The Latest (Useless) Wine Technology

I’ll apologize right off the bat for being a horrible skeptic when it comes to these sorts of things. I thought I might be able to refrain from bashing this product that I’d never seen or experienced, especially as I was beginning the article about it. But as I was reading about a new kiosk solution for grocery store wine aisles that helps consumers find wines that they would like, I eventually got to the following description of the technology:

“What kind of matches does the system make? An occasional wine drinker who heavily salts his food, puts plenty of sugar in his coffee and hates diet sodas would be matched with Beringer White Zinfandel, 2004 or St. Supery Moscato, 2004, an Italian wine. Someone who doesn’t use salt, takes their coffee strong and black and drinks diet beverages would get a recommendation for B.V. Cabernet Sauvignon, 2001 and Franciscan Oakville Cabernet Sauvignon, 2000.”

It was at this point that I was forced to throw up all over this idea. Now I’m all for helping shoppers connect with products that they will like. The premise behind this solution is a good one, namely that there’s a dizzying array of choices available at your local grocery store when it comes to wine, and it’s hard to be sure you’re gonna get one you like unless you’ve tasted it before. But I would put hard cash on the fact that asking someone whether they drink diet soda, put cream in their coffee, and whether they tend to put salt on their food doesn’t get you any closer to knowing what sort of wine to recommend to them.

As someone with a good deal of experience in interactive design, especially for retail e-commerce, I can tell you that it’s UNBELIEVABLY difficult to make good recommendations to people using an automated system, even if that system is driven by highly complex artificial intelligence. And that’s true for something like shoes, or jewelry, which are both a lot simpler than tastes in food.

People’s tastes are so subjective, AND so difficult for them to describe, that even the most elegant, multivariate, multi-step wizard can’t begin to approach a solid recommendation, even if you do have some statistical data that correlates cream-with-their-coffee drinkers and White Zinfandel. Not to mention the added variable of food pairings, which can turn even the most accurate recommendation based on someone’s personality or general food preferences on its head.

“But it’s better than nothing,” some people will claim. I completely disagree. People are better off just learning what they like by trial and error, or by asking someone knowledgeable for help.

End of rant. Read the full story.