It was a cheap stunt, but effective. Leo McCloskey lined up a dozen vintners from the Robert Mondavi Winery in Oakville, Calif. in the corporate tasting room and asked them to rank the contents of ten bottles. When they'd finished swirling, spitting and recording their verdicts, McCloskey pulled out a sealed envelope. Inside was a list of the same wines ranked by a computer system he'd programmed to determine quality. The list matched dead-on the Mondavi judges' winner and loser.
The gimmick was vintage McCloskey. He's used it dozens of times to drum up business for his Sonoma, Calif. wine consultancy, Enologix. "We have mathematically defined wine quality," he says.
Which is something akin to reducing love to an algorithm. Just how it's done is a closely guarded secret. McCloskey, 53, allows that he breaks wine chemistry down to such molecules as the anthocyanins that give wine color, the catechins that give it some bite and the esters that impart aroma (and thus taste). By tracking such elements he can tell vintners how to blend grapes to get the best taste scores. He pools the data inside a cluster of Unix servers in his office and sells online access to clients. For as little as $600 a month, depending on how many cases they produce, they can nab data on the quality, aging potential and treatment of their grapes in order to manage their crops.
McCloskey says his predictions have a fair degree of correlation with ratings by such authorities as Robert Parker and Wine Spectator. "We can tell the winemaker what he's got fermenting in the barrel long before the market knows," says McCloskey. In December he was already crunching results for cabernets from the 2003 harvest.
The idea for Enologix dates to 1984, when McCloskey and his wife, Susan Arrhenius, were working on their doctorates in chemical ecology at UC, Santa Cruz. Acting on a hunch that the chemicals in wine might be linked to its price and quality, McCloskey asked the head of the math department to build a statistical model. He spent the next decade working as a vineyard manager and consultant, collecting data from clients' cellars and testing his theory. He even patented a piece of his analysis for measuring alcoholic esters, called "enzymic assay for acetate ions."
A crucial test
By 1990 he was ready to put his model on the line. He approached the late Richard Graff, one of the American winemakers who, in 1976, nosed out the best French white burgundies-- Seabiscuit-style -- in a blind tasting held by Parisian judges. Graff, who had an irreverent streak and great connections, arranged for McCloskey to test the relationship between price and chemistry, in the vaunted cellars of Château Lafite Rothschild in Pauillac. The results: In a dozen benchmark vintages, from $200 to $800 per bottle, changes in chemical concentrations shadowed prices just as McCloskey had predicted. He was in business.
Today Enologix stores the chemical fingerprints of 50,000 wines, plus data on flavors of oak barrels, climate and thousands of samples of vineyard soil. McCloskey linked four robotic grape and wine testers and a battery of chromatographs to the database. They assay murky samples of wine so young that their features can't be discerned by the most effete palates and provide data on quality and aging potential; they can simulate blends, even pick the right kind of oak barrel. His staff of 14 includes an analytical chemist and a couple of chemistry Ph.D.s, as well as a chief information officer. "We can make great wines in an orderly fashion," McCloskey says.
Great-- meaning wines garnering high marks from the fearsome Robert Parker. This former Maryland lawyer, who has insured his nose for $1 million, is the self-appointed custodian of American wine taste. Since the U.S. is the biggest wine importer, his 100-point rating scale effectively makes the world market. "Declining Parker scores are the kiss of death for many wineries from Tierra del Fuego to the Langhe Hills" in Italy, says wine writer Lawrence Osborne, author of The Accidental Connoisseur. A 10-point drop near the top could mean as much as $100 less per bottle at retail. A bad rating coupled with a poor harvest can be exceedingly painful to a vineyard.
McCloskey claims that Parker's nose can be predicted and plotted. "It already exists today in the barrel," he says. There are believers among McCloskey's 75 clients, which include Diageo's Beaulieu and St. Francis; the expensive boutiques Diamond Creek and Ridge Vineyard; and Kobrand, the large French distributor of U.S. wines. "I can tell you from his data that I've looked at and from his work with other wineries that he's got a very high correlation," attests William Leigon, president of Hahn Estates winery and a McCloskey client. "It tends to be plus or minus five points"-- which, to the uninitiated anyway, sounds like a lot. Then, again, the reviews of vaunted critics can vary as much as six points on the same bottle of wine.
'Like the word processor'
Leigon contends that McCloskey helped save him from ruin. In 2001 the 1,000-acre Hahn vineyard was producing way too many grapes and making wine that everyone agreed was awful. Using Enologix data for comparable vineyards and a grape varietal as a yardstick, Leigon reduced the crop from 5 tons per acre to 3. Last year he got intense cabernet grapes, just as he wanted, and sent the fruit to McCloskey for flavor analysis. To time his harvest he relied on flavor data to determine ripeness, in addition to measuring sugar content, which may fluctuate with weather spells. When McCloskey's computer run indicated that the grapes fell within 5% of the desired Bordeaux-style cabernet Leigon wanted to make, he sent out pickers.
Leigon says that his business has tripled since 2001, shipping 260,000 cases last year and shooting for 400,000 cases in 2004. His critic ratings went up 10 points. Although he needs to score consistently over 90 for at least three years to be able to raise prices, Leigon says he was able to eliminate heavy discounting.
Diamond Creek, whose vintage cabernet sells for $400 a bottle, is also a believer. Enologix is helping its winemakers select and blend the best grape juice at the very early stage so the premium crop will go into the best barrels. "The palate can tell you that the wine is tannic, but it can't tell you what the number is," says Diamond Creek's founder, Al J. Brounstein. "Enologix is a tool-- like the word processor. It gives us more power, but ultimately the winemaker has to make the wine."
McCloskey, meantime, says he is making money, netting $650,000 pretax on revenue of $1.6 million last year. His goal is to turn Enologix into the Dun & Bradstreet of winemaking. "The trading model, whether it's cars or wine, says that the consumer doesn't believe the producer," says McCloskey. But will they believe the guy who turns one of the most sensual experiences into binary code?