Concurso mineiro de cachaças e aguardentes: uma abordagem estatística dos dados de treinamento de jurados.

Abstract

Sensory analysis is fundamental in evaluating food quality, especially in cachaça competitions, where it enables detailed characterization of samples and selection of the best beverages. The reliability of results depends on the consistency of the evaluators, requiring adequate training and standardized protocols. This study aims to perform a statistical analysis of sensory data gathered from the training process of 26 judges to assemble a reliable panel for a cachaça competition in Minas Gerais, correlating this data with chromatographic analyses of the samples and the evaluators’ profiles. Five distinct samples were used: Cachaça 1 (white, stored in stainless steel vat), Cachaça 2 (aged in bálsamo wood), Cachaça 3 (white, stored in stainless steel, with higher alcohol content than Cachaça 1), Cachaça 4 (aged in oak), and Cachaça 5 (aged in amburana wood). The training process lasted three months and included an introductory meeting on the competition, basic sensory analysis concepts, explaining overall test rules, and laboratory structure. Triangular and multiple comparison tests were conducted to assess the judges’ ability to distinguish, quantify, and describe sensory differences. Judges were also given the competition evaluation form to clarify further doubts and suggest improvements. Gas chromatography (SPME/GC-MS) was performed to identify and quantify volatile compounds, correlating this data with the judges’ comments. A structured questionnaire collected sociodemographic data and consumption habits of the judges. Data analysis included descriptive statistics, Analysis of Variance (ANOVA), Tukey tests, and Hierarchical Cluster Analysis (HCA) to identify groups of judges with overlapping preferences. Consistency of evaluations was assessed by Cronbach’s alpha and Intraclass Correlation Coefficient (ICC). Word association analysis was also conducted to identify sensory descriptors linked to the samples. Results indicated that judges could statistically differentiate the samples and recognize the standard sample. Cluster analysis revealed three judge profiles, some performing with more consistency and others with greater variability. Exploratory research enabled a better understanding of the panel, identifying factors influencing sensory perception. Correlation between sensory and chromatographic data showed that compounds such as esters and phenols directly impacted aroma and flavor, confirming the relationship between chemical composition and sensory quality. Cachaça 5 had a Cronbach’s alpha of 0.696, indicating moderate consistency, while other samples had values below 0.6, suggesting low consistency and greater response variation across three sessions. ICC indicated low reliability in individual evaluations due to high variability in scores, but reliability improved when considering mean evaluations, especially for Cachaças 2, 3 and 5, which showed ICC above 0.5, indicating moderate variation. The results indicate that although judges could distinguish among the cachaças, low individual consistency and reliability suggests a flaw in the current training process and further evaluation is required. Focused training on sensory attributes, protocol standardization, and continuous panel performance monitoring are essential. Integration of sensory analysis with statistical and chromatographic data is crucial to understand factors determining cachaça quality, ensuring more reliable and consistent evaluations.