Anatylical models shed light on the inner workings of wine competition scoring

In 2021, The prestigious Concours Mondial de Bruxelles (CMB) wine
competition began using artificial intelligence (AI) developed by Winespace, a Bordeaux-based firm founded by Sylvain Thibaud and Julian Laithier in 2015, as part of its judging protocol as a value added feedback mechanism for producers. (CMB also collaborates with UC
Louvain in Louvain-la-Neuve, Belgium, on further analysis that evaluates the performance of its juries and judges.)

I first experienced Winespace’s rubric for sensory analysis when judging the México Selection by CMB, held in Guanajuato, Mexico, in 2021. At the time, its technology was a promising prototype, and the collaboration with CMB helped Thibaud and Laithier to commercialize what is now a robust platform called Tastee AI that has since been adopted by CMB and others.

In April 2024, I used the tasting grid again during the Sauvignon Selection by CMB held in Leibnitz in the Austrian state of Styria, where a panel of judges recorded tasting notes in as many as eight different languages. The technology seamlessly translated and analyzed those notes and compiled a sensory analysis for each wine.

According to Thibaud, the resulting reports, which have been shared with producers for the past two years, offer a collective picture of the wine’s quality and style in the form of summarized tasting notes, an aroma wheel, a list of specific strengths and weaknesses, and comments for improvement.

“From the CMB’s point of view, this approach also reflects a desire for transparency in the wine assessment process, thereby reinforcing the
industry’s confidence in the seriousness and value of the medals wine competitions award,” he observes.

Deeper analysis of performance by wine competition judges

In the future, Winespace plans to provide feedback to each judge on their taste preferences, scoring style, and other criteria. “One of the aims of analyzing comments is to identify the preferences and writing habits of each taster and to better understand the frequency of the concepts and descriptive vocabulary they use,” explains Thibaud.

“We can also identify the criteria they seem less sensitive to and areas where they are less expressive.” This type of rigorous analysis soundly
debunks the commonly held belief that wine sensory evaluation is
purely subjective. For example, analysis of my five-person jury
for the 2024 Sauvignon Selection competition (provided by CMB and
prepared by Christian Ritter of Ritter and Danielson Consulting along with UC Louv revealed a highly correlated panel working with astounding consistency.

The eggshell plot shown here illustrates the performance of a coherent wine competition jury, one that arrived at a consensus in their scoring when evaluating wines of the same or similar quality (namely Sauvignon Blancs from world-class regions). The panelists’ close alignment is represented by the cluster of jagged curves just above the smooth curve, which is the control metric for a theoretical jury with identical scores.

This is one of half-a-dozen metrics reported by Ritter, all of which are
immensely useful in understanding my performance and the performance of my colleagues over the course of the competition. Metrics of this caliber could help the industry address valid concerns about inconsistency in scoring caused by the different models and scoring systems used by commercial wine competitions.

In a further exploration of Tasteee.ai in practice, I conducted a study on sensory training methods for novice wine consumers which was presented at the American Society of Enology and Viticulture on May 18, 2025. The three-phase study determined that despite consumer and marketing research pointing to preferences for gamification and hedonic learning experiences, a Gen Z cohort showed no preference for gamification over the traditional rubric/lexicon models used by the Wine & Spirit Education Trust or the Court of Master Sommeliers.

Also read: