It had been a long week in Charlotte investment banking and Kurt Taylor needed to unwind.

After driving to Wilmington, his hometown, Taylor and his father went to an Italian restaurant. They wanted wine with dinner and perused the wine list but none stood out. The waiter offered help. He asked about their likes and dislikes, then recommended a wine with an”oaky finish” and a 92 score on the scale of famed wine reviewer Robert Parker.

Taylor and his father tried it. They would not have scored the wine so generously.

“We weren’t going to pour it down the drain but neither of us felt like we got our money’s worth,” Taylor said.

That disappointing experience more than 18 months ago happened not because the waiter was not knowledgeable. He was. But the waiter’s tastes and preferences were different. No matter how knowledgeable the sommelier, there’s no universal language for wine. But now, there’s software.

Taylor’s Wilmington company, Next Glass, is developing software to help consumers choose wine. The software learns a user’s preferences and makes wine recommendations accordingly. It does so through science.

Wine selection meets science

When Taylor started Next Glass, he decided that any recommendation tool must start with scientific analysis of wines. Next Glass contracted the David H. Murdock Research Institute to do that analysis. The Kannapolis institute, which specializes in researching food and nutrition, used its mass spectrometers and chromatography equipment to break down wine samples. Chemical analyzers tagged and identified the components. So far, Next Glass has identified more than 22,000 components in wines. Next Glass has a patent pending on its analysis processes and methods.

The next step was to develop software to tap that data. Next Glass found a software development partner in Sean Owen, founder of London-based machine learning startup Myrrix which was acquired in July by Cloudera. Machine learning involves building systems that can learn without being programmed. What Taylor wanted was software that can learn an individual’s wine preferences and then make personalized recommendations. When Next Glass’ app recommends a wine to a user, it is recommending a wine whose scientifically identified components match with components in wines the user has already liked.

Other companies already market software that offers wine recommendations. But Taylor says their recommendations aren’t based on scientific analysis of wine, nor are they truly personalized. Currently available software uses cooperative filtering algorithms, the kinds of algorithms and Netflix use for recommendations. After seeing a user likes a particular book or movie, their software recommends movies or books based on what that consumer has seen and liked before as well as what others like and recommend. While this software can be helpful, it needs a mass number of users to provide sufficient data on which to base recommendations. And these recommendations aren’t based on any scientific analysis.

“We see our software as an objective approach to the problem,” Taylor said. “We know what you like so it’s irrelevant what other people like or recommend. I think that’s an important distinction.”

The business model

More than 100 million Americans identify themselves as wine drinkers – double those who say they are beer or spirits drinkers, according to the Wine Market Council, an industry group. While Next Glass aims to place its software in the hands of these wine aficionados, Taylor plans to make money by striking deals with merchants and restaurants.

The Next Glass software can be programmed to be location aware. For example, if a Next Glass user is at a Harris Teeter grocery store, the app can identify which wines currently available in the store are closest to the user’s wine preferences. This capability will require restaurants and stores to be clients of Next Glass and they must provide the company a list of their wine inventory. But Taylor expects they’ll gladly do so if it boosts wine sales.

Next Glass will always be free to consumers. But Taylor sees revenue opportunities in advertising or a small fee each time a wine recommendation leads to a consumer purchase. There could also be additional revenue for facilitating direct messages, such as notifications of specials.

Raising money

The Next Glass app is not yet available but the software is already starting to attract notice. In June, the company won an Elance pitch competition. The first place prize included an online meeting with venture capitalists from New Enterprise Associates and Stripes Group.

Next Glass is not seeking venture capital investment right now but Taylor is considering it for the future. He opened a fundraising round in July targeting $1.8 million. A former investment banker with Fennebresque & Co., Taylor said he already has connections to local angel investors. Proceeds will be used to recruit software developers as well as marketing staff and a laboratory technician. That round is expected to close at the end of September. The software remains in development and testing; programmers are still refining the algorithms. Taylor’s goal is to launch the software in the first quarter.

In the meantime, Taylor has spent the summer traveling. Besides meeting investors he has also visited grocery stores, wine sellers and restaurants. He said restaurants are excited because a diner who has had a good dining experience based on a wine recommendation is more likely to return as a customer. Wine retailers see Next Glass offering a new way to distinguish wines from each other. Right now, the sure-fire way to boost wine sales is by cutting prices. But Next Glass software could give retailers a way to latch on to customers’ tastes without eroding profit margins.

“They want to push more wine but they don’t know how,” Taylor said. “They see us as a tool to help build more wine sales.”