Can we use wine back labels to figure out wine quality?

wine back labels
Mark Thornton: “The words — and not what they mean — on wine back labels are a clue to wine quality.”

Because, finally, someone has discovered a way to measure the relationship between what’s written on wine back labels and the quality of the wine.

The breakthrough came from a Harvard Ph.D. student named Mark Thornton, who took data from 75,000 wines in the Wine.com inventory, and compared what was written on their back label — and not what the words meant — with ratings from the site’s users and from wine critics.

The findings? That certain words appear on the back labels of wines of lesser quality, while certain words appear on the back labels of wines with higher ratings. Thornton told me he knows this isn’t perfect, given how scores and wine ratings work, and he wants to improve that part of the study. In addition, he wants to refine the way his software decides which words to analyze, perhaps eliminating regions and better understanding phrases, like grilled meats instead of grilled and meats.

What does matter is that Thornton’s work is apparently the first time anyone has done this kind of research, making it as revolutionary as it is helpful in deciphering the grocery store wall of wine.

Thornton, whose parents teach in Cal State Fresno, says this study interested him because it’s about wine, which he likes, and because it ties into his PhD research, which deals with how we describe things. One of the concepts this study takes into account is called “naive realism,” in which we assume that what we sense has to be true for everyone, when it obviously isn’t. Which dovetails neatly with wine.

Thornton’s findings confirm many of my suspicions about wine back labels, as well as how critics use descriptors. The word clouds on his site summarize the results; I’ve set them up so you can see them more easily here for the consumer ratings and here for the critic ratings.

These are among the highlights of the study:

? Restaurant food pairings or terms like pasta appear on the labels of the lowest-rated wines. Thornton says this may well be because the wine doesn’t have any wine-like qualities to recommend it.

? Words used to describe sauvignon blanc — grapefruit, herb, clean — show up on the critics’ lowest-rated white wines. This is not surprising, given that sauvignon blanc has always garnered less respect from the Winestream Media than chardonnay.

? A location on the back label seems to indicate lower quality white wine; “handcrafted” is in the higher quality word cloud. For reds, “value” and “soft” are poor-quality words, while “powerful” and “black,” probably used to describe black fruit, infer higher quality. Handcrafted is especially interesting, since it doesn’t mean anything in terms of wine production.

Finally, a word about prices, which is also part of the study. Thornton divided the consumer ratings into five price ranges, and there was little difference in perceived quality between the first three ranges. In other words, you get more value buying the cheapest wine. Shocking news, yes?

The critic price-value rankings were even more bizarre. The worst value came from wines that got scores in the mid-90s, while wines in the high 90s (and even 100) were less expensive, and the best value wines were around 90 points. Thornton says he isn’t quite sure why this is true, though it may have something to do with critic bias. My explanation is simpler: Wine scores are inherently flawed.