Fresh fruits are always recommended as ingredients in healthiest diets. However, there is a tendency for consumers to move their consumption towards transformed fruits, which are integrated in many food products. Quite commonly fresh fruits are difficult to handle and store but they also do not have regular quality when they reach consumers. There are many other elements besides the physical characteristics, which are very important for consumers, and they can be promoted through marketing actions. It is very important to understand why consumers make elections of fresh fruits in order to increase their consumption. The aim of this study is to understand how consumers make their purchasing choices based on the most important peaches’ attributes and levels. In Spain there are 20 fruits with the Protected Designation of Origin (PDO) label. Among those PDOs, only one brand certifies the peaches’ origin and it is called “Calanda Peaches”. This fruit has been selected to test several hypotheses about consumers’ fruits choice. The survey collects information from questionnaires applied to PDO Calanda peaches` consumers that were attending two hypermarkets in Zaragoza city, in 2009. An attribute‐level best‐worst experiment was undertaken, respondents stated the most and the least important characteristic in their purchasing. Each characteristic, or alternative, is an attribute associated to a level of that attribute. In our case, nine hypothetical products were presented from different combinations of 4 attributes, with 3 levels in each attribute, (price: 1.2 €/kg, 2.4 €/kg and 3.6 €/kg; origin: PDO Calanda, non PDO Calanda and non Calanda; packaging: bulk, conventional packaging and active packaging; and fruit size: small, medium and big) to allow main effects estimation. Data were analysed using Weighted Least Squares (WLS) by in Best‐Worst Paired (BWP) and Best‐Worst Marginal (BWM) methods. Both models allow the attribute and attribute’s levels impact estimation on consumer purchase decision. They also have similar measurement properties, but as Paired models have more observations per respondent, they present smaller standard errors. Results show that both models have good performance. Consumers give different weights to the attributes when they buy peaches. There is an overriding influence of the origin especially for the attribute‐level Calanda in comparison with the rest.
