Web Service Goes Date a-Mining
Much like Netflix can suggest movies, an Internet recommendation engine called Wings points you toward dating prospects.
A computer might be able to discern your tastes in romance even better than you can.
A new dating site called Wings is trying to push the bounds of machine learning and statistical models for better matchmaking recommendations. Unlike sites that rely on questionnaires, Wings tries to understand who you are by picking up the social media bread crumbs you leave online. Among the intriguing findings: whether you have a tight-knit group of online friends tends to predict what sort of person you might like.
Launched this year, Wings turns the model of traditional online dating sites like eHarmony or Match.com on its head. Wings doesn’t ask you about yourself. It tells you. The service requires a Facebook account, which it uses to study your online social network, but it also can connect to your Netflix, Pandora, Last.fm, Twitter, and Foursquare accounts to access data like movie-rental history and places you’ve been.
All that data is fed into the service’s recommendation engine. That system combines Bayesian modeling, a type of mathematical analysis that lets computers draw inferences from huge data sets, and machine learning, where the more data and feedback the algorithm is fed, the “smarter” it gets.
The idea is that the computer’s analysis of your behavior provides a richer analysis than what you’d say about yourself. “We serve as our own blind spot in that it’s difficult to accurately answer questions about oneself without biasing toward recent experience, current mood, etc.,” says Sunil Nagaraj, chief executive and cofounder of Triangulate, the company behind Wings.
Nagaraj founded Triangulate with two other Harvard graduates. The company raised $750,000 in July to expand and improve the service.
Because the dating service utilizes a recommendation engine instead of a simple search or questionnaire, it can draw some interesting and counterintuitive correlations on what leads to successful matches. For example, Nagaraj says, the density of one’s social network turns out to be an important factor. If your group of Facebook friends tends to be more closely knit, meaning that your friends are often friends with each other, you’re more likely to match with someone who also has a tight network of friends, rather than a loose association of acquaintances.
Wings has also found that couples tend to be well suited if they have similar percentages of friends from their own country versus other countries. It matters as well whether your Netflix rental or music playlist history tends toward the mainstream or underground. And couples that have lots of overlap in the types of people they follow on Twitter tend to match well, too.
Another company, IntroAnalytics, whose engine is used by dating sites like those run by the lad magazines FHM and Maxim, also has applied recommendation technology to online dating. But it differs from Wings in that it uses the data provided by the sites to which it licenses its technology–like profile information and user browsing patterns. IntroAnalytics cofounder Gavin Potter says that this method is very effective. On one site, for instance, Potter found that 60 percent to 70 percent of users’ navigation employed the recommendations provided by his technology instead of the search function, in which people look for traits they think they want.
To be sure, science can’t wave an algorithmic wand to find the perfect mate. Recommendation engines tease out correlations from huge data sets to make smarter suggestions, not perfect predictions. And the data set is never perfect. Nagaraj concedes that Wings could be drawing from more data streams to get a fuller picture of a person, and he is working to integrate more online services with Wings.
One other problem is that people aren’t consistent in their opinions, says Caterina Fake, a cofounder of a Web recommendation engine called Hunch. It tries to tackle questions that Google doesn’t easily answer, such as, “What movie should I see?” Fake gives an example of how people change their minds: “We may predict that you will rate some movie five stars, and right after you watch it, you do give it five stars. But a month later, you might come back and rate it three stars, because it’s faded in memory.”
Still, collecting and analyzing social data the way Wings does could be a new branch in the evolution of Web services that make smarter recommendations without having to be told something twice –or even once.