Prediction of Shiraz wine flavour from volatile and odour intensity profiles

Abstract

ABSTRACTWine flavour is a critical driver of customer impression of quality associated with liking the wine and tendency to repurchase. However, winemaker and expert assessments poorly reflect customer impressions of wine flavour, and currently, available predictive capabilities are limited. This study developed predictive models for customer perceptions of Shiraz wine flavour, utilising the proportion of customer wine reviews mentioning a descriptor as the indicative measure to be predicted. To achieve this, two strategies were explored, which involved polynomial regression sourcing input terms from either: (1) volatile profiles directly, or (2) odour intensity (OI) values, produced via transformation of the volatile profiles using modified vector modelling (MVM). Neither strategy significantly outperformed the other (P = 0.834), however across the two strategies, a subset of 18 models from the 169 models generated had very good predictive potential (10-fold R2 > 90%). The 18 models presented herein provide a novel capacity for winemakers to predict customer perceptions of their Shiraz wine prior to market launch, to guide a marketing strategy to maximise customer satisfaction and thereby product success.KEYWORDS: Odorflavormodified vector modellingconsumer perception AcknowledgementsKW and JSP carried out statistical modelling, odour vector modelling, constructed figures, and drafted the manuscript. JL and KH assisted with manuscript drafting, editing, and revision.Disclosure statementNo potential conflict of interest was reported by the authors.Additional informationFundingKW received a Royal Melbourne Institute of Technology (RMIT) scholarship and was supported by Nutrifield Pty Ltd. JL was supported by Cannabis & Biostimulants Research Group (CBRG) Pty Ltd.