As part of Travel + Leisure's new technology and innovations vertical, The Grid, we're rolling out a video series all about the ideas, products, and people shaping the future of travel. Next up: Entrepreneur Philip James on the algorithms of taste.
Penrose Hill, a wine commerce site, has been innovating the winemaking business for the modern age. One of the ways the company — and its founder Philip James — is doing that is by implementing algorithms and surveys that better-predict what wines its customers will love.
The key to doing this well, James recently told T+L, is by pulling more useful information. He says most recommendation tools use surveys that ask questions that, well, just don't matter. For example: other wine recommendation sites like to ask how its users drink their coffee, reasoning that coffee tastes indicate wine tastes. James insists that these tidbits of information are red herrings when it comes to wine.
“I used to drink frothy, sweet lattes because I loved the mouth feel and the sugar. But for dietary reasons, I switched to black coffee,” says James. “My taste in wine has nothing to do with that.”
It's a cop-out for the company, James says. “It’s easy for a consumer to say how they [drink] their coffee, if they prefer sweet food or salty food, if they like apples or berries. But those questions don’t really have any relevance on the kind of wine you like.”
That’s why Penrose Hill and its wine club Firstleaf (a Time Inc. partner), uses an algorithm sort of like the radio and music recommendation site Pandora.
Pandora starts out with a "seed" artist or song to populate a playlist. While listening, users can rate songs up or down. Pandora doesn’t generate playlists based on a questionnaire in which you reveal your ideal beats per minute or the kind of lyrics you like, instead it uses an algorithm to find that information based on the songs you like. It's way easier for somebody to give a song a thumbs up than it is to pick out a syncopation pattern.
“We do the same in wine,” James says. “It’s so much easier to watch how a consumer reacts to a specific bottle.”
For more on how algorithms can predict taste, check out the video, above.