Nearly all of us have spent time pondering what makes two people a compatible couple, but few have put as much time, money and Ph.D.s behind this question as Match.com.
When the site started in 1995, online dating was an obscure (and somewhat dubious) practice. Now one in every five new relationships starts with online dating, according to a study it commissioned.
Matchmaking has never been a simple business, but when you add 16 years of data from a site that Compete.com estimates has about 14 million unique visitors per month, there’s endless opportunity for complexity. About three years ago, the company decided to delve into that data to help improve its matches.
It brought on current VP of strategy analytics Amarnath Thombre to head up the charge. Since then, he and a team of 12 have been hard at work developing an equation (well, hundreds of equations) for successful match recommendations. It’s a uniquely challenging pursuit.
“It’s easy to predict who likes The Godfather, but in this case, the Godfather has to like you back,” Thombre says.
The team’s first algorithm, despite the challenge, seems to be working. The first changes it made resulted in a doubling of “yes” matches on the site.
We sat down with Thombre and Match.com President Mandy Ginsberg to learn more about the math behind improving matches. These are the four main components of the equation.
1. What You Say
When you sign up for Match.com or pretty much any other dating site, you fill out a survey about yourself, your preferences and what you’re looking for in a partner. The most basic part of Match’s suggestions result from simply matching people with characteristics that they both said they liked.
2. What You Do
What people say and what they do don’t always match up. Of the women who say that a partner’s desire to have kids is a “must have” criteria on Match.com, for instance, 57% still have emailed men on the site who don’t want to have kids.</span> 51% who say the same about a partner’s income have ended up breaking their criteria.
Because of this dissonance, Match.com’s recommendation engine considers what “must have” criteria you will compromise on and when you make those compromises. Maybe you say that you don’t want to date someone who has been married before, but you consistently email divorced men as long as they want to have children. Match.com’s algorithms might begin to include some men who fit into that category in your recommendations.
“Instead of trying to create the perfect algorithm, we try to create the perfect algorithm for you,” Thombre says.
3. What People Like You Do
In addition to tuning into your behavior to decide who you might like, Match.com also tunes into the behavior of people who are like you. The site looks for other users whose behavior mirrors your own (i.e. they have communicated with the same people who you have). It considers people who your behavior twins have interacted with to be more likely matches for you.
It’s another way of determining what you like even if you can’t articulate it.
“We don’t know exactly what it is,” explains Thombre. “It could be their sense of humor, it could be the way they smile, the way that their facial structure is — all of these abstract patterns … we just try to identify the pattern and find more people like that for you.”
4. A Bit of History
In 16 years of business, Match.com has collected a mass of data that any human behavior researcher would envy (competitor eHarmony has been at it for 14 years). It uses that data to make some guesses about who you like before you even rate your first match.
The site’s math team has documented hundreds of correlations. Here are just a sampling:
- Republicans are more likely to reach across party lines to contact a Democrat than vice versa.
- A woman who smokes daily is less likely to email a man who doesn’t smoke.
- Women of “other” religion show positive propensity to email a Jewish man.
Neither the matchmaking company nor its algorithms claim to have the answer to any of these questions.
“We don’t pretend to know who is right for you, but we use mathematics to quickly learn how your complexity shows itself on the site,” Thombre says. “We are like a bartender who is always observing particularly which types of people are talking to each other and hitting it off.”
Image courtesy of iStockphoto, adventtr
More About: algorithms, dating, love, Match.com, online dating
For more Social Media coverage:
- Follow Mashable Social Media on Twitter
- Become a Fan on Facebook
- Subscribe to the Social Media channel
- Download our free apps for Android, Mac, iPhone and iPad
Continue Reading...
