The Love Algorithm Match.com thinks it has found a formula for ruling the fiercely competitive online dating business: A scientific way to find Mr. or Ms. Right.
By Susan Orenstein

(Business 2.0) – In a conference room called cuddle bug at the Dallas headquarters of Match.com, three men and a woman sit around a long table talking about lust. But this is not the conversation that the topic brings to mind. "You can do exploratory or confirmatory nonparametric statistics branching," says the group's leader, a weary-looking psychologist named Mark Thompson, between sips of a supersize soda. A visitor new to the jargon might quietly hope Thompson is alluding to some kinky practice known only to consumer-survey mavens. But no. As the meeting's focus turns to the papers that are spread across the table, it becomes clear that "comma delimited" means more to this group just now than Kama Sutra. Although some of the papers carry tantalizing scribbles--"Sexually passionate" and "Believes sex is crucial to a relationship"--most are a blur of sterile notations like "Mean: 1.4144," "Std. Dev.: 0.2477," and "N: 744." Not what you'd call the stuff of romance. ¶ Unless you're Mark Thompson. A consultant to Match.com, the country's largest online dating service, Thompson has been pulling late nights for months at Match headquarters, refining what could be the industry's ultimate trade secret: a scientific method for helping customers find true love. Thompson, a 36-year-old bachelor, has been working with his team on what he calls a "personal-attraction test." The goal, he says, is to cre-ate search algorithms that would fix up Match customers with dates who not only don't smoke and do live within 10 miles, but also have the right je ne sais quoi. Call it a Google for Love.

Don't laugh. Online dating is a serious business. The industry brings in more revenue than any other category of legitimate paid Web content--more than digital music or business and investing advice. Revenues for online dating should hit $313 million this year, up from $228 million in 2002, according to Jupiter Research.

As the most visited and richest of the sites--its $125.2 million in 2002 revenue accounts for 55 percent of the entire industry's--Match is the poster child for online matchmaking's commercial possibilities. Already it is one of the most profitable units of Barry Diller's sprawling InterActiveCorp, formerly USA Interactive. Company executives talk boldly of a time when every unattached person in America--61.6 million between the ages of 18 and 54--will have a profile on Match.com and go to parties armed with a card bearing his or her user name. "It'll feel very universal," says Tim Sullivan, Match's president. "It will be quite a natural thing to be using Match in some way."

Before all that big talk turns into reality, however, Match has to overcome a critical challenge: The customer experience is spotty. The dating scene, like a flea market, seems like one of those fragmented marketplaces truly enhanced by the Internet, with its speed, anonymity, and ability to connect vast numbers of buyers and sellers. But ironically, Match's main competitive advantage--its size--has a distinct downside. Searchers are often deluged by opportunity; it's not uncommon for female users in particular to receive dozens of come-ons. And too often the suitors aren't even close to being Mr. or Ms. Right. As one woman who tried Match in Washington, D.C., puts it, "The matches sucked." Among her friends who tried out various sites, second dates were rare. One of her friends, who is gay, searched for potential love interests on Match; among the candidates he got back was himself.

While Match gets the most traffic--for now, anyway--dozens of competitors are circling. Yahoo Personals and AmericanSingles.com, for example, are not all that far behind. Edgy niche players like Toronto-based Lavalife and New York City's Spring Street Networks are stealing younger users. As the industry matures, Match cannot remain the leader by delivering its subscribers a mass of poorly differentiated eligible singles. To succeed, the service will have to convince customers that it can really increase their odds of scoring a romantic bull's-eye. That's where the love algorithm comes in.

When Match was getting off the ground back in 1995, it was hard to imagine online dating services as anything but a niche industry. Personal ads, after all, were associated with desperate divorcees and smarmy swingers. This wasn't how users of the brave new medium saw themselves.

From the start, Match attacked this problem head-on. It was so focused on finding a mate, not just a playmate, that it was "almost puritanical," says John Pleasants, president of information and services at InterActiveCorp. (Match still screens out postings that contain loaded words like "discreet," which could signal a search for an affair.) "We decided early on that we would introduce the category and not just Match.com," says Trish McDermott, whose title is "VP of romance." Being single on Match, the company aimed to say, is nothing to be ashamed of.

As a pioneer, Match can take at least some credit for turning online dating into a better alternative than singles bars and personal ads in newspapers and magazines. But much of the change stems from the Internet itself. Because they charged by the word, print personals had a transactional quality, observes Rufus Griscom, chairman of Spring Street Networks, which manages online personals for Salon.com, the Onion, and other sites appealing to the urban sophisticate. "They were about an SWF looking for a round, curly-haired man," he says. Online ads, by contrast, offer plenty of room for self-expression.

