Skip to Main content Skip to Navigation
Master Thesis

Analysis of consumers' preference drivers: can we go beyond existing practices?

Abstract : One main concern for product developers is to answer those three questions: “What products do consumers like?”, “Why do consumers like these products?” and “How can we deliver them what they like?”. Within this context, identification of consumers’ preference drivers is of utmost importance to guide and support product innovation and renovation, in order to develop better and successful products that meet consumers’ expectations. Quantitative consumer product tests allow gathering knowledge on consumers’ preferences regarding the products, while sensory descriptive profiles provide insights into the sensory characteristics of the same product space. Main objective is then to relate consumers’ liking to sensory properties of the products. Current practices in Nestlé are based on internal preference mapping and linear correlations. However, Consumer Specialists pointed out several limitations, especially regarding the identification of consumers’ preference drivers. Hence, alternative methods have been investigated, and main focus was on Partial Least Squares Regression (PLSR). Its ability to analyze data with numerous predictor variables, even if they are noisy and highly correlated, makes it a promising approach. Besides, PLSR offers good compromise between a good summary of sensory attributes and a selection of the ones which best explain consumers’ liking. As all consumers do not like the same products, they are clustered into homogeneous groups prior to analysis. PLSR of the mean overall liking per cluster on the sensory attributes have been performed. Non-linear effects, including quadratic terms and interactions, can also be included by augmenting the predictor matrix. For each cluster of consumers, significant sensory attributes are highlighted, and a biplot, based on the PLS components, displays on a same representation space, the products, the sensory attributes and the consumers. This method has been implemented as an Excel macro, and outputs have been adapted to provide a user-friendly interface. PLSR outperforms current practices to uncover the sensory characteristics driving consumers’ preferences, and the results prove its ability to take account of non-linearity. However, further investigations could be considered regarding the expression of significant PLS coefficients, alternative ways to include non-linear effects in the model or the adjustment of the methodology to the product category.
Document type :
Master Thesis
Complete list of metadata

Cited literature [26 references]  Display  Hide  Download
Contributor : Agrocampus Ouest <>
Submitted on : Thursday, October 22, 2015 - 9:41:50 AM
Last modification on : Monday, July 6, 2020 - 3:38:21 PM
Long-term archiving on: : Saturday, January 23, 2016 - 12:39:16 PM


 Restricted access
To satisfy the distribution rights of the publisher, the document is embargoed until : 2025-10-31

Please log in to resquest access to the document


  • HAL Id : dumas-01219121, version 1



Chloé Barthomeuf. Analysis of consumers' preference drivers: can we go beyond existing practices?. Life Sciences [q-bio]. 2015. ⟨dumas-01219121⟩



Record views