Abstract
Taking consumer quality perceptions into account is very important for new-fruit product development in todays competitive food market. To this end, consumer-oriented quality improvement models like the Quality Guidance Model have been proposed. Implementing such mod- els in the agro industry is challenging. We propose the use of Bayesian Structure Equation Modelling for parameterizing the Quality Guid- ance Model, allowing for the integration of elicited expert knowledge. Such casual modelling would furnish important insights for determining the opti- mal fruit product in terms of consumer avour-quality perceptions. In the context of tomato breeding, where we have data about metabolites, sensory- panel judgments, and consumer avour-quality perceptions, we estimated a benchmark Bayesian SEM using non-informative priors, starting from an initial causal model derived from the data with a score-based Bayesian Network learning algorithm. The results so far have given some in- dication of the importance of accounting for consumer heterogeneity in the modeling process.