As online dating edged into the cultural mainstream during the past eight years, numerous sites sprang up, catering to niches from Ivy Leaguers to Christians to tall people to aging hippies. Match has gone for the broader play, gobbling up smaller competitors like Kiss.com and Udate, and striking key deals with AOL and MSN that provide a steady flow of traffic. In May alone, Match clocked 5.1 million unique visitors. (AmericanSingles follows at No. 2 with 3.6 million.) Not all of these are paying customers, however. Visitors are free to post their own profiles and search other profiles, although only subscribers can send messages to prospects. Match's paid subscribers total 766,000 (most of its rivals, including Yahoo, do not disclose subscriber or revenue figures).

Maintaining a brand identity in this crowded field is a continuous challenge, and sites strive to set themselves apart any way they can. Both Match and Yahoo charge subscribers by the month, while Lavalife and Spring Street charge by the transaction, meaning that users pay for each e-mail or instant message they send. As for product positioning, Match, AmericanSingles, and Yahoo all play to the middle of the road; they're like meeting at a barbecue. Lavalife and Spring Street, by contrast, are like going to a neighborhood bar--a pretty risque one, in Lavalife's case. One of its communities, Intimate Encounters, lets users post nude photos and search by their preference for "conventional sex" or "dominance and submission," among others.

For all that, the bigger sites tend to be more or less interchangeable. They all give users the chance to post descriptions of themselves, post photos, and connect via e-mail or instant message. The competition is so relentless that when one service introduces a function, such as the ability to keep a list of "favorite" profiles, you can count on seeing the same option pop up almost immediately on rival sites.

Neither Match nor its competitors, however, have yet succeeded at providing the service that customers want most--the ability to reliably find a good match. While a few smaller sites, like eHarmony.com and eMode.com, have already come out with personality tests, the mainstream sites search only on blunt criteria like age, income, and hobbies. Such measures tell users next to nothing about whether the people who turn up will make decent dates, let alone long-term partners.

The opportunity for a site that can truly help its users find real matches is clear. "The next chapter of personals is about to begin," Pleasants says. "For me, that means product innovation." The killer innovation, of course, would be Thompson's love algorithm. If it works.

Mark Thompson has been listening to doubters since he was a doctoral student at the University of Indiana, when he first became fascinated with the intersection of artificial intelligence, advanced math, and human behavior. He sighs at the oft-repeated objection that some human mysteries are beyond the reach of science. "A lot of what we think is elusive is very measurable, tangible, and predictable," he says. He cites the example of emotions, which seem to well up out of nowhere, "but have clear biochemical underpinnings." It's only logical, he insists, that certain traits influence attraction. "A lot of that spark and magic is an expression of parameters we're now getting a handle on," he says.

The notion that attraction can be predicted has some basis in research and is not as crazy as it might sound. Studies have shown, for instance, that people tend to be attracted to those who resemble them in socioeconomic class, education, and family background. "You're drawn to things you esteem," says Pepper Schwartz, a sociologist and author of Everything You Know About Love and Sex Is Wrong. "How many upper-middle-class girls actually want a motorcycle guy who refinishes cars as their heartthrob? Not too many." It's a long way from there, of course, to determining which small subset of non-motorcycle-riding, non-car-refinishing guys is most likely to make any particular upper-middle-class girl's heart go pitter-pat. But those are the kinds of details Thompson's personal-attraction test aims to address.

On a Friday morning in April, Thompson demonstrates the test at Match headquarters. The phrase "Usually I am ..." flashes across the computer screen, followed by a menu of choices: "Creative." "A little nervous." "Open to a one-night stand." The options change every few seconds, giving him little time to overthink his answers. The test, meanwhile, is supposedly making adjustments with each answer to better measure the respondent's extremes. Just how averse to intimacy is Thompson? When he clicks on "Getting close can make me nervous," it feeds him "Fear commitment" to see if he'll click on that too. (He doesn't.) Thompson isn't promising to find you your perfect match. "Maybe we won't be really good at predicting who to marry," he says, "but this is a good way to get you to date the right people." He claims that as his test collects untold gigabytes of data about users, it'll become better at doing what most people can't: articulating the traits that make them more or less likely to be attracted by or attractive to someone else. It apparently works well enough to have won over one important backer--Match CTO Mike Presz. "I'm a tech guy," he says, "and I'm shocked at how accurate this test is."

