I came across an advertisement the other day for a company called Bright Cellars - a subscription wine service started by two MIT grads that sends you wines based on your 'palate quiz'. After a fairly short quiz, half of which has nothing to do with taste, you get four recommendations that they'll send to you for $15 a bottle plus shipping.
Not being familiar with any of the vineyards recommended, I can't comment on the quality of their palate matching. But I am curious about their model, which seems to be a common one popping up. After a few questions can you really match my palate and feel comfortable recommending based on that? Is it really possible to break down wine preference into individual attributes that you try to suss out based on the type of juice or tea I like? I'm skeptical.
After so many iterations of recommendation algorithms - both in food and wine - I'm curious why there don't seem to be models based more around "people like you also liked ...". If it works for Amazon's purchase recommendations, why not here? I'd much rather see restaurant reviews on Yelp from people who enjoy restaurants I do. If you're more of a steak and burger person, your recommendations aren't as well suited to me since I don't eat beef. The same seems like it would apply to wine.
An app like Vivino already knows what else people who liked XYZ wine also rated highly. I'd be curious to see what a recommendation like that looks like and how accurate it might be.
No comments:
Post a Comment