Uncorking white wine liking: Combining analytical chemistry and chemometrics with crowd-sourced data to predict quality ratings

Quality perception is a crucial aspect of wine research, particularly regarding consumer acceptance and quality impression. These assessments are challenging because consumers frequently incorporate extrinsic cues when evaluating hedonic quality. A unique perspective on wine quality is provided by the platform Vivino, which utilizes crowd-sourced consumer reviews on a scale of 1 to 5. This study investigates the relationship between chemical composition and consumer liking, using Vivino ratings as quality endpoints. German white wines were analysed using gas chromatography–mass spectrometry and Fourier transform infrared spectroscopy to establish comprehensive chemical profiles. Mid-level data fusion was applied, followed by variable selection to identify significant chemical features used to construct a Partial Least Squares regression model for predicting Vivino quality ratings. Our findings demonstrate that chemical properties can predict consumer-perceived quality parameters, validating our hypothesis and underscoring the potential for further research to refine and expand predictive models in wine quality assessment. • Assessing consumer quality perception can be challenging due to extrinsic factors such as price, labeling, and region, which influence quality ratings. • GC–MS and FT-IR techniques offer a detailed chemical profile of white wine that aligns with the genetic relationships of wine grape varieties. • A multimodal model combines aroma, physicochemical parameters, and crowd-sourced quality ratings to evaluate wine quality. • Consumer quality ratings can be predicted based on chemical markers.