Influence of label information on purchase intent for wines – analysed by con-joint analysis and L-PLSR

Line Holler Mielby¹, Michael Bom Frøst¹ & Hildegarde Heymann²
¹University of Copenhagen, Department of Food Science
²University of California – Davis, Department of Viticulture and Enology

To elucidate the impact of labelling on purchase intent of wine, two studies were conducted, with Cabernet Sauvignon (CS) and with Chardonnay (Chard) labels.

For both wine conditions three label attributes (origin, back label information and price) each at three levels were used. For each wine condition 27 hypothetical wine labels were created and evaluated by 138 CS consumers and 120 Chard consumers.

To study the relationship between wine knowledge and purchase intent, each consumer completed a wine knowledge questionnaire. For conjoint analysis (CA) consumers were segmented based on their wine knowledge.
 

For each wine an L-PLSR analysis was performed using three matrices; a wine label design matrix X, a purchase intent matrix Y, and a consumer background (wine knowledge) matrix Z. The Y matrix shared one common dimension with the X and the Z matrix.
 

The two data analysis methods gave quite similar results. Regardless of wine type, overall purchase intent was most affected by price, (less expensive led to higher purchase intentions). Back label information also drove purchase intent, and wine origin was not an important driver. Consumers’ wine knowledge did not have much effect on their purchase intent though there was a weak tendency for the most wine knowledgeable consumers to be less affected in purchase intent by the wine price.
 

Although the two analyses were similar, they reached the target differently. CA simplified the results by denoting relative importance values for each factor and level. With L-PLSR the visual geometrical display made interactions and relations clearer. Both methods are highly usable, but L-PLSR analysis has the advantage of displaying the data for each individual consumer and it can cope with collinear data. However the results can be more difficult to interpret, since consumers are often positioned a swarm rather than nicely clustered segments.