Blog Firstleaf The Data Science Behind Firstleaf Wine

The Data Science Behind Firstleaf Wine

The Changing World of Wine 

Breakthroughs in technology seem to be coming faster and faster, changing the landscape of everyday life, making complex tasks seem routine, easy or even mundane. It used to take weeks to travel from New York to London by ship, but now we can fly across the Atlantic in a few hours. We have maps on our phones, television on the internet, and cars that drive themselves. So why is finding good wine still so hard for most people?

The Firstleaf Difference

One of the reasons is that other wine clubs and the everyday wine enthusiast don’t have the technology available to Firstleaf members. Our custom algorithm is unparalleled in the wine industry and offers the best way to understand an individual's unique tasting profile.  Every rating a member makes influences what wines are recommended for the next box by utilizing the algorithm that was built by Dr. Matthew Martz, our on-staff computational biochemist. It sorts every member rating through thousands of simulations in less than a second to pick the best wines for that individual. It’s one of the reasons that we can say with 100% accuracy that Firstleaf is personalized to you. It’s also why we say that Firstleaf gets better with every box.

To join other wine clubs, you might have taken quizzes asking about the types of chocolate you like (to gauge your preferred sweetness vs. bitterness levels) or what kind of citrus you eat. These clubs match your answers with subjective, large-group “tasting profiles” that never change, which means that your club experience really isn’t personalized.

We work exclusively at the individual level, no grouping members with other people’s preferences—our algorithm takes into account each unique set of taste-buds. At Firstleaf we update profiles with every rating and every box, giving members recommendations with unprecedented specificity. With this technology we can give differentiated and accurate suggestions for every club member. So, how exactly does this high-tech wine algorithm work?

This picture shows chemical attributes in a multidimensional space, as the algorithm is classifying several new wines.

How Our Algorithm Works

The process began with a chemical analysis of all our wines. We needed to explore how to group each bottle based on shared and unshared components. The results give us a base-line of hard information that our wine experts can use to discover what members love and what kind of wine we should be looking for. The program searches the data to recognize all the patterns and relationships between the different wines. Dr. Martz's job was to apply this new information and to determine how to arrange and translate these numbers into something that our wine experts can understand.

This is a picture of our alignment algorithm that takes the chemical attributes and sorts them based on varying factors into suggestions for a member.

New wines are placed relationally in the three-dimensional graph and are thus “mapped” into a location based on their chemical makeup. By adding in a specific member’s likes and dislikes to the algorithm, the program is able to group and measure the relevant information and make recommendations for that club member, running thousands of outcomes in less than a second! There is no “collaborating” or short-cut built in that groups a member with anyone else, everything is individual.

What Firstleaf Members Can Expect

Over time the algorithm “learns” what a club member likes to drink. It’s possible that it will suggest something that may seem like outside choice, but actually shares many of the favored characteristics of other wines. Half the fun of a wine club is discovering new things that you will love!

At the top of every suggestion is the swap button, if you don’t like the look of a wine, swap it out!

Swap It Out!

Our members can take in the full spectrum of a wine varietal’s composition. Don’t like the over-extracted, buttery, California Chardonnay “oak bombs?” There's no need to write off the grape entirely! Our algorithm might suggest a Chablis style Chardonnay that shares the high acidity of your other favorite wines. By hitting “swap” the algorithm recalculates to pair a different wine that still shares some of these characteristics. Remember that you are in control! The more wines you rate, the better the suggestions become. 

Think about the algorithm as taking math and translating it into wine tasting notes. Dr. Martz has essentially built an online translator for wines. Even if you don’t know why you like a certain wine, our program is determining what you like at a chemical level.

What do I have to do? 

Your feedback matters! It’s why every wine, every thumbs up, and every thumbs down plays into predicting the best wine for you. Your comments and ratings are the best tools our experts have in making sure that Firstleaf gets better with every box!

You may also like

%d bloggers like this: