Artvoice: Buffalo's #1 Newsweekly
Home Blogs Web Features Calendar Listings Artvoice TV Real Estate Classifieds Contact
Previous story: Getting Away With Murder
Next story: Last Call for Romance

Crunching Data for Love

How Dating Sites Find Your Match

Whether looking for love or just a date, thousands of competing dating sites and apps promise everything from a huge selection of potential mates, to personally tailored matches for maximum compatibility. Online dating companies would never make public the details of the algorithms they develop to decide who to match with whom, but information about the concepts they use is widely available.

Matchmaking algorithms analyze vast troves of information to create matches. A service can get information from you, in the form of questions, tests, and assessments, or gather information about you, by observing your behavior on the site.

Anyone who has ever tried sites such as OkCupid,, and especially eHarmony, has encountered the endless barrage of questions and tests. The questions cover preferences in potential mates and include religion, politics, lifestyle choices and personality traits. Some sites make the questions optional. OkCupid recommends answering at least 250 questions to match you with compatible people, with the average user answering about 400. Many people who have answered over a thousand or more report receiving fewer matches, but the matches they do get all score over 90% compatibility. The exact number of questions on OkCupid is unknown–it’s rumored to be over 30,000.

Users can rank the importance of a potential mate’s answer to each question, i.e. smoking–not important or very important. The more similarities between two people the higher their compatibility score. With the exception of Tinder, most online dating services ask members extensive questions.

The situation at eHarmony is different from other sites in that it requires users to answer 400 questions before even being allowed to join. This pre-screening assesses 29 “dimensions” of compatibility. However, eHarmony also uses these answers to determine whether or not one will even be allowed to join and the site rejects about 20% of applicants without explanation. Eharmony has faced numerous lawsuits over its refusal to match gay applicants. Your chances of being accepted diminish if you are not Christian, with some claiming the site automatically disqualifies all atheists, regardless of how you answer any of the other 399 questions. Artists and creative types also have slim chances. Since eHarmony caters to people seeking marriage particularly, they also reject anyone who is married or separated. If you are under the age of 60 and have already been married 4 or more times, there is no chance of getting in the club. By rejecting people, they are curating a membership base that is highly compatible with their matching system. At least they let you down easy before wasting your money.

Behavior analysis has been increasingly incorporated into dating site algorithms to see if you are actually attracted to the type of people you say you’re looking for. For example, if a user states a certain age bracket as a mandatory requirement for their mate, but frequently views profiles of people outside that age bracket, the computer infers that they would probably be open to meeting people outside their stated age preference.

Other behavioral patterns factored in are frequency of site usage and interactions with other users. If you exchange a lot of messages with person A, but the conversation with person B is short the computer presents you with more matches similar to person A. Perhaps the computer observes that a female user hardly ever responds to messages from men under 5’8”. It will conclude that she prefers tall men and factor that in to her matches.

Tinder is the most striking example of behavior analysis. 83% of its users are 18-34 and the app asks no questions whatsoever about one’s preferences or dislikes, and has minimal room to describe yourself in the profile. There is no way to search or choose who you see. The app presents photos of other users. Swipe left for ugly and right for hot. If two people swipe right on each other, they can start messaging. It seems simple, maybe a little too simple, but under the hood, Tinder is anything but. The exact details of Tinder’s algorithm are kept secret, but there seems to be a general consensus that users are scored in three major categories.

The attractiveness score is determined by measuring the percentage of people who swipe right for a given person. It isn’t just about quantity though. Being liked by someone with a high attractiveness score will benefit one’s own score more than being liked by someone less attractive. In addition, they compare you to people like you and can anticipate whom you might find attractive by looking at the patterns of people similar to you, like maybe tattooed women swipe right on guys with motorcycles in their photos

Tinder also scores each user on activeness, or how much time they spend using the app. Basically, the app rewards those who use it a lot with more profiles to view, and shows fewer profiles to those who only log in rarely.

Perhaps the most interesting score in the Tinder algorithm is the pickiness assessment. Swipe left too much? Tinder will think you’re too picky, and you risk running out of profiles to view. Many have been in Tinder limbo, being told by the app that there is nobody to view in their area, but seeing their friend sitting in the same room get a steady stream of photos to view. To prevent this, some people, guys mainly, swipe right on everybody. Although this can produce more matches initially, the computer will know it’s being gamed if conversations are not occurring between you and your matches. It knows you didn’t really think that person was hot. It also knows that you don’t engage with people much. People use Tinder to meet other people; someone who doesn’t try to meet anyone is on the road to Tinder limbo. Instead of rejecting undesirable users like eHarmony, Tinder simply makes the app less appealing to use for those who are unappealing to their users.

Algorithms can study millions of users over the course of years and find large trends over time. This gives the program something to work with when a new user joins, and the computer is expected to immediately generate matches. is well known for studying members’ behavior in great depth. Having been one of the first dating sites on the internet, it has accumulated a database of human behavior that even Facebook and Google should be jealous of.

Interestingly, OkCupid conducts numerous statistical studies on their user base and publishes it all in their OkTrends section of the site. The trends that are revealed are both eye-opening and at times, comical. For example, according to OkTrends, 40 year old iPhone owners have had an average of 16 sexual partners in their lives, whereas Android users average about 10. People who use Twitter every day have shorter relationships on average than everyone else, and are twice as likely to masturbate on any given day. It goes on and on. No stone is left unturned.

Are people finding love more often when math does the matchmaking? Quite possibly. Of all couples that married in the past year, 17% met in online dating sites. There has also been evidence suggesting that married couples that meet online have longer-lasting and more fulfilling marriages. According to a study commissioned by eHarmony and published in Proceedings of the National Academy of Sciences, 35% of couples that married between 2005 and 2012 met online through dating sites, Facebook, Craigslist, etc, and were marginally more likely to stay married. Only 5.96% of couples who met online were divorced or separated, compared to 7.67% for those who met in “real life.”

Online dating has long since shed the stigma it had in its early days. With 42 million people in the U.S. having tried it at least once, it will be part of mainstream culture for the forseeable future. Perhaps online dating will evolve to become not just another way to meet people, but a superior one.

blog comments powered by Disqus