Thompson's test has been live on the site for four months, but so far all it does is produce a personality profile for the test taker. (A beta version that matches testees with one another is in development.) Still, visitors to Match.com apparently enjoy investing 10 minutes in Thompson's questionnaire to be told, "You have an insatiable curiosity" or "You can get so caught up in a conversation that you talk more than most." Match's brass is delighted. "More than a million people have already taken the test. That's amazing," Pleasants says. "People are fascinated with this stuff."

While Thompson refines his analysis, Match continues trying to improve its product in other ways. In June the company unveiled a redesign, largely aimed at making the site more engaging, flirtatious, and easier to use. "We want quicker ways to get visitors into the game," says Match's VP for marketing, Melanie Angermann. The company started offering features that some of its competitors already had, like a virtual wink--in the form of a smiley face--that a user can send with one click to a prospective date to indicate romantic interest. And, ever sensitive to the stigma problem, Match dropped the word "singles" from the homepage--at the suggestion of no less than Barry Diller himself.

Meanwhile, Match's competitors are working equally hard to make their own services irresistible. Yahoo's general manager for personals, Katie Mitic, says customer surveys showed that users wanted to be able to put voice and video in their profiles; Yahoo started offering those features in January. Spring Street Networks, meanwhile, tries to turn profiles into entertainment by posing sexy essay questions.

Few have gone quite as far as Match, however, in the quest to create the Google for Love. (eHarmony and eMode are the exceptions.) Neither Yahoo nor Spring Street offers a personality test. Others have considered tests but have not gone beyond the talking stage. Joe Shapira, co-founder of AmericanSingles.com, has hired a Harvard professor to cook up something at once informal and sophisticated. "It's going to be similar to the zodiac, only based in psychology," Shapira promises vaguely. Matchmaker--owned by Terra Lycos--has hired a group of psychologists to add feedback on "what styles go together," says Meredith Hanrahan, who runs the site. "We want to have an algorithm that keeps the door open for human nature," she says. "It's very complicated."

It's easy to dismiss what Thompson is attempting. Several psychologists and sociologists who took his test for Business 2.0 were skeptical. Susan Sprecher, a sociologist at Illinois State University and editor of Personal Relationships Journal, argues that "no matter what a test can indicate, there are so many factors that determine whether two people like each other in an intense way, from communication style to the way they smell. Personality is just one factor."

Thompson doesn't disagree. Indeed, he's working on another test that aims to assess a person's taste in strong chins or curvy figures. Unfazed by the skepticism, he argues that his personality test captures nuances like communication style. His computer models are designed to learn by collecting feedback on whom users contact and how they rate their dates. Giddy about the windfall of research subjects Match can provide, Thompson says his ultimate vision is to be able to tell someone, "Hey, you're introverted and say you like extroverts. But most introverted people actually get along with people who are a little more introverted than you say you want."

While Thompson is a true believer, Match president Sullivan knows the danger of overpromising. He says Thompson's analysis is no substitute for "that intangible chemistry." But he has good reason to sound excited anyway. Self-assessments like Thompson's help broaden Match--they even appeal to married people. "Match isn't going to win by being big like eBay," Sullivan says. "We want to be a destination, not just a transaction service."

But that's only one reason that Match is aggressively promoting Thompson's test, describing it as "scientific, accurate, fully customized." Diller himself even plugged the test in a recent earnings call. Why? Because in the end it doesn't really matter whether the algorithm works. Online dating is a bit like the dieting business. It thrives by making a plausible promise, not necessarily by delivering. After all, when a person finds love, he or she is lost to Match.com. As Griscom of Spring Street Networks points out, "The flirting business is a lot bigger than the matchmaking business."

Thompson's system provides added encouragement to keep looking--and paying. Jay Novella is a perfect example of a customer that Match doesn't want to lose. A member for a year, he knows the high of getting new e-mail, but also describes how "day in and day out, it's the same thing, looking at all the pictures." At this point, he's ready for something new to offer hope.

Which brings us back to Thompson, toiling late hours in conference rooms like Cuddle Bug. In some ways, he's on the same kind of quixotic search as someone looking for love. It's a good thing too. For what if, by some improbable feat of nonparametric statistics branching, Thompson succeeds and unlocks the mysteries that have stumped poets through the ages? Deep in its heart, Match can only hope it doesn't happen.

Susan Orenstein (sorenstein@business2.com) is a senior writer at Business 2.0.