News article

Date: 20. February 2017

Find Love with Machine Learning on eHarmony
Matchmaking site makes finding love more advanced

A dating site without photos sounds like a mad idea. However, this was the reality for matchmaking site eHarmony when they began 16 years ago. Singles could fill ou a profile (back then it had over 400 questions). Matches would then be suggested... however you couldn't see what your matches looked like!

Of course, over time the dating platform improved by letting singles upload photos and making the interface more user friendly. The likes of dating  apps (Bumble and Tinder among them) of course took a lot of users away... for the operators, this was certainly aggravating.

"People do end up on those sites looking for relationships, and we see that as our challenge,"Grant Langston, eHarmony’s CEO, told

Langston states that nowadays most users are more interested in what he calls "swipe friendly" dating apps. Naturally they are a lot cheaper and very easy and quick to download. According to the eHarmony CEO, however, there is a distinct advantage that these services are lacking. That is, namely, data: they've got over a decade and a half of information on users that includes psychological, behavioural and demographic information.

EHarmony and Their Machine Learning Algorithms

Now, users only have to answers 150 questions. There is often a completion rate of 80% (which does mean that many people are actually interested. Along with this high rate there are many members who will also go for the monthly subscription to the site.

Thankfully these days the questionnaire on eHarmony isn't as long as it used to be (now there are only 150 questions to answer). 80% of singles who do it end up completing it - which is of coure helpful. Along with this, a significant number of them will also purchase monthly subscriptions to the site, allowing them to fully communicate with other members and see where things go.

The most interesting thing, however, is what is done with the data by the site. Demographic information (location, gender, age) as well as psychological details (habits, likes, interests) and even behavioural data (what actions are taken by users on the site) are all collected by eHarmony. A research team also investigates couples who met up through the site, and in-house psychologists and scientists give this information to machine learning algorithms in order to help match people who are compatible.

"That data lets us optimise the site and the process for a number of variables," Langston says. "We are about relationships, and we track the quality of the relationships we make."