Stumbling through overwhelming aisles of wines, it can be daunting to find the perfect match to that asparagus-heavy dish. There are pairing tables, books and websites filled with descriptions and ratings, but they aren’t quite an exact science. Seeking out and buying a wine, only to realize you’ve gone the wrong direction, is something most people experience at least once. What makes pairing and choosing wine easier?

Big data. Not just brief tables or small collections of suggestions, but data-driven algorithms and analysis. It might sound unromantic, but several wineries already use data extensively during the growing and planning processes. It can help evaluate existing wines, and even predict what the next wine will taste like.

First, one question: how do you like your tea?

Many existing matching systems simply ask which typical wine flavors the reader already enjoys. Do you like fruity tastes? Cedar tones? Sweet tobacco? These don’t quite get at the root desire or palate. They’re descriptions. While the average person makes purchases based on memory, label, price, and a hint of randomness, data can look beyond what may be “popular” or familiar and return a wine that matches based on an individual’s tastes. By asking questions like “how do you like your tea?” they can determine something much more fundamental about overall preferences. The result is that they don’t just give you popular rieslings, or famous pinot noirs. They return very specific wines of different types.

Using a machine learning algorithm, results get even better with time. With each returned survey and rating, matches can become even more exact and reliable. This is the technology millennials are ready for. No more guessing. No more disappointing buys. An app that can accurately suggest wines for users based on a multitude of factors (price range, origin country) will be a go-to for wannabe connoisseurs. There are several sites and apps that aggregate ratings to determine what a “good” wine is, but wine is hardly a “one size fits all” market. There are folks who love heavy, full-bodied wines and those who want their wine to taste like grape juice. Plus, more often than not, casual wine drinkers don’t actually know what they want. They may say they like “fruity” wine, when they actually mean “sweet.” They also tend to value combinations of flavors, rather than just one single flavor. Objective data can cut through incorrectly used words and flawed suggestions. Customers almost always have a smart phone by their side, and are faced with a countless number of wine options. The future needs apps to help us find the perfect wine with ease. Wine selection isn’t just for connoisseurs anymore.

Mapping patterns for profit

For connoisseurs and wine makers, there are even more options available with big data and machine learning. Mapping weather data and well-known top vintages of Bordeaux, one man was able to crack the data code on how to tell if a Bordeaux will be good. H20’s Alex Tellez used 60 years of weather data from the growing region and let the machine learning algorithm do the work. By having the machine learn weather patterns from the average years, it should be able to detect anomalies and indications of an extraordinary vintage—and it did. The machine could correctly identified every year that produced a great vintage. That in itself is incredibly cool, but not yet useful. He had quantified something that was always left as an art form, but he wasn’t quite finished. The next step was to use that algorithm on current datasets and predict how the next vintage would fair. It seems even the most experienced wine snob can’t beat data and machine learning. More importantly, predicting those great vintages could lead to some incredible investing opportunities—the project’s new end-goal. Where else could big data be useful?

Vineyards. Many large vineyards are already equipped with plenty of sensors that relay information to the cloud. Enjoying a glass, it is easy to forget just how much science is put into creating that wine. Grapes are carefully planned, evaluated and watched. Science and concrete knowledge has always been important to growing crops, but big data algorithms are now being leveraged to do everything better. The age-old practice of growing wine may have developed over the years, but there hasn’t necessarily been a revolutionary change in a very long time. In fact, many vineyards also rely more heavily on experience and guesswork than science and data. Combining the IoT, big data, algorithms and machine learning, technology has opened a new world of opportunity for wine growers.

In California where most crops are irrigated, it can be tough to know when and how to actually water the vines. Though it seems absolutely fundamental to have concrete answers to these questions, the process is often based on guesswork. This is both shocking and a great opportunity for tech companies, given how even smallest variations can make a huge impact in a wine’s flavor. One of the biggest selling points of near continual gathering and analysis of data is the chance to make realtime decisions. This huge hole in the market that is quickly being capitalized on. Gathered plant data is sent to the cloud, where is is analyzed. The results of such companies have been great, though not necessarily staggering. It seems vineyards are still slow to adapt and, of course, need time to fully utilize the information.

These platforms are also relatively new. Additional research and technology is still needed to perfect these systems. Big data and algorithms are not omniscient, and can’t accurately predict in a vacuum. Luckily for consumer-based programs, there will be no shortage of feedback. Each time a customer likes or dislikes a wine, the machine learning algorithm only gets better. They’ve already begun taking power from the industry, where the words of a few famed connoisseurs often determine what is “good” or “bad.” That’s one way to strike a profitable chord with millennial wine drinkers.

image credit: Jordan Johnson

Like this article? Subscribe to our weekly newsletter to never miss out!

Previous post

Beginner's Guide to the History of Data Science

Next post

Why Your Data Scientist Isn't Being More Inventive