We have evaluated many existing digital and traditional systems, but none of them fulfilled our core criteria: Being able to deliver better recommendations to anyone, no matter their experience in wine.
wineMatcher is not another wine ranking system: Wine ranking systems try to define from an objective point of view how good a specific wine is, but intentionally ignore that every human being has a different taste. wineMatcher is the exact opposite: It only focuses on your personal palate to define whether you will like the wine or not.
We analyze every wine we recommend using a combination of proprietary analytical and flavor chemistry methodologies, specifically designed to understand how the unique and complex characteristics relate to the human palate.
Why are we able to give better recommendations with wineMatcher than a Sommelier in a restaurant or a sales person in a store?
wineMatcher changes this completely: With generic questions about your taste, we can recommend wines that match your personal taste.
Tastry’s solution is unique in the market: It analyses the wine chemically for more than 100’000 criteria to feed its artificial intelligence engine. Based on the palate of thousands of experts and wine lovers, it “learns” which type of wines match to which palate.
To learn your taste, Tastry’s AI asks you 10 initial questions, which allows us to predict wines you will like to a very high degree. With every wine you taste, you provide us with a simple feedback (“I like”, “Neutral” or “Dislike”). This way, the AI learns your personal taste and will deliver you better and better results over time.
We have asked Katarina Axelsson, CEO of Tastry and Elian Kool, Co-Founder of winemaker.com about the concept of wineMatcher and how it works.
Taste is personal - how can a computer know my preferences?
Artificial Intelligence (AI) or "machine learning" refers to technical processes that gain insights based on a large amount of data (in the case of wineMatcher, over 200,000 data points). Good AI algorithms are characterised by the fact that they constantly improve and refine the accuracy of their statements based on feedback.
How does wineMatcher work?
Through simple questions, wineMatcher learns your taste preferences. On the other hand, there are over 200,000 measuring points, which are determined using a complex procedure. wineMatcher does not make a statement about absolute good or bad, but about whether a wine is particularly well suited to a preferred, personal taste.
Why 10 questions?
Of course, one could make even more precise statements with 50 questions. However, the accuracy rate is already very high with 10 questions and is only slightly improved by asking additional questions. Your individual feedback on the selection, once the wines tasted, has a greater influence on the algorithm accuracy improvement.
Are the questions asked always the same?
wineMatcher by Tastry regularly reevaluates what the best questions are to evaluate wines and their characteristics based on all the wines tested.
How many wines were tested?
About 50 wines were analysed for the current beta phase. We plan to analyse a large part of our wine portfolio by Q4/2021.
Why are there different questions for red and white wine?
White and red wines are completely different worlds in terms of aromas and taste (we might like a "fruity" white wine - a sweet, fruity red wine rather not) - which is why wineMatcher asks different questions to assess individual preferences in reds and whites.
Do wineries pay for the analysis and scoring of their wines?
The costs for the analysis are borne entirely by winemaker.com and the results therefore are completely independent. The selection of the wines in the beta phase was made by the winemaker.com team, with the aim of conveying the widest possible range of styles.