal of this work is to advance the state of consumer preference modeling to account for both emotion and experience.
Traditional conjoint-based approaches to preference modeling require a participant to judge preference for a product based upon a static 2D-visual representation or a feature list. While the aesthetic forms and functional features of a product are certainly important, the decision to buy or not to buy a product often depends on more, namely the experience or feel of use. To address the importance of the product experience we introduce the concept of experiential conjoint analysis: a method to mathematically capture preference for a product through experience-based (experiential) preference judgments.
Experiential preference judgments are made based upon the use, or simulated use, of a product. For many products, creating enough physical prototypes to generate a preference model is cost prohibitive. In our work, virtual reality (VR) technologies are used to allow the participant an interactive virtual product experience, provided at little investment.
In order to provide participants with a virtual product experience we created a custom solution. The virtual environment was created using the Open-source Graphics Rendering Engine (OGRE). 3D models of all the products used in the studies were created in Solidworks and converted to the appropriate formats using Blender. The Polhemus Patriot was used to track the 3D position and orientation of the hand and the 5DT Data Glove 5 Ultra was used to capture the motion and bending of the partisipant's individual fingers. A custom set of interaction was created for each product. Together these components enabled participants to interact with and gain an understanding of the aesthetic and experiential attributes of a product.
In order to capture preference we designed and ran a conjoint-based study. Two products, truck dashboards and coffee mugs, were chosen to examine. The designs of these products were analyzed to identify the attributes that, when changed, would most likely influence a partisipant's experience. With these attributes identified, design of experiments was performed to determine the specific product configurations to be experienced by the participant during the study. Next, a survey was conducted in which participants were allowed to interact with and experience the use of the various product configurations in a VR environment. The experiential preference judgments provided by the participant in the survey were then employed to generate preference models whose predictive ability was validated using the Pearson product-moment correlation coefficient (PCC) the mean absolute error (MAE). Two studies were performed. The first study was designed to capture layout preference for truck dashboards. The second study was designed to capture form preference for coffee mugs.
The results of this work first and foremost demonstrate that preference models can be derived based on experience in a VR environment. Further the results demonstrate that the preference judgments of virtual product representations are more similar to preference judgments of real products than preference judgments of 2D product representations are. When examining similarity of modeled preference, experiential conjoint is found to be superior to previous 2D visual-based methods with respect to mean absolute error, but with respect to correlation no significant difference is found. Additionally, we find that providing additional interaction-based information about a product through a product experience does not hinder the ability of participants to provide consistent preference judgments nor the ability of the resulting preference models to predict preference.
Emotion has a strong influence on both preference and consumer behavior. Thought once to be un-important, emotion is now considered at all stages of product development. Traditionally, emotional responses to products are measured through analysis of self-reports provided by consumers.
Our work introduces a method to passively measure emotional responses to product evaluations by capturing changes in the facial expression of the evaluator. A product evaluation experience is created that allows for the evaluation of high-quality 2D-color product images. Video is recorded during the evaluation with both consent and prior-knowledge. These videos are processed in order to track the locations of landmarks on the eyebrows, mouth, and nose. The locations of these landmarks are then used to evaluate the facial expressions that appear during the product evaluations.
This work specifically explores the connection between preference, choice, and emotion. As part of a study, participants are asked to evaluate 45 products: five classes of products each with nine example products. During the product evaluation partisipants provide a numerical rating representing their preference for the product. They are additionally asked to identify the product within each class that they would be most likely to purchase for themselves. This work demonstrates that consumers exhibit emotional reactions to products through changes in their facial expressions. The focus of this work is now moving toward a system that can infer important information, such as preference and willingness-to-purchase, based upon the facial expression exhibited during product evaluation.
Primary Researcher: NOAH TOVARES