We have exciting news. It has been a fascinating year of wine discoveries and delights for us, and it is finally time to share with you some of the fruits of our labor.
The first set of surprising results are coming out of our algorithm, and the data is worth exploring. As a reminder, our Artificial Intelligence collects data based on user’s blind ratings of our wines, and uses it to recommend better wines for each individual in the future. In addition to allowing us to better serve our customers, it has given us unique insights into the wine world. And the capacity to challenge certain long standing assumptions.
For one thing, American style wines do not enjoy the privileged position wine that industry ‘wisdom’ would have you believe. In fact, customers are giving more favorable treatment to European style wines, even if they come from here in California.
How the data is gathered
Before we send even a single bottle to a consumer, we carefully evaluate the wine on a number of different scales. Experienced sommeliers like myself tackle the difficult process, not of judging the wine, but of establishing its complex profile. With up to 200 data points per bottle.
These are the data points that the algorithm then uses to line up wines with consumers, based on the traits of the wines they have enjoyed in the past. With more customers comes more data: we began to observe patterns in their behavior. Patterns which help us understand their preferences.
We noticed right away that we were able to group wines into different clusters based on their style. Wines themselves fit into broad categories depending on their shared characteristics, which more or less broadly fit into established ways of viewing wines.
These style clusters break the norm in that instead of grouping wines by their region or grape, we group them by their taste. We may mention a familiar wine in the description of the cluster as a benchmark, but it’s the data that brings each wine in that cluster together. For example, we can group red wines together that have high alcohol, full body, and lean towards darker, riper red fruits—think Napa Cabernet. However, not all the wines in that cluster are necessarily from Napa or even Cabernet. It could be a ripe Nero d’Avola made in Virginia!
Once we associate the personal data with the wine clusters it becomes even easier to observe, and understand, what the data is telling us.
What the initial results are telling us
Our best performing red ‘clusters’ are the Baron (Bordeaux style reds), Savage Beast (like Northern Rhône Syrahs), and the Mediterranean (for that Rioja feel). This runs counter to the wine industry rhetoric of American’s preferring big oaky Napa-style cabernets like those favored by influential critique Robert Parker. Instead American’s appear to be leaning towards styles commonly associated with European wines.
The fascinating insight is that this isn’t a bias in favor of foreign wines, because American wines with non-traditional profiles are also increasingly popular. The customers don’t seem to be favoring a particular region but instead a particular style of wine. This information could shake up an industry defined by decades of ‘Parkerization’.
For white wines there appears to be a similar trend. Our members prefer more acidic, fresher wines in lieu of the traditional oakey Californian style. The top white clusters were the Salty Aquaholic (Chablis style whites), the Barista (think Pinot Grigio), and Oversized & Opulent (similar to Condrieu). It’s important to note that Napa Chardonnay still ranks quite highly showing that the orthodox style remains popular, but hardly dominates. Instead the results show a wide open field, very different from the ranked hierarchies so many wine aficionados adhere to.
There is room for a lot more information
The initial results are certainly promising, and surprising. They are mere hints at the vast amount of intriguing information out there. We will continue to arduously gather more as customers continue to rate wines, and new members and bottles join our adventure.
Already we have demonstrated that the received wisdom in the wine community is not backed up by the evidence, but that leaves a huge gap. If Americans do not prefer the classic California/Napa style, what do they want? What makes a wine stand out when you take away the label, the medals, the branding and marketing? The taste, of course.
Ultimately taste is unique and subjective, and each member will enjoy each wine differently. This is the fundamental truth behind our model, as we seek not to rank wine, but to match wines to members who will enjoy them. And while we continue to deliver to our members wines they will enjoy, they will continue to feed us the information needed to uncover trends and solve mysteries in the wine world.