Sam: Beer, Norm?
Norm: Have I gotten that predictable? Good.
Brewers know much more now about the compounds responsible for aroma and flavor in beer, and scientists have a better idea of how our “most enigmatic” sense—smell—works. Because there are thousands of choices today, doesn’t it seem as if there should be an app that knows what beer we want to drink next?
You could call it “Pandora for beer.” Oh, wait, that’s how every new recommendation app gets described these days. Some take a scientific approach similar to Pandora when collecting data, but others are more like Spotify, which uses collaborative filtering. Netflix and Amazon are the best-known services using that technology, which is based on the assumption that if two users enjoy overlapping sets of movies or books (or songs or beers), their tastes are probably similar.
In contrast, feature-based systems begin by measuring specific attributes of songs (or beers). Pandora created the Music Genome Project and employs musicians to catalog key elements of every song in its database. A new app called Next Glass, which launched earlier this year, continues to build its own “cellar genome” of beers and wines. High-tech machinery in the laboratory at Next Glass analyzes the contents of each bottle, mapping the chemical makeup of individual beers and forgoing any effort to quantify a beer based on its descriptive characteristics. The app uses an algorithm to blend the attributes of the beer in question with what it knows about each consumer and assigns a number predicting how much the consumer will like that beer.
“The only way to do this is science,” says chief operating officer Trace Smith. Some drinkers obviously gravitate toward that science. It took BeerDeCoded only three weeks to reach its goal of raising $10,000 on Kickstarter to map the DNA of ingredients in 1,000 beers and begin to build a tree of them. Project director Gianpaolo Rando at the University of Geneva admits it is only the first step toward creating a “genomic sommelier” and eventually a smartphone app. The project’s press kit also acknowledges the larger challenge that “taste is a perception and not an intrinsic quality of beer (as in DNA).”
There’s the rub for any app creator, no matter how deep the data or how well conceived the formulas to generate recommendations. None has found a way to map the human olfactory genome, and as a result of genetic variation, two people may perceive the same odors differently. “You end up with a barcode situation, whereby each individual has a slightly different barcode,” says geneticist Doron Lancet of Israel’s Weizmann Institute of Science.
So while these apps often provide excellent recommendations, they aren’t necessarily as reliable as Pandora or Spotify. One of them, BeerMapper, uses a hybrid approach in order to improve its success rate, although it still can’t account for genetic differences. It draws its database from the user-generated reviews at ratebeer.com and, much like other collaborative filtering systems, looks for like users. It also extracts the adjectives used to describe each beer—“We feel the best way to analyze how beer tastes is crowd sourcing,” says Joe Sheahan at BeerMapper—and that provides a feature-based component. BeerMapper generates a visually striking personal beer flavor map as well as recommendations.
The map may tell consumers who don’t want to be as predictable as Norm Peterson something about themselves—that’s one way any app may be of use beyond simple recommendations. But it’s still possible, for now, that the best beer discoveries are the ones made without a smartphone in hand.