:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/63701815/wine-scanning.0.1462687000.0.jpg)
Every technology company has its origin story. This one starts with a bad bottle of wine.
Kurt Taylor and his father, George, were having dinner at a Wilmington, N.C., restaurant and ordered a bottle of wine on the suggestion of a sommelier who knew them well. They hated it.
Their subsequent conversation about finding a better way to offer wine recommendations was the genesis of Next Glass, an app launching Thursday in Apple’s App store that bases its recommendations on a scientific analysis of some 20,000 bottles of wine and 4,500 bottles of beer.
Next Glass asks users to rate beer or wine through a series of yes or no Tinder-like swiping cards; the software matches up a person’s preferences against an ever-expanding database of booze. Using what’s known as a high-resolution mass spectrometer, Next Glass analyzes the thousands of chemical compounds that make up a glass of beer or wine to determine which ones impart the flavors you like.
“Not every molecule gives a beverage its flavor, but thousands do, and you can’t cherry pick — you have to run the test for the whole spectrum, whether they factor into taste or not,” Chief Operating Officer Trace Smith said.
That allows Next Glass’s algorithms to compare the “hoppiness” of one brew — say, a Firestone Walker Pivo Pils — to a Pilsner Urquell or a Budweiser.
“We’re learning about what chemically matters to them, what they like and do not like,” Smith said.
When browsing the liquor store or supermarket, the user points the smartphone at a bottle and launches the app; the camera automatically reads the label and the app generates a 1-100 recommendation.
Smith said that in tasting trials, Next Glass has shown 96 percent accuracy.
This article originally appeared on Recode.net